THE VIBRATION ANALYSIS HANDBOOKTABLE OF CONTENTS CHAPTER ONE: INTRODUCTION TO MACHINERY VIBRATION Theory of Vibration Harmonic Motion . Periodic Motion Random Motion The Relationship between Time .... Frequency Amplitude Measurement Sources of Frequencies ....... Generated Frequencies . . Excited Frequencies ; oe Frequencies Caused by Electronic Phenomena. Forcing Function . . oe Combinations of Machine Problems .. Mixing Frequencies . Electrical and Mechanical Relationship Time and Frequency Domain . Relationship between Vetocty, Displacement, and Acceleration Units of Measurement . 5 : Relationships... Ways of Measuring Vibration . Relation between Diameter, Speed, and RPM How To Determine Machine Speed in FPM from the Vibration Data Conclusion and Efficiencies . eres CHAPTER TWO: TIME AND FREQUENCY ANALYSIS TECHNIQUES Introduction . Basic Physics . . Single Frequency Single Frequency with Harmonics Clipping ..... : Square Wave .. Natural Frequencies ..... . : Multiple Frequencies - Linear Systems. High Frequency Riding a Low Frequency . . Multiple Frequencies - Nonlinear Systems... . Amplitude Modulation... . ‘Sum and Difference Frequencies : Pulses . .. Frequency Modulation | Conclusion . 41 35 36 52 52 55 55 55 56 37 61 : 67 68 ixCHAPTER THREE: HARDWARE AND SOFTWARE REQUIRED FOR ACCURATE Hardware . Data Collection Software .. DIAGNOSTICS Introduction... Personal Computer ..... Real-time Analyzer Printers . .. Transducers Displacement Transducers Velocity Transducers ........ Accelerometers. . : Pressure Transducers Microphones . Once-Per-Revolution Markers. Multiplexer Gauss Meter... . Introduction. Group 1, Toolbox Software . Signal Analysis Program .... . Vibration Calculation Program . Resonance and Deflection Calculator (RADC) Bearing Calculation Program ....... Gears Program .... Roll Ratio Program and Rusch Chart. Group 2, Machine Doctor (MACHDOC) MACHDOC . Polar Plot Time Plot .... Balancing .. Diagnostic Database Diagnostic Modules Roll Quality Assurance Program 109 (Groups eeeereee ree eeee sees 110 Group 4. ; 110 CHAPTER FOUR: ACCURATE EVALUATION OF MACHINERY CONDITION Introduction m1 Theory . m1 Calibration 115 Frequencies Generated 116 Data Collection . . 117 Transducer Selection... 118 Continuous Monitoring ....-. ug Common Problems .. ; 19Imbalance ...... Bent Shaft . Soft Foot .. Misalignment Looseness ...... : Bearings Loose on the Shaft . . a Bearings Loose in the Housing ............0..++ Common Forms of Looseness . . : Noise Diagnosis of Looseness « Resonance . : Rubs ..... Problems That Cause Pulses Oil Whirl... Analysis of Electric Motors. Motors Out-of-Magnetic Center . Broken Rotor Bars Turn-To-Turn Shorts in Windings - ee Vibration Problems in Synchronous Motors ....... Siren Effect ..... : a Solo Data on Motors Steam Turbines ........ Pumps ... 50000 Impeller Contact . Starvation Cavitation Compressors .... Fans... Special Tests . : Startup/Coast Down Data. Bump Tests . tieeeeee Noise Recording .. . ‘Synchronous Time Averaging (STA) . Relative Motion Measurements (RMM) CHAPTER FIVE: ACCURATE DIAGNOSIS OF ANTIFRICTION BEARINGS Introduction . Data Collection ..... Transducer Selection ........ Generated Frequencies . Fundamental Train Frequency .... Ball Pass Frequency of Outer Race Ball Pass Frequency of Inner Race Ball Spin Frequency . Application of the Bearing Formulas Outer Race Analysis Inner Race Analysis Ball Spin Frequency Analysis. : Fundamental Train Frequency Analysis .. 167 167 - 167 167 168 = 169 -170 +170 172 175 176 176 177 xiVCI Bearing Calculation Program Bearing Defects .. tees ; Raceways Outer Race . Inner Race . . Outer and Inner Race ‘Amplitude. Modulation of Ball Pass Frequency Rolling Elements, Balls, and Rollers. Cage... . pebeE Multiple Defects... Progressive Bearing Failure Defect Severity . . vee Bearing Behavior... Inner Race Defect Length . Defect Length Calculation . Deep Fatigue Spall vs. Shallow. w Flaking . Problem Sources . . Acid Etching ... Fluting Inadequate Lubrication Looseness .... Bearings That Have Excessive Internal Clearance . Bearings That are Turning on the Shaft . Bearings Loose in the Housing . Testing for Bearing Frequencies A Conclusion CHAPTER SIX: ACCURATE EVALUATION OF GEARS xii Introduction Data Acquisition How To Take Data Transducer Selection . Gear Vibration Theory... Evaluation of Gear Ratios Factoring . Gearmesh Frequency . Fractional Gearmesh Frequency Hunting Tooth Frequency Planetary Gears... Digression ...... End Digression Gear Life Expectancy Amplitude Modulation . . The Gears Program .... : ‘The AGMA Quality Number System . Gear Problems and Causes . Eccentric Gears Meshing Gears That Have a Common Factor and One Gear Is Eccentric . 17 178 178 178 181 185 185 186 189 189 193 196 196 197 201 203 - 208 208 210 214 » 216 . 216 219 220 + 220 225 225 2225 226 1.27 . 27 : 27 228 229 230 231 . 233 236 + 242 = 243 - 244 + 245 - 246 - 246 . 246Gears That Do Not Have a Common Factor and One or Both Gears Are Eccentric . wee - 257 Gears That Are Out-of-Round or Have Several High Places + 262 Gears Installed on a Bent Shaft : ; a + 265 Loose and Worn Gears . + 268 Misaligned Gears . : = 269 Backlash Problems or Oscillating Gears. = 270 Broken, Cracked, or Chipped Teeth .. = 274 Conclusion... . . = 285 CHAPTER SEVEN: ANALYZING AND SOLVING PRESS ROLL AND NIP PROBLEMS Introduction .........2.... . - 287 Vibration Theory of Rolls In Nip : = 288 Hardware . + 288 Software + 289 Synchronous Time Averaging - + 289 Hardware Setup . +292 Dynamic Measurement of Rolls. 294 Problems Associated with Rolls In Ne = 297 Eccentric Rolls... : : = 297 Improper Ratios of Rol +297 Resonant Frequencies . ... . eeeeree - 299 Installation of Improperly Ground Rolls . 301 Diagnosing Problems ........... = 302 Data Collection and Analysis of Overall Vibration Data . +302 Relative Motion between Rolls . ; : + 306 Conclusions and Recommendations . + 328 References . . 329 Appendix A = 331 Appendix B. Glossary . Index... xiiiCHAPTER ONE: INTRODUCTION TO MACH! THEORY OF VIBRATION The physical movement or motion of a rotating machine is normally referred to as vibration. Since the vibration frequency and amplitude cannot be measured by sight or touch, a means must be employed to convert the vibration into a usable product that can be measured and analyzed. Electronics, mechanics, and chemical physics are closely related. Therefore, it would logically follow that the conversion of the mechanical vibration into an electronic signal is the best solution. The means of converting the mechanical vibration into an electronic signal is called a transducer. The transducer output is proportionate to how fast the machine is moving (frequency) and how much the machine is moving (amplitude). The frequency describes what is wrong with the machine and the amplitude describes relative severity of the problem. The motion can be harmonic, periodic, and/or random. All harmonic motion is periodic. However, all periodic motion is not harmonic. Random motion means the machine is moving in an unpredictable manner. Harmonic Motion Harmonic motion is characteristically a sinusoid or some distorted version, depending. upon the harmonic content, as in Fig. 1-1. All harmonic motion is periodic, meaning it repeats at some point in time. Ina linear system, imbalance in rotating equipment could generate harmonic motion. However, with many variables such as gear problems, looseness, bearing defects, misalignment, etc., such sinusoids are not often found. It is important to understand that a sine wave is simply a plot of a circle against time. Notice how the circle in Fig. 1-1 can be plotted as a sine wave, proving that linear motion is harmonic. All harmonic motion is repeatable and is just one form of periodic motion. Fig. 1-1. Harmonic Motion. Periodic Motion Periodic motion is all motion that repeats periodically. This includes harmonic motion, pulses, etc. Periodic motion is any motion that repeats itself in equal time periods. For example, a misaligned motor coupling that is loose could have a bump once per revolution of the shaft. Although this motion is not harmonic, it is periodic. The time signal will have one pulse every x seconds as indicated in Fig. 1-2.nod M wealCHAPTER 1 Introduction to Machinery Vibration Random Motion Random motion occurs in an erratic manner and contains all frequencies in a particular frequency band. Random motion is any motion that is not repeatable. Popcorn in a popper, rain hitting a roof, and bowling pins being knocked over are examples. Random motion is also called noise. When random noise is generated by a machine, a recording, of the noise played back ten times faster than it was recorded can sound like a TV set after the station has signed off the air. A time signal of random noise will contain all frequencies in a given range. The frequency spectra from such time signals will be up off the baseline as indicated in Fig. 1-3. Often, random motion in a machine is caused by severe looseness. ‘THE RELATIONSHIP BETWEEN TIME AND FREQUENCY Time When we say that AC line frequency is 60 cycles per second, this means if a one second time period was observed, 60 cycles would be present as indicated in Fig. 1-4. However, it is not always practical to observe one second of time and count the number of cycles. ‘We can measure the time period for one cycle and calculate the frequency. We can also calculate the time period for one cycle if the frequency is known. Time and frequency are the reciprocal of each other. For example, if 60 cycles occur in one second, divide one by 60 to get the time period for one cycle. When determining the frequency from the time period for one cycle, divide the time period for one cycle into one (1): S HlM ui AN CATT Ni Fig. 1-4. Time and Frequency.CHAPTER 1 Introduction to Machinery Vibration If 60 cycles occur in one second and the time period for one cycle is 0.0167 seconds, the calculation can be verified by: F x T = 1 or 60 x 0.0167 ~ 1. Please note that the time period for one cycle of all frequencies above 1 Hz will be less than one second. Also note that if frequency is in cycles per second, time must be measured in seconds (generally fractions of a second). Frequency Frequency is the number of cycles that occur in one time period, usually one second. Until a few years ago, frequency was identified as cycles per second (CPS). CPS was changed to Hertz, honoring the man who developed the frequency theory. Today Hertz (cycles per second) is the standard measurement of frequency. Machine speed is measured in revolutions per minute (RPM), but the frequencies generated by those machines are measured in Hertz. From the above discussion, the formulas for frequency and time can be derived: 1 Fe r-+ Fr- F 1 T For the beginner, it may be helpful to construct a triangle such as in Fig. 1-5. To solve for 1, F, or T, simply cover the unknown and the formula can be seen. For example, to solve for F, cover the F, and 1 over Tis left. Where F equals frequency or the number of cycles that can occur in one second, T equals the time period for one cycle, and (1) equals 1 second in this case. ' T = time (frection of @ second) T Fo traquancy (oyon/oe) fad Fig. 1-5. Relationship of Time and Frequency. Example 1-1: What is the time period for 1 cycle if the frequency is 29.6 Hz? Answer: T = 0.0338 sec or 33.8 msCHAPTER 1 Introduction to Machinery Vibration Most electronic instruments measure time in milliseconds or thousandths of one second. To convert milliseconds to seconds, move the decimal point to the left three places. For example, one millisecond (ms) is equal to 0.001 seconds. Therefore: T = 1 ms x (0.001 secims), T - 0.001 sec Example 1-2; What is the frequency of a time period of 50 milliseconds? Answer: T = 50 ms x 0.001 - 0.05 sec, F-1, F-— 1 or 20 kz T 0.05 sec In the above formula, when determining frequency in cycles per second, time must be in seconds. Example 1 Answer: What is the frequency of a signal if the time period is 0.0338 seconds? 1 F- = F206 Ki 010338 sec 296 He 4 T The conversion of cycles per second or Hz into cycles per minute or RPM to determine machine speed is often required. One minute contains 60 seconds and frequency is measured in cycles per second. Therefore: 1 cyclsec x 60 secimin - 60 cyclmin = 60 RPM: Therefore, multiply Hertz by 60 to obtain CPM or RPM and divide CPM or RPM by 60 to obtain Hertz, Please note the industry standard for measuring machine speed is revolutions per minute (RPM). ‘The industry standard for measuring frequency is Hertz (cycles per second). In this text, the industry standards shall be used. Example 1-4: What is the speed of a machine that generates a fundamental frequency of 29.6 Hz? Answer: 29.6 Hz x 60 secimin - 1776 RPM Example 1-5: What is the fundamental frequency a machine will generate if the machine speed is 1180 RPM? Answer:CHAPTER 1 Introduction to Machinery Vibration AMPLITUDE MEASUREMENT. The four different ways to express the vibration amplitude level are: peak-to-peak, zero- to-peak, RMS, and average. Peak-to-peak is the distance from the top of the positive peak to the bottom of the negative peak. The peak-to-peak measurement of the vibration level is shown in Fig. 1-6. This type of measurement is most often used when referring to displacement amplitude. HARMONIC MOTION Fig. 1-6. Peak to Peak. HARMONIC MOTION |=——- one cycle, ——= Fig. 1-7. Zero to Peak. Zero-to-peak or peak is the measurement from the zero line to the top of the positive peak or the bottom of the negative peak. The zero-to-peak value of the vibration level is shown in Fig. 1-7. This type of measurement is used to describe the vibration level from a velocity transducer or accelerometer. ‘The Root Mean Square (RMS) is the true measurement of the power under the curve. In Fig. 1-8, the RMS value is the cosine of 45 degrees times peak (0.707 x peak only applies to pure sine waves). The true RMS value is calculated by the square root of the sum of the squares of a given number of points under the curve. For example:CHAPTER 1 Introduction to Machinery Vibration True RMS ~ HARMONIC MOTION |=—— one cycle ——= Fig. 1-8. RMS and Average. ‘When calculating true RMS, the crest factor and duty cycle must be considered for signals that contain pulses. The crest factor (CF) is the ratio of the peak value to the RMS value the DC component removed. For example: A crest factor of 7 is normally required for accurate measurement of pulses. The duty cycle is the ratio of the pulse width (PW) to the pulse recurrence frequency (PRF). For example: Pw le = Duty cycle = Soe Several forms of pseudo RMS are used in some equipment. For example: 0.707 x PEAK, AC? + DC? Analog meters measure average amplitude. Various constants are then used to calculate peak, peak-to-peak, or RMS. Most measurements that are not true RMS measurements are either overstated or understated. When describing the vibration level of a machine, the RMS value should be used if possible. However, some cases require peak-to-peak measurements, for example, when measuring mils of displacement. Other cases require zero-to-peak displacement 7CHAPTER 1 Introduction to Machinery Vibration measurements such as high places on a roll. The average value is 0.637 times the peak of a sine wave. See Fig. 1-8. Average values are measured by analog meters. Average is then converted to peak by multiplying a constant of 1.57. These calculated values are accurate only when measuring pure sinusoids. The following constants may be helpful. However, they apply to true sine waves only. The more the signal deviates from a true sine wave, the more error is introduced. Average = 0.637 x Peak Average = 0.90 x RMS Peak to Peak = 2 x Peak Peak ~ 1.414 x RMS Peak = 1.57 x Average RMS = 0.707 x Peak RMS ~ 1.11 x Average SOURCES OF FREQUENCIES The three sources of frequencies in machines are: generated frequencies, excited frequencies, and frequencies caused by electronic phenomena. Generated Frequencies Generated frequencies, sometimes called forcing frequencies, are those frequencies actually generated by the machine. Some examples are imbalance, vane pass frequency (number of vanes times speed), gearmesh frequency (number of teeth times speed), various frequencies generated by antifriction bearings, ball passing frequency of the outer race, ball passing frequency of the inner race, ball spin frequency, and fundamental train frequency. Generated frequencies are the easiest to identify because they can be calculated if the internal geometry and speed of the machine are known. Some of the calculated frequencies may be present in most machines without indicating a vibration problem. ‘These frequencies, at acceptable levels without sidebands, include but are not limited to: imbalance, vane pass frequencies, blade pass frequencies, and gearmesh frequencies. Other calculated frequencies should not be present in any form at prescribed calibration levels. These frequencies include, but are not limited to: ball pass frequencies of the outer and inner races, ball spin frequency, fundamental train frequency, etc. Calculated frequencies should not be modulated with any degree of significance by other frequencies. If any of these frequencies are generated, a vibration problem exists. When a rotating unit has a mass imbalance, it will generate a sine wave that has very little distortion. “This signal can be observed in the time domain. The frequency domainCHAPTER 1 Introduction to Machinery Vibration spectrum will have a spectral line at one times speed of the unit. For example, a 1776 RPM fan that is out of balance will have one spectral line at 29.6 Hz. Most pumps and fans can generate vane or blade pass frequency, which is the number of vanes or blades times the speed of the unit. A high vibration at this frequency could bbe the result of buildup on the vanes or blades, the vanes or blades hitting something, or looseness associated with the rotating unit. Example 1-6: What is the blade pass frequency of a 1776 RPM fan with four blades? Answer: ae = 29.6 Hz 296 Hz x 4 Blades = 118.4 Hz Gearmesh frequency is normally seen in data taken from a gearbox or gear train. The frequency is the number of teeth on a gear times the speed of that gear. For two gears in mesh, the gearmesh frequency will be the same for each gear; the ratio of number of teeth to gear speed is a constant. In a gear train, all gears will have the same gearmesh frequency. This vibration is caused by teeth rotating against each other. Multiples and submultiples of gearmesh frequency are sometimes observable in the frequency spectrum and will be discussed later. Example 1-7: A bull gear with 67 teeth is in mesh with a 22-tooth pinion gear. The bull gear is rotating at 6.4 Hz. a) What is the gearmesh frequency (GF)? b) What is the speed of the pinion gear? Answer: GF = no. of teeth x gear speed 0) GF ~ 67 x 64 He ~ 428.8 He b) Since GF is the same for both gears, the speed of the pinion is: 428.8 Hz = 195 Hi 22 teeth * There are many other generated frequencies such as misalignment, bent shafts, bearing frequencies, looseness, etc. These sources will be discussed in detail later. Excited Frequencies Excited frequencies, (natural frequencies), are a property of the system. Amplified vibration, called resonance, occurs when a generated frequency is tuned to a natural frequency. Natural frequencies are often referred to as a single frequency. Vibration is amplified in a band of frequencies around the natural frequency, as in Fig. 1-9. ‘The amplitude of the vibration in this band depends on the damping. When we refer to the natural frequency, we often mean the center frequency. Natural frequencies can be excited by harmonic motion if the harmonic motion is within the half- power points of the center frequency and contains enough energy. The half-power points are down 3 dB on either side of the center frequency. The 9CHAPTER 1 Introduction to Machinery Vibration NATURAL FREQUENCY empiiicaton Fig. 1-9. Relatively Damped Frequency. frequency range between these half-power points is called the bandwidth of the natural frequency. The half-power point, or 3 dB, is 0.707 times peak at the center frequency. It is a general rule to stay at least 10% away from each side of the center frequency. If some frequency is within the bandwidth of the natural frequency and this frequency contains enough energy to excite the natural frequency, the natural frequency will be present. The term "critical speed" means the rotating speed of the unit equals the natural frequency. When this occurs, the natural frequency is considered unacceptable by some experts. Damping is the measure of a machine's ability to absorb energy. Therefore, a relatively undamped vibration signal can be high in amplitude and relatively narrow banded. A relatively damped signal can be low in amplitude and relatively wide banded. Fig. 1-9 displays a relatively damped frequency. F. is the center frequency and is equal to 18 Hz. F, is the low frequency at the half power point (0.707 times peak) and is equal to 14 Hz. F, is the high frequency at the half power point and is equal to 22 Hz. From this information, an amplification factor (AF) can be calculated: F 18. 18 oar - 2.25 "@-F) @-4 8 Some experts agree that the amplification factor should be less than eight (8). Note that the result is the same if RPM is substituted for Hertz. Fig. 1-10 contains a relatively well- damped frequency where: Fy = 17 He, Fy ~ 19 He, + Borar-8 Some standards for fluid film bearing machines require an amplification factor of less than eight (8). However, for rolling element bearings, it is hard to get less than twelve (12). In Fig. 1-9, machine speeds could vary from 0 to 840 RPM and from 1320 RPM up to the second critical. Therefore, the operating range is restricted. However, if you operate on 10CHAPTER 1 Introduction to Machinery Vibration NATURAL FREQUENCY Fon tHe complication _ tector Reine poten Fig. 1-10. Relatively Well-damped Frequency. the critical, the problem is less severe. In Fig. 1-10, the operating range is less restricted. However, if you operate on critical, the problem could be severe. Natural frequencies can be helpful when they act as a carrier, transporting the source of excitation to the measurement point. This normally occurs when natural frequencies are excited by periodic motion such as hits, bumps, or impacts. In such cases, the natural frequency can be present with spectral lines on each side. The number of these spectral lines, or sidebands, is determined by damping and distortion of the wave form. The difference frequency between the spectral lines identifies the frequency or source of excitation. More than one natural frequency may be present and harmonics of the natural frequencies can occur if distortion is present. Natural frequencies can be identified because they are generally not a calculated frequency, but they are modulated by a calculated frequency. Fig, 1-11 is an example of an excited natural frequency. In such cases, if the source of excitation is removed, the natural frequency will not be excited. When the oil film is destroyed in antifriction bearings, a natural frequency can be excited by the metal-to-metal contact between the balls or rollers and the raceways, as in Fig. 1-12. This problem can be solved by adding oil or changing the viscosity of the oil. The worst situation exists when the natural frequency equals the generated frequency. In such cases, the amplitude can be quite high. The most simple solution is to change the generated frequency or raise the natural frequency above the generated frequency. An excited natural frequency is a resonant condition. Resonance in rotating machinery is the same as an amplifier in electronics. Therefore, excessively high vibration amplitudes are often encountered. The Resonance and Deflection Calculator Program (RADC) is Necessary in every predictive maintenance program to identify and calculate natural frequencies, critical speeds, mode shapes, and wavelengths, Critical speeds and mode shapes require some explanation. For explanation purposes we shall use a simple H- beam. The theory applies to more complex systems such as motors, turbines, gears, ete However, the complex explanations are beyond the scope of this book. iCHAPTER 1 Introduction to Machinery Vibration @ 60 Xen AN weae ven Sn FL Ke Wr nsec Fig. 1-11. Excited Natural Frequency. 0.58 : 0.42 0.28 0.14 R 0.00 ‘4. 0.00 Hz y=0.00 EU Fig. 1-12. Inadequate Lubrication. 12CHAPTER 1 Introduction to Machinery Vibration A critical speed occurs when a machine speed equals a natural or resonant frequency of the machine or some machine part. ‘The mode shape explains the way the beam or machine is bending. The first mode equals the first critical, the second mode equals the second critical, etc. The second or higher criticals are seldom a harmonic of the first or higher critical. The wavelength or lambda (1) is the physical length of a beam required for one cycle to occur. The wavelength should not be confused with the natural or resonant frequency, because the natural frequency occurs at one quarter or three quarters of a wavelength. ‘The reason for this seeming paradox is loops and nodes. A node or a nodal point is the place where the curve crosses the zero point. A loop is the area where the curve is not crossing zero. Zero motion or vibration occurs at a node and maximum vibration occurs at the peak of a loop. In order to keep the mathematics to a minimum, the RADC software package is used. This discussion explains the natural frequencies and mode shapes of a 6 by 6 by 0.25 inch H-beam. The H-beam is six feet long and one end is fixed in concrete. Fig. 1-13 contains the mode shape for the first natural frequency. The first natural frequency is 26.86 Hz. ‘The node is at near zero and below. The maximum loop is at 72 inches and the quarter wavelength is also 72 inches. The second, third, and fourth natural frequencies are 170.58, 482.47, and 946.42 Hz respectively. Fig. 1-14 contains the mode shape for the second natural frequency of 170.58 Hz. Nodes occur at zero and 62 inches, which are zero and one half wavelength, respectively. ‘Maximum loops occur at about 36 inches and 72 inches, which are one quarter and three quarters wavelength, respectively. UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 NATURAL FREQUENCIES & MODE SHAPES REPORT RADE MODEL, 6 Natural Frequencies (Hz): 26.86, 170.58, 482.47, 946.42, > NORMALIZED MODE SHAPE WL Natural Frequency: 26.06 He 72/00 a oop 1/4 3 ae Fig. 1-13. Normalized Mode Shape #1. 13CHAPTER 1 Introduction to Machinery Vibration i ‘UCT RESONANCE AND DEFLECTION CALCULATOR 2,00 NATURAL FREQUENCIES & MODE SHAPES REPORT AOC MODEL 6 Natural Frequencies (Hz): 26.86, 170.58, 482.47, 946.42, © NORMALIZED MODE SHAPE 42 Natural Frequency: 170.58 Hz [ Nodal points Tatoo 2ato0 Fz00 Fig. 1-14. Normalized Mode Shape #2. UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 NATURAL FREQUENCIES & MODE SHAPES REPORT aoe MODEL 6 Natural Frequencies (Hz: 26. 6, 470.58, 482.47, 946.42, > NORMALIZED MODE SHAPE HS Natural Frequency: 492.47 Hz Node 13 i200 za'00 72/00 Loop 14 3 Node 1/2 t Loop 1/4 > Fig. 1-15. Normalized Mode Shape #3. 4CHAPTER 1 Introduction to Machinery Vibration Fig. 1-15 contains the mode shape for the third natural frequency of 482.47 Hz. This mode shape contains three loops and three nodes. One wavelength is about 68 inches Jong at a frequency of 482.47 Hz in this case. Please note the first mode or natural frequency contains one loop and one node. The second mode contains two loops and two nodes. The third mode contains three loops and three nodes. This pattern is repeatable at higher modes. This software package removes the mystery and the complex mathematics from analyzing natural frequencies. This software also permits you to simulate a repair to determine if the proposed repair will solve the problem, and allows you to evaluate any new problems that may be generated from the repair. For example, what would the effect be if the beam was shorter or longer, or if it was heavier or lighter? What if the H-beam was made into a box or channel, or was braced at various points, etc? Following are some simple rules to remember when working with resonant frequencies: 1. Mass - If mass is added, the frequency goes down. If mass is removed, the frequency goes up. 2. Stiffness - If stiffness is increased, the frequency goes up. If stiffness is decreased, the frequency goes down. 3. Increase in length lowers stiffness. 4. Decrease in length increases stiffness. 5. Braces increase stiffness. Another major feature of the RADC software is the ability to calculate the resonant frequency and the deflection ot sag in a beam or shaft. Fig. 1-16 is a 20-inch diameter 4140 steel pipe 300 inches long with a wall thickness of 0.5 inches. Fig. 1-17 contains the first, second, and third natural frequencies of 24.35, 96.74, and 205.39 Hz, respectively. 4140 ste! ro 0.5" shal rckness pinned ends Fig. 1-16. Steel Pipe Model. 15CHAPTER 1 Introduction to Machinery Vibration ‘UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 NATURAL FREQUENCIES & MODE SHAPES REPORT 2 Natural Frequencies (Hz 24.95. 96.74, 205.30 NOAMALIZED MODE SHAPE #1 Natural Frequency: 24.35 Hr Fig. 1-17. Normalized Mode Shape #1. Table 1-1. Static State Values. SS UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 STATIC STATE REPORT Location shear Moment Slope Def tection cinehes TBF, canehbe> = des ‘Cinches) 9.00 4300 2.0128 9.00) 38:00 i890 co: = -.0088 82:00 700-16 cerotat So:tee 90.00 $20.44 “Bicays = “Bbla 130:00 360.65 Bios = 818188 159.00 9.00 9.00 “0.02 fe0'00-200 88 9.098 -0 0188 219.00 © ~820/1 80073 = 00170 24:00 ~700:16 O:os08 =0°0324 370.00 iea8 Ocoiat ‘0088 30000 1200 0.0128 0.00 Maxinun State Values hear: | 1300 at L = 0.00 fonent: 97519 at L_= 150.00 Siope: -0.0390 SEE > 8:80 Deflection: =0.02 “ae "=" $50.00 cd 16CHAPTER 1 Introduction to Machinery Vibration Table 1-1 lists the static state condition for this pipe. The maximum deflection occurs at 150 inches and the pipe deflects or sags 0.0209 inches. The Resonance and Deflection Calculator Program puts high technology in the hands of the predictive maintenance department. Until now, this technology was only available to the more sophisticated engineering departments. Frequencies Caused by Electronic Phenomena In certain situations, false or misleading signals can be present. For example, when a sinusoid is clipped (as occurs when the input to the Real-Time Analyzer (RTA) is overloaded slightly), it can cause a string of harmonics. The amplitude of the harmonic content is normally quite low as in Fig, 1-18. If the clipping is serious enough to cause a square wave or near square wave, only the odd harmonics may be present, or the odd harmonics can be higher in amplitude than the even harmonics. See Fig. 1-19. Also, if a sinusoid is distorted and the time period for each halfwave is not equal, harmonics can be generated as in Fig. 1-20. When true harmonic motion is present, the number of positive going peaks observed for one time period is equal to the true harmonic content. The amplitude of these harmonics is normally high in the frequency domain. When distortion occurs with true harmonic content, the number of harmonics in the frequency domain can exceed the number of positive peaks for one period in the time domain as in Fig. 1-21. If the construction of a machine is sound and the various parts are not defective, the machine will behave in a linear manner. Pure imbalance will generate a single frequency 80.00 460.00 i 2400 ‘3000 Fig. 1-18. Clipped Time Signal. 7CHAPTER 1 Introduction to Machinery Vibration a 2.00 : 1.80 1.20 2 0.80 0.40 0.00 5.00 tz 0.00 Ht eu 1.02 © 0.61 0.20 ovnd 0.20 Fo 0.61 “02 © ni ‘600 ‘0 =f AN nese Ven 78 FIL Fig. 119, Square Wave, eu 1” Fig. 1-20. Distorted Wave Form. 18BRBeS gsCHAPTER 1 Introduction to Machinery Vibration without harmonics or modulation as in Fig 1-22. There are several forms of nonlinear problems in a machine. Misalignment, for example, is nonlinear and is not speed related above certain speeds depending upon shaft size, coupling type, amount of misalignment, etc. It is general knowledge that at higher speeds, less misalignment is allowed. After the speed increases, making misalignment a problem, the vibration level will increase to a given point and remain relatively constant at higher speeds. Looseness is probably the most common form of a nonlinear machine problem. Looseness is also the most common problem in operating machines. When some form of looseness is present in a machine, the machine becomes “tuned for maximum sensitivity.” When this occurs, a small amount of imbalance can cause high vibration levels. Also, the machine is easily shaken by a small excitation from other machines. Looseness causes distortion in the signal. The distortion causes harmonics, and if clipping or truncation ‘occurs, sum and difference frequencies will be present in the spectra. A good time domain signal is necessary to identify all frequencies in the frequency spectra and accurately diagnose machinery problems. Forcing Function fant shot Mochine = | ‘ se UL IY Fig. 1-23. Physical Nature of Vibration. FORCING FUNCTION When a machine vibrates, the cause must be determined. The cause of the vibration is often referred to as the forcing function. A number of possible problems, any combination of these problems, and an infinite degree of each problem can cause the machine to vibrate. The amplitude of each problem can be either overstated or understated depending on transfer function, resonance, damping, frequency addition, and frequency subtraction. Therefore, very low amplitudes can be very serious problems in some cases and very high amplitudes can be relatively minor problems in other cases. Frequency measurement and analysis makes vibration analysis work. Each forcing function has its own frequency, as indicated in Fig. 1-23. If the problem is present, the frequency is present. The frequency can manifest itself as a discrete frequency or as a sum and/or difference frequency. Resonance, damping, frequency addition, and frequency subtraction do not affect the frequencies caused by vibration. This fact explains why frequency analysis is accurate, while amplitude measuring and trending are not accurate when diagnosing problems in rotating machinery. 20CHAPTER 1 Introduction to Machinery Vibration Defects in rotating machinery are caused by normal wear, improper lubrication, overloading, improper installation, improper manufacturing, etc. Vibration analysis identifies the problems or problem areas. Management, engineers, and technicians must determine the cause of the problems and direct the necessary repairs, or find the appropriate solutions to the problems. Vibration analysis of rotating machinery involves calculating the frequencies the machine can generate and measuring the frequencies the machine is generating. Then, relating the measured frequencies to the calculated frequencies and machine installation can identify the problem. If the forcing function is a hit or an impact, the hit or impact may not be measurable. However, the machine response to the impact can be measured. Problems in machinery induce vibration and result in wear, malfunction, and/or structural damage. COMBINATIONS OF MACHINE PROBLEMS, ‘The amplitudes of various frequencies and the frequencies in rotating machines can add and subtract. These frequencies mix, detect, amplitude modulate, and frequency modulate in much the same way as signals in electronic mixers, detectors, and modulators. An electronic amplifier performs in much the same way as resonance. An attenuator performs in much the same way as damping. The simplest form of amplitude addition and subtraction is imbalance. When a rotor becomes out of balance, one side is heavier than the other side. This imbalance condition or heavy place generates a signal at one times RPM, as shown by the solid line in Fig. 1- 24. Correction Fig. 1-24. Simplest Form Imbalance. Balancing the fan is accomplished by adding an equal amount of weight to the light place. This weight generates a one times RPM equal in amplitude and 180 degrees out of phase with the imbalance, as indicated by the broken line in Fig. 1-24. Since we live in an imperfect world, the signals may not cancel completely. Therefore, a small amplitude of the resulting imbalance may be present as indicated in Fig. 1-25. 2CHAPTER 1 Introduction to Machinery Vibration Fig. 1-25. Residual Imbalance. MIXING FREQUENCIES When two frequencies are present in a machine and a cause and effect relationship is not Present, the high frequency will be riding the low frequency and the Fast Fourier Transform (FFT) will yield spectral lines at frequency one and frequency two. If there is @ cause and effect relationship and the two frequencies can mix together, the result is amplitude modulation. Without getting mathematical, amplitude modulation is a time varying amplitude. Amplitude modulation is caused when the equipment has some form of nonlinearity. This nonlinearity permits the amplitude of the two signals to add together when the signals are in phase, or subtract when the signals are out of phase. With amplitude modulation, the carrier frequency will be the frequency with the highest amplitude. The envelope of the varying amplitude will be the difference between the two frequencies. An FFT of these signals can yield spectral lines at frequency one, and frequency one plus and/or minus frequency two. For example, suppose gearmesh frequency is modulated by gear speed, gearmesh frequency is 1200 Hz, and gear speed is 20 Hz. An FFT of this signal would then yield Spectral lines at 1200 Hz, 1200 + 20 = 1220 Hz, and/or 1200 - 20 = 1180 Hz. Descriptions of these frequencies are: 1 1200 Hz is gearmesh frequency. 2, 1220 Hz is gearmesh frequency plus gear speed. This is a sum frequency. 3.1180 Hz is gearmesh frequency minus gear speed. This is a difference frequency. 4. The difference between 1200 and 1220 Hz, or 1200 and 1180 Hz is 20 Hz, and this is also a difference frequency. 5. The source of excitation, or the problem shaft or gear is usually expressed as a difference frequency. If the amplitude modulated signal is clipped or truncated, the FFT can yield an array of sum and difference frequencies of the two frequencies, including frequency one plus 22CHAPTER 1 Introduction to Machinery Vibration frequency two. There are many forms of sum and difference frequencies. These frequencies will be discussed later. ELECTRICAL AND MECHANICAL RELATIONSHIP As mentioned previously, the basic physics of electronics and mechanics are the same. This similarity explains why we can convert a mechanical motion into an electrical signal, then diagnose the mechanical problem with the electrical signal. For example, Table 1-2 displays the equivalent electrical and mechanical terms. Table 1-2. Comparison of Electrical and Mechanical Terms. SS ELECTRICAL MECHANICAL E = Voltage F = Force 1 Current V = Velocity R = Resistance or Impedance (Z) P = Power HP = Horsepower F = Frequency (CPS) S = Speed (RPM) Eee) In electronics there is electromotive force or voltage; in mechanics there is force. In electronics there is current; the mechanical counterpart is velocity. In electronics there is resistance, which is purely resistive resistance, and impedance which is the total opposition to current flow. This includes resistive, capacitive, and inductive resistance. ‘The mechanical counterpart to resistance is friction. In electronics there is power; in mechanics there is horsepower or work. In electronics frequency is used; in mechanics speed is used. Table 1-3 contains some of the electrical formulas established by Ohm's Law and the mechanical equivalents. Some liberties were taken with the mechanical equivalents for ease of explanation. Each mechanical action or occurrence is related to an electrical action or occurrence. Vibration analysis begins with mechanical energy and ends with electrical energy. In examining a rotating object mechanically, a force causes the object to rotate. Electrically, voltage causes the object to rotate. We can mechanically measure the velocity of the rotating object, or electrically measure the current of an electric motor. For example, to compare the mechanical and electrical worlds, we can slow down the rotating object by mechanically applying friction, or electrically applying resistance. Speed (mechanical) or frequency (electrical) measures how fast the object is moving. As explained in the introduction of this text, a transducer is used to convert mechanical energy into electrical energy. The electrical energy is then processed and analyzed for possible machine problems. Even though mechanical and electrical terms may appear to be different, they 23CHAPTER 1 Introduction to Machinery Vibration are very much alike. Table 1-3, Electrical and Mechanical Equivalents. F=VR TIME AND FREQUENCY DOMAIN The frequency domain is all around us. However, sometimes we call frequency by another name. For example, light is a frequency. The color red is a frequency. Sound is a frequency. We do not refer to these items as frequencies; we call them light, color, and sound. As light, color, and sound are varied, the frequency changes. The audio range of frequencies is from 20 to 20,000 Hz. It is called the audio range because it is the frequency response of the human ear. However, most people cannot hear 20 or 20,000 Hz, and some cannot hear several frequencies between these limits. Low-pitched sound contains low frequencies, as in a baritone singer. A high-pitched sound contains high frequencies; an opera singer is an example. ‘Some things are better measured in time and others are better measured in frequency. For example, consider the sentence, "We go to work once a day.” We would not say that “once a day’ is the frequency. Rather, we say we work eight hours each day or 40 hours each week, which are amounts of time. Life and the world are full of misconceptions because we do not understand our surroundings. For example, we say it is dark. There is no such thing as darkness, only the absence of light. Darkness cannot be measured. We say it is cold. There is no such thing as cold, only the absence of heat. However, heat and light can be measured. These misconceptions carry over to other areas. Many times the author has heard a client say "I looked at the bearing (or equipment) and I did not see anything wrong," The client was telling the truth, but this does not mean the problem is not there just because he did not see it. Many times a mechanic, supervisor, or engineer has used a screwdriver to listen to a machine after the machine has been diagnosed as having a problem. This person would then say: “It sounds okay to me." However, the machine docs have a problem; the frequency may be out of the audio range or it could be masked by other noises. 24CHAPTER 1 Introduction to Machinery Vibration Fig. 1-26. Relationship between Time and Frequency. Analysis of some problems such as imbalance can be diagnosed in the time domain. However, time domain signals from rotating machines are often very complex. Such signals must be analyzed in the frequency domain. If imbalance is diagnosed in only the frequency domain, errors will often occur. In vibration analysis, use of both the time domain signal and the frequency domain spectra are required for complete, accurate analysis. To move from the time domain to the frequency domain, we must perform a Fast Fourier Transform on the time domain signal. Mr. Fourier was a French mathematician who proved all complex wave forms could be broken down into their individual frequency components mathematically. However, this brilliant technology was not used extensively until the advent of the computer. Utilizing the transformation of the time signal to a frequency spectrum, Mr. Fourier’s technology, and the computer's rapid capabilities, can produce a Fast Fourier Transform or FFT. Fig. 1-26 contains a time signal and the corresponding FFT. The first (left-most) time signal contains the fundamental, second, and third harmonic. This statement is true because each time period has three positive going peaks. The first spectral line is the fundamental frequency and its time signal. The second spectral line is the second harmonic and its time signal. The third spectral line is the third harmonic and its time signal. If the time signals of the three frequency components were added together, the result would be the first (left) time signal, as indicated in Fig. 1-26. RELATIONSHIP BETWEEN VELOCITY, DISPLACEMENT, AND ACCELERATION Velocity is the measurement of how fast an object is moving from zero-to-peak and is normally measured in tenths of one inch per second (IPS). The effective frequency range of most velocity transducers is from about 10 to 2,000 Hz. Velocity is the most accurate measurement because it is not frequency related. For example, 0.15 IPS is the same at 10 Hz as it is at 2,000 Hz. Displacement is the measurement of how far an object is moving from peak-to-peak and is normally measured in thousandths of one inch (mils). Displacement is frequency 25CHAPTER 1 Introduction to Machinery Vibration related. Therefore, any measurement of displacement must be at a specified frequency. For example, three mils at one times RPM of a 1200 RPM motor is a meaningful statement. However, an overall measurement of three mils is a meaningless statement. Three mils at 1200 RPM is about 0.2 IPS of velocity. However, three mils at 3600 RPM is almost 0.6 IPS of velocity. The effective frequency range of noncontacting displacement transducers is from about 0 to 600 Hz. For contacting displacement transducers, the effective frequency range is about 0 to 200 Hz. Acceleration measures the rate of change of velocity from zero-to-peak and is normally measured in units of gravitational force (g's). This means that high frequencies generate high g levels, and acceleration is frequency related. For example, 3 g’s at 20 Hz equals. about 9 IPS of velocity. Three g’s at 2,000 Hz is about 0.09 IPS of velocity. The effective frequency range for low frequency accelerometers is from about .2 to 500 Hz. The effective range of high frequency accelerometers is from about 5 to 20,000 Hz. The Vibration Calculator Program is useful for comparing various Engineering Units (EU). Fig. 1-27 displays the frequency response curves for displacement, velocity, and acceleration. All curves are plotted against a constant value of 0.15 IPS of velocity. {splacemer ‘Aceelefotion Velocity (0.15 1s Displacement (mi ¢ 16 Too 1059 70.00 Frequency (Hs) Fig. 1-27. Frequency Response Curves. Please note the displacement curve is downward and outward sloping. This indicates low frequencies generate high levels of displacement and high frequencies generate low levels. of displacement. Therefore, the displacement transducer most effectively measures lower frequencies. ‘The frequency response of the velocity transducer is relatively flat from about 10 Hz to about 2,000 Hz. It is the most accurate transducer to use in this frequency Tange. The acceleration curve is an outward and upward sloping curve, which means that high frequencies generate high levels of acceleration. The accelerometer must be used for frequencies above 2,000 Hz and may not be as effective for frequencies below 100 Hz. A strong word of caution is due at this point, Two transducers must be used when making measurements on equipment that generate both high and low frequencies, one transducer for measuring low frequencies and an accelerometer for measuring high frequencies. If this procedure is not used, the high levels from the high frequencies require equipment setups that could mask the low frequencies. If so, low frequency problems may not be identified. Except for real-time monitors, velocity should be used for most vibration monitoring and analysis. Velocity levels remain relatively constant over a wide frequency range. Therefore, velocity should be used whenever possible; displacement and acceleration are both frequency related. The frequency of the vibration source must be 26CHAPTER 1 Introduction to Machinery Vibration known for the measurement in either displacement or acceleration to be meaningful. UNITS OF MEASUREMENT The following symbols are used in the vibration field: D= _ Displacement in inches peak-to-peak V= Velocity in inches per second zero-to-peak &= Acceleration in g’s zero-to-peak F= Frequency in Hertz RELATIONSHIPS The relationships between these measurements and the conversion from one engineering, unit to another can be accomplished with the use of the following equations: 1, V~ nFD 2 v~ 6144 £ F 3. g - 0.0511 FD 4. g ~ 0.0162 VF 5. D- 0.3163 & F 6. D- 1957 & F The following examples may be helpful: Example 1-8: The vibration level on a variable speed motor operating at 400 RPM is 0.12 IPS (measured with a velocity transducer). The displacement in peak-to-peak can be calculated by using equation 5: D - 0.3163 £ F First, frequency in Hertz must be determined. 400 RPM 0.12’ F - ———— - 6.67 Hz, D - 0. aa 60 0.3188 (er) Then, D = §.7 mils 7CHAPTER 1 Introduction to Machinery Vibration Example 1-9: If the vibration level for the motor in Example 1-8 is determined with an accelerometer to be 0.013 g’s, then the displacement can be calculated by using equation 6 D ~ 19.87 |-£.) ~ 19.57 (9913) - 5.7 mus (4) (a8 o7™ Example 1-10: A pump driven by the motor in Example 1-8 is turning at 111.4 RPM. The vibration level is measured with an accelerometer which reads 0.01 g’s. The vibration level in IPS can be calculated from equation 2: v~ 6144 & F Again, F must be calculated. For this problem, 111.4 RPM 60 F = 1.86 Hz, and - 61.44 & ~ 61.44 v- 61.44 £ 6) (¢ Warning: These calculations are true only for 0.01 g’s at the given motor speed. Example 1-11: The vibration of the pump in Example 1-10 is 40 mils when measured with a displacement probe. The vibration level in terms of velocity (IPS) can be calculated directly from equation 1: V = xFD = (1.86)(40 x 10°) - 0.23 IPS Example 1-12: The vibration level on a motor running at 400 RPM is 0.10 IPS. The corresponding acceleration is calculated using equation 4: g - 0.0162 VF - 0.0162 (0.10) (3) = 0.011 gs Example 1-13: The vibration level for the pump in Example 1-10 is 10.1 mils (measured with a displacement probe). An accelerometer would read, from equation 3: g - 0.0511 F*D ~ 0.0511(1.86)* (10.1 x 10°) - 0.0018 g's Example 1-14: The vibration level on a motor running at 1800 RPM is 0.3 IPS (measured with a velocity transducer). The displacement in mils can be calculated by using equation 5: 28CHAPTER 1 Introduction to Machinery Vibration p - 03183 ¥ F First, frequency must be calculated in Hertz. re 1800 RPM = 90 He Substituting this value in equation 5: D - 0.3183 (33) = 3.18 mils Example 1-15: Vibration level of the motor in Example 1-14 can be determined in g's by using equation 4: g ~ 0.0162 VF ~ 0.0162 (80) (0.3) - 0.146 g's Example 1-16: If the vibration level of the motor in Example 1-14 is measured with an accelerometer to be 0.146 g's, then the displacement can be calculated using equation 6: D- 1957 £ ~ 19.57 fer Example 1-17: The vibration level of the motor in Example 1-14 can also be computed in IPS by using equation 2: v~ 61.44 £ - 61.44 = 0.3 IPS The above examples are useful. However, it may not be convenient to perform the mechanical calculations each time conversion is needed and the formulas and constants may not be available. A simpler, better, and faster method is the Vibration Calculator Program. The main screen contains three frequency response curves for a pure imbalance condition: displacement, velocity, and acceleration. A menu is also provided. Table 1-4 contains the menu contents. ‘The operator enters the frequency of interest and the vibration level in engineering units: cither displacement, velocity, or acceleration. The program then calculates the other two engineering units. The program also calculates engineering units from dB and transducer sensitivity or calculates dB from transducer sensitivity and engineering units. Fig. 1-28 contains the frequency response curve for displacement. The x axis is frequency in Hertz and the y axis is amplitude in engineering units. Please note the displacement is 1 mil at 20 Hz, about 2.3 mils at 30 Hz and 4 mils at 40 Hz. 29CHAPTER 1 Introduction to Machinery Vibration Table 1-4, Vibration Calculator Program. A MAIN SCREEN MENU Frequency (Hertz): 20.00 Displacement (Mil): 1.00 Velocity (IPS): 0.06 Acceleration (G): 0.02 Sensitivity (mV/EU): 375.00 Vibr. Level (EU): 1.00 eB: 51.48 Fig. 1-29 contains the frequency response curve for velocity. The x and y axes are the same as in Fig. 1-28. Please note the velocity is 0.07 IPS at 20 Hz, 0.21 IPS at 30 Hz, 0.45 IPS at 40 Hz, and 1 IPS at 50 Hz. Fig. 1-30 contains the frequency response curve for acceleration. The x and y axes are the same as stated previously. Please note the acceleration is 0.02 g’s at 20 Hz, 0.1 g’s at 30 Hz, and 0.34 g's at 40 Hz. Figs. 1-28 through 1-30 represent the effects of imbalance in displacement, velocity, and acceleration as speed is increased or decreased. ‘These graphs can be compared with actual coast down data to identify imbalance, misalignment, bent shaft, and resonance problems. man 7:00 4.20 0.00 a ‘e105 30:00 30:00 ‘3000 ‘40 100°" g0"00 =20 04 Hz ye1.00 iI 174=48.90 nsec Hertz Fig. 1-28. Imbalance Response for Displacement. 30CHAPTER 1 Introduction to Machinery Vibration IPs. 1.00 0.80 0.00 oo" 5700, 3000 ‘30°00 "40 :00°""'6000 30.08 Hz y=0.24 TPS 174933.27 nsec Herts Fig. 1-29. Imbalance Response for Velocity. 0.64 0.00 —— ‘8 10:00 30.00 ‘3060 "40200" $0700 X=30.06 HE y=0.40°G 41M233.27 nsec Hertz Fig. 1-30. Imbalance Response for Acceleration. 31CHAPTER 1 Introduction to Machinery Vibration ‘The program and graphs provide quick conversions between the three engineering units without having to remember formulas or perform calculations. The program also calculates vibration level from transducer sensitivity and dB, or dB from transducer sensitivity and vibration level. WAYS OF MEASURING VIBRATION, Vibration can be measured by displacement, velocity, and acceleration. Transducers, meters, data collectors, and real-time analyzers are some of the tools capable of measuring vibration levels. Displacement is measured with a contacting displacement transducer. ‘The contacting displacement transducer is a mechanical device that measures relative motion between two surfaces. The tip of the contacting displacement transducer is a plunger which moves back and forth with the relative motion between the two surfaces. ‘The moving plunger changes the transformer's differential output. The resulting signal is the measured displacement. ‘The noncontacting displacement transducer measures relative motion with magnetic flux. (Magnetic flux is generated by the driver.) The noncontacting displacement transducer measures the dynamic displacement using magnetic fields. The tip of the noncontacting displacement transducer is a small coil which measures the change in magnetic flux when the probe tip is near a ferromagnetic material. The contacting displacement transducer and noncontacting displacement transducer must be used when the frequencies generated are below 10 Hz. The contacting displacement transducer should not be used above 200 Hz. The noncontacting displacement transducer may not be useful above 600 Hz. Velocity is measured with a velocity transducer that has a relatively flat frequency response between 10 and 2,000 Hz. The velocity transducer measures the vibration using permanent magnet that is fixed, and a coil of wire mounted on a spring. When a vibration source is applied to the transducer, the coil moves over the magnet, producing a signal. In most applications, the velocity transducer is the best tool with which to measure and record vibration. Acceleration is measured with an accelerometer. The frequency response is greatest when. measuring high frequencies. An accelerometer measures the vibration generated by using piezoelectric crystals. A piezoelectric crystal produces a voltage signal when the crystal is deflected. A change in speed will cause the crystal to deflect and display a change in vibration. An accelerometer must be used when the expected frequencies are greater than. 2,000 Hz. It is possible to measure displacement with an accelerometer, acceleration with a velocity transducer, etc. However, the signals must be either differentiated or integrated. These conversions can be calculated with the equations given above or with the Vibration Calculator Program. RELATION BETWEEN DIAMETER, SPEED, AND RPM. ‘The speed paper or other material travels through a machine can be obtained at various readout locations on the machine. Paper speed is useful in determining the RPM of a specific roll. The displacement of an element of paper is equal to the displacement of an 32CHAPTER 1 Introduction to Machinery Vibration element on the surface of the roll. Therefore, the speed of the paper over a roll is equal to the speed of the roll surface. The equation for the circumference of a circle,C = xd, indicates how far the roll element travels during one complete revolution, where d equals the roll diameter in feet. Given the paper speed in FPM, the RPM of a roll can now be determined from: q-—-F ps = xd(RPM) RPM ~ -PS. (RPM) . (xd) where ps is the paper speed. Example 1-18: Paper speed through a dryer section is 1750 FPM. The diameter of a dryer roll is 5 feet. The speed of the dryer in RPM is: = BS. MO a4 RPM Ta nO) , HOW TO DETERMINE MACHINE SPEED IN FPM FROM THE VIBRATION DATA Vibration data from a felt roll or the fundamental frequency generated by a felt roll are typically from about 2 Hz to about 18 Hz. Look for these frequencies in the spectrum. Ifa spectral line is found that could be the roll speed, do a trial paper speed calculation. For example, if a spectral line is present at 6.5 Hz and the felt roll diameter is about 16 inches, convert 6.5 Hz to RPM by multiplying by 60: 6.5 x 60 = 390 RPM. Next, convert roll diameter in inches to feet by dividing by 12: 16 a6 = 1.333 12 Multiply roll diameter in feet by pi: 1.33 x x = 4.178. This is the roll circumference in feet. Now multiply the roll circumference, 4.178, by the roll speed of 390: 4.178 x 390 = 1629 feet per minute (paper speed). This speed can be verified by gearmesh frequency from the dryer roll. For example, if the gearmesh frequency is 138.4 Hz, divide the gearmesh frequency by the number of teeth on the gear (80 teeth): 1984 1.79 He (dryer speed) 1.73 x 60 = 103.8 RPM is the dryer speed. Most dryer cans are 15.7 feet in circumference. ‘Thus, 15.7 x 103.8 = 1629 FPM. This is the same speed obtained in the first test. Paper machine speed can also be verified with bearing frequency. For example, an SKF 22318 bearing that has a defect on the outer race can generate a frequency of 40.17 Hz, if the roll is rotating at 65 Hz or 390 RPM. A fundamental frequency at 40.17 Hz and harmonics in a spectrum from a felt roll would indicate a defect on the outer race. Divide 33,CHAPTER 1 Introduction to Machinery Vibration 40.17 by 6.18 (the ball pass frequency of the outer race at 1 Hz): 40.17/6.18 = 65 Hz. The roll must therefore turn at 6.5 Hz or 390 RPM in order to generate a frequency of 40.17 Hz if a defect is on the outer race. Finally, 390 x 4.178 (roll circumference) = 1629 FPM machine speed. CONCLUSION AND EFFICIENCIES Thus far, this textbook has explained the basics in machinery vibration analysis: Why and how a mechanical signal can be converted into an electrical signal and analyzed; various types of motion; different ways to measure motion; sources of frequencies; complex machine problems (mixers, detectors, resonance, and damping); relationships between velocity, displacement, and acceleration; units of measurement for velocity, displacement, and acceleration; and problem examples. These basics are the first step to understanding vibration analysis. Vibration analysis is one of the best ways for companies with heavy industrial rotating machinery to save money. With the use of the above basics, companies can stop fixing, equipment that is not broken, improve the quality of spares, predict impending failures, and reduce downtime. 34CHAPTER TW TIME AND FRE IS TECHNIQUES INTRODUCTION Accurate diagnosis of problems in rotating machines requires a thorough understanding of the time domain signal and the frequency domain spectra. Following are some of the reasons: 1, The time signal is a plot of amplitude versus time. This signal contains all frequencies, harmonics, and subharmonics. The phase relationships of these frequencies are also contained in the signal. Pulses, amplitude modulation, frequency modulation, truncation, and distortion are also present. 2. The frequency spectra are plots of amplitude versus frequency. These spectra contain frequency, harmonics, subharmonics, and sum and difference frequencies. The FFT produces the frequency spectrum from the time signal, based on electronic physics. However, during the process, some information is lost. For example, phase, true amplitude of pulses, nature of the pulses, bandwidth, and the various forms of modulation are not easily identified in the frequency spectrum. 3. A mechanical machine may not generate a fundamental plus harmonics in the same way as in the electronic world. However, a rotating machine does generate relatively linear signals when a linear problem exists, such as imbalance. A machine can generate a distorted signal as a result of a nonlinear problem. This distorted signal is a composite signal, as would be obtained after various frequencies and harmonics are combined. 4, For the above reasons, various time signals can produce the same frequency spectrum. This explains why the time signal must be considered. Costly errors in diagnostics and loss of credibility could occur if the time signal is not analyzed. Before analyzing the time signal, an understanding of how frequencies add and subtract, and the effects of the phase relationships is required. It may be helpful to remember that multiplication is a series of additions, and division is a series of subtractions. This chapter explains the time signal in a nonmathematical way. However, extensive use of a computer and the Signal Analysis Program (SAP) are necessary to provide graphic representations of various signals and how they mix. The SAP software performs the required mathematics. BASIC PHYSICS Alll things in the universe obey the basic laws of physics. Vibration signals from rotating machinery must obey these same basic laws of physics. This is why we can take data from either side of a motor and receive the same results. (Some slight variations can occur in nonlinear systems because of transfer functions.) Ina pure linear system, data taken in different directions around a motor should be the same, except for phase. CHAPTER ONE made the point that a sine wave is the plot of a circle against time. All complete circles contain 360 degrees and all complete sine waves contain 360 degrees. ‘The phase of a signal can be anything from 0 to 360 degrees, depending on the referenceCHAPTER 2 Time and Frequency Analysis Techniques point. For example, in comparing a signal taken from the horizontal direction with a signal taken from the vertical direction, one signal should lag the other signal by 90 degrees. This is because the positions from horizontal to vertical are 90 degrees apart on the machine. This phase relationship should also apply to other data taken at various points around the machine. The time signal is continuous from start to stop of the machine. Concerning the phase of the fundamental frequency, the starting point happens purely by chance, depending on the exact instant of time when data collection is started. Once again, we are saved by the basic laws of physics. We can start taking data at any instant of time or location, and it does not make any difference. The reason for this is the phase relationship between the fundamental and other frequencies, or a once-per-revolution marker, will remain constant. These frequencies will add and subtract, depending on the phase relationship, When the signals are in phase, the amplitudes will add. This is why positive sidebands occur on some frequencies. When the signals are out of phase, they will subtract. This is why negative sidebands occur. This also explains how and why truncation occurs. Because of the above physics, a given frequency spectrum and time signal can take many forms, depending on the phase relationships of the various signals, including harmonics. In rotating machines, several different problems can generate the same frequency spectrum. For example, a machine that is loose can generate a fundamental and the second harmonic. A machine that has a bent shaft can also generate a fundamental and the second harmonic. The only way to determine which problem exists is to determine the phase relationship between the fundamental and the second harmonic. If these two signals are in phase, the shaft is bent. If the two signals are out of phase, the machine is loose. Currently, the only way to determine this phase relationship is with the time signal. The meanings of various phase relationships for specific problems are discussed in more detail in other chapters. SINGLE FREQUENCY A single frequency is often referred to as a discrete frequency and is the simplest form of frequency data. Fig. 2-1 contains the time signal of a single frequency and the resulting frequency spectrum from the Signal Analysis Program. The time period for each cycle is 0.01667 seconds and the signal is sinusoidal. Sinusoidal simply means the signal follows the sine function. Mathematically, the time signal is: y ~ A cos (360 ft + 8), where y= instantaneous value of the signal A = the zero-to-peak amplitude © ~ the phase angle (degrees) t= the time variable (seconds) > the frequency variable (Hertz)CHAPTER 2 Tim nd Frequency Analysis Technique: 0.40 i re Fi 6608" "TB07GS 0.00 He y=0.00 EU 170 meee Mertz eu 2 m Dacha Le 4780. =.16.87. m8 —3 0.88 o 0.22 § 0.2 = : -0.68, +110 @ : 0:00 49:00 0 30.00 40.00" 0.00 X60.00 nsec” Y=t.00 EU siete nee Freq fret Frea (fmol Phase Freq (fmp1 Phase 90°00 1"000 3 8°03 0°00 ‘o'8 —6'69 "000 ‘a0 0:00 0:000 8:00 "000, 8:6 8:00 0-000 8:0 8:00 9000 8°0 0°00 8000 8:0 0:00 0-000 8:9 8:00 8000 8:5 9°00. 8000 8:6 6:09 8000 9:0 8:00 8:000 8:5 8:60. 8000 8:8 0:00 8000 8:0 Fig. 2-1. Single Frequency. The time signals for SAP are based on the cosine function. The reason for using the cosine instead of the sine is because the starting point at time zero of the fundamental is the highest point. The cosine is the same as the sine, except for a 90 degree phase shift. Fig. 2-1 is a cosine function with a peak amplitude of 1.0 engineering units (EU). In the real world, the signal can start at any point between 0 and 360 degrees. Fig 2-2 is a representation of a sine function with a peak amplitude of 1.0 EU. Comparing this signal to the signal in Fig. 2-1, a phase shift of -90 degrees is present. The signal in Fig. 2-2 represents the sine wave with the phase starting at zero degrees. The difference between the signals in Fig. 2-1 and Fig. 2-2 is the effective difference between taking data in the horizontal and vertical positions. The signals are 90 degrees out of phase. Cosine is used for SAP because the real part of the FFT is based on the cosine function. The following mathematical description of the FFT is given for those readers interested in the mathematics used to break the complex wave forms down into the discrete frequencies. The FFT is based on: cos (x) - jsin (3), where c0s(zx) = real component of signal J ~ imaginary variable (j? = -1) sin (2) = imaginary component of signal When the time signal is processed by the FFT, it is divided into the amplitudes and phases of the individual cosine and sine functions. When a spectrum is displayed, the plot is an amplitude plot of the frequencies. If a complex FFT is performed, both the 37CHAPTER 2 Time and Frequency Analysis Techniques Freq (fmol Phase = Freq Arpl = Phase = Freq fimpl Phase. eo"G0 1-000 "80-0 003-000 80 = 0.00 0-000 3-0 ‘0:00 6000 0:0 © 0:00 9 :000 9:0 © 8°00 0"000 8:0 0°00 0:000 9:0 = B'00 G00 8:0 9°00 0:00 oo 9:00 8.000 9:0 0:00 9.000 8:0 8:00. 8-000 a0 8:00 8.000 8:0 0:00 8-000 8:08:00 82000 a8 Fig. 2-2. Single Frequency -90 Degree Phase Shift. amplitude and phase are available, but only the amplitude is displayed. In the introduction, the frequency spectrum was said to lose the phase. In actuality, it may not be lost. It is just not listed or is discarded. Unless the phase is retained and viewed, the time domain signal must be used to identify phase relationships. To reconstruct the time signals from the frequency domain, a starting point must be selected. When time equals zero, the cosine starts at maximum amplitude and the sine starts at zero amplitude. It is more consistent to start at the maximum amplitude of all signals, so the signals add at a time equal to zero. The signal in Fig, 2-3 starts at maximum negative amplitude. This is a cosine function with a 180 degree phase shift. In rotating machinery, data taken from opposite sides of a motor should be the same, except the phases of the fundamentals should be 180 degrees apart. This obeys all laws of physics, Amplitudes of signals in rotating machinery will add or subtract, depending on their phase relationships. The top grid in Fig. 2-4 contains two frequencies at 60 Hz which are 180 degrees out of phase. The time signal in the bottom grid has zero amplitude because the two signals are equal in amplitude, 180 degrees out of phase, and one signal cancels the other signal. This is the analytical proof. “The empirical proof can be observed in balancing. Equipment is out of balance because one side is heavier than the other side. This generates a signal, as in Fig. 2-1. The equipment is balanced by placing a weight on the light side equal to the heavy place. This generates another signal 180 degrees out of phase, as in Fig. 2-3. If the two signals were equal in amplitude and 180 degrees out of phase, they would cancel, as in Fig. 2-4. In the real world, such perfection does not exist. However, it does prove that the amplitudes of out-of-phase signals subtract, 38CHAPTER 2 Time and Frequency Analysis Techniques e385 28° nor X=0.00 He ‘ye0.00 EU yy 100: 2 0: 0.00 nsce vs Freq fmpl Phase = Freq anol = Phase = Frea fmt = Phase 9°00 1°000 © f80-6 003. 0-000 0-5 =a 0-00 3-0 ‘0:00 0.000 0:0 © 900 0-000 8:0 bce” S00 0:0 9:00 0.000 9:0 000 G00 8:0 0'0u v.00 9:0 9:00 8:000 9:0 0:00 8000 8:0 0:00 3000 0:0 8:00 8:000 8:0 ©0700 82000 820 0:00 8:000 8:0 Fig. 2-3. Single Frequency 180 Degree Phase Shift. U60=16.67 msec U60=16.67 msec 0.04 0.00, -0.12 -0.20 0.00 10.00 30.00 40.00" "80.00 x:0.00 swe ¥=0.00 EU nsec Frea 1 | Phase Freq mpl Phase = Freq fnpi Phase o'8a s%6bo + 85S GISSo"Sb0 0 © Grog 0-000 ‘370 0:00 11000 ¢180:0 9:00 8.000 8:9 9:00 0.000 8. ‘0:00 07000 0:0 0:00 9.000 8:0 0:00 0.000 80 8:00 0-000 8:0 6:00 0000 0:0 © G:b0 0.000 8:0 8:00 07000 8:0 © 0:00 8000 8:0 0:00 0:00 8:0 Fig. 2-4. Two Signals 180 Degrees out of Phase. 39CHAPTER 2 Time and Frequency Analysis Techniques eu tito alee 0.22 -0.22 =0.66 et) 0.00) 10:00 20.00 30:00 ‘40-00 $0.00 0.00 nsec ye1.00 EU ne He nsec eu. [t= 16.67 mse 2.20 H 1 £ £ Sos L 10:00 20.00 30:00 v2.00 60 wee Freq (fmol Phase Freq. 90'S 0-000 S75 o00 8:60 8.000 8:0 9:00 8 8:00 8:000 8:0 8:00 8 I 8:00 8:00 8:08:00 a! 0:00 0.000 9:00. 0:000 0:0 G:00 0 Fig. 2-5. Two Signals Equal in Amplitude and in Phase. Fig. 2-6. Boiler Feedwater Pump with Imbalance. 40CHAPTER 2 Time and Frequency Analysis Techniques The top grid in Fig. 2-5 contains two in-phase signals equal in amplitude. The top grid appears to contain only one signal. The signal in the bottom grid in Fig. 2-5 is the result of signal 1 plus signal 2, and the amplitude has doubled. The cases discussed in Figs. 2-4 and 2-5 prove that the amplitudes of frequencies can subtract when out of phase, and add when in phase. This data also proves the overall vibration measured in most machines can be either over or understated. Therefore, phase must be accounted for to accurately identify the severity of a problem. This explains why "amplitude trending” is misleading. Fig. 2-6 is an example of a single frequency. The data was taken from a 3600 RPM Boiler Feedwater Pump that is out of balance. This example has some small differences from the theoretical single frequency. The real world case is not a perfect signal and shows some effects of distortion. The distortion in the time signal may not be visible to the unaided eye, but it is present. The spectral line at 60 Hz is the frequency of the time signal. The low frequency noise and the harmonics of 60 Hz are caused by distortion. In this case, the frequencies at these levels are considered insignificant because they could be caused by very minor nonlinearity in the machine or some other source. SINGLE FREQUENCY WITH HARMONICS The same physics that permit two signals to add and subtract apply to a single frequency with harmonies. A harmonic is some exact multiple of a discrete frequency. The discrete frequency, called the fundamental, is the first harmonic. The second frequency, which is two times the fundamental frequency, is the second harmonic. The second, third, fourth, etc., harmonics can be either in phase or out of phase with the fundamental. The phase relationships between the fundamental and the harmonics are valuable in diagnosing problems in rotating machines. Failure to understand and use the time signal and harmonic phase can result in diagnostic errors. A single frequency without harmonics will have one positive-going peak per time period, as indicated in Fig. 2-1. The number of positive-going peaks in one time period of the fundamental frequency identifies the highest number of true harmonics. This is true for a single frequency with harmonics only, and is true regardless of the phase relationships between the fundamental and the harmonics. The amplitudes of the fundamental and the harmonics determine the amplitudes of the positive-going peaks. However, the phase relationships of the harmonics to the fundamental determine the locations of the positive- going peaks in the signal. For example, the top grid in Fig. 2-7 contains a fundamental and second harmonic. The second harmonic is in phase with the fundamental and goes completely out of and back into phase with the fundamental once each time period of the fundamental. The bottom signal in Fig. 2-7 is the result of the top two signals adding and subtracting. It is important to note several points: 1. The amplitude of the positive-going composite signal is doubled because both signals are the same amplitude and the positive-going signals are added. 2 The amplitude of the negative-going composite signal is less because one cycle of the second harmonic is 180 degrees out of phase with the negative-going portion of the fundamental 41CHAPTER 2 Time and Frequency Analysis Techniques 40-0050: Fig. 2-7, Single Frequency with an In-Phase Harmonic. 3. The second positive-going peak is at the bottom of the fundamental, and the amplitude of the composite signal is zero. Once again, the reason for this is both signals are equal in amplitude and they subtract. 4, The composite signal is truncated on the bottom or negative side. The time period for each frequency is marked on each grid. The phase of the fundamental is changed 180 degrees in Fig. 2-8. The two signals are still in phase because the harmonics are referenced with respect to the fundamental, as can be seen in the top grid. Both signals in Fig. 2-7 start at the maximum value. In Fig. 2-8, the fundamental starts at the maximum negative value, and the second harmonic starts at maximum positive value, Itis important to note that the phase of the harmonic in reference to the fundamental is at the start of the cycle. In Figs. 2-7 and 2-8, the harmonic is in phase with the fundamental. The highest point of both signals occurs at the same point. At any other point in the cycle, the harmonic goes out of phase and back into phase with the fundamental. The net effect of the 180 degree phase shift is that the fundamental in the top grid and the composite signal in the bottom grid are shifted 180 degrees. The overall effect does not change the shape of the composite signal. The phase relationship between the fundamental and second harmonic is unchanged. In rotating equipment, the signals in Figs. 2-7 and 2-8 are the same result as could be obtained by taking data on opposite sides of a motor. That is, the fundamentals would be 180 degrees out of phase, but the second harmonics would still be in phase with the fundamentals. Again, this proves that data 42‘and Frequency Analysis Techniques U0" 16.67 30.00 40.00" "$9.00 160=1667 26.00" 30:00 ‘4070080500 Ante Hz meee fmol Phase Freq anet Phase 9-000 0-0 © 120'00 a.o00 + *"o-0 8:000 8:0 “9:60 O:000 3:0 8.000 8:0 9:00 9.000 8:0 9:000 6, 8:00 0:000 0:0 000 8:00 0%000 8:0 Fig, 2-6. Single Freq. with a 180 Degree Phase Shift and Harmonic. By act AbEBe ak ke MO 6 Yc —— 20.00 30:00 80.00 ee mee Loe. W60=16.67msec 0.00 z ; $0.00 %:0.00 see wee meee Freq anet Freq mpl Phase = Freq np! Phase. g0:69, 1000 0°00 0-000 30 "38 0-000 ‘o6 120°00 1000 00 0:00 0:0 8:00 9.000 80 0:00 07000 8:00 9:000 8:0 9:00 8.000 8:0 8:69 0:00 8:00 8.000 8:9 6:00 8.000 8:0 0:00 8:000 8:00 0:00 8:9 8:00 8.000 80 Fig. 2-9. Single Freq. with a 180 Degree out-of Phase Harmonic. 43CHAPTER 2 Time and Frequency Analysis Techniques can be taken from either side of a machine without affecting the phase relationships of the harmonics. The top grid in Fig. 2-9 contains a fundamental and a second harmonic. The second harmonic is 180 degrees out of phase with the fundamental. When this occurs, the second harmonic is in phase with the fundamental when both signals are maximum negative, and out of phase when the fundamental is maximum positive. This can be seen because the signals add when the fundamental is at maximum negative amplitude and subtract at maximum positive amplitude. The result is a higher negative amplitude in the composite signal in the bottom grid, and the two positive-going peaks are at the top. The top grid contains both signals and displays how they go into and out of phase as they did previously. However, the second harmonic subtracts from the fundamental at the start of the fundamental cycle. Please note the composite signal still has two positive-going peaks that identify the fundamental and second harmonic. However, both positive-going peaks are at the top of the signal. This can only occur if the second harmonic is 180 degrees out of phase with the fundamental. Several points concerning amplitude must be made for Fig. 2-9: 1. The zero-to-peak amplitude of the negative-going peak below zero is equal to the zero-to-peak amplitude of F, + F: 2. The zero-to-peak amplitudes of each positive-going peak are equal to the zero-to-peak amplitudes of F, and F;. 3. These amplitude relationships are true only for two signals equal in amplitude and 180 degrees out of phase. These phase and amplitude relationships hold true for linear systems. However, most real applications contain nonlinearities, called distortion. The distortion can appear in the signal as a phase shift in one or more of the harmonics. Distortion of the signal can also generate additional harmonics in the frequency domain which are not true harmonics of the signal. Therefore, the number of peaks in the time signal must be checked for true harmonic content. Continuing with phase relationships, the next step is to observe a phase shift of 90 degrees. Figs. 2-10 and 2-11 contain signals with a plus 90 and a minus 90 degree phase shift of the second harmonics, respectively. In Fig. 2-10, the composite signal still has two Peaks per cycle, denoting the second harmonic. However, instead of a discrete high and low peak, the lower peak is shifted to the left, on the downward slope of the signal. Fig. 2-11 has the same uneven peaks, except the lower peak has shifted to the right and is on the upward slope of the signal. It can also be shown that any angle between 0 and 180 degrees can be approximated by looking at the various phase shifts, For example, a positive phase shift of the second harmonic will move the lower peak to the left. As the phase approaches 180 degrees, the lower peak will approach the same amplitude of the higher peak and the signal will look like the signal in Fig. 2-9. In addition to phase shifts, the amplitudes can change. Fig. 2-12 shows the second harmonic changed to one half the amplitude of the fundamental. The starting amplitude of the composite signal is still the sum of the two amplitudes: 1.0 EU +05 EU = 15 EU. 44CHAPTER 2 Time and Frequency Analysis Techniques 0.00 100,00 0.00 20.00 60.00 80,00 “100.00 0.00 nse ve1.d0 60 rv, see Frea Phase Freq imp! = Phase Frea (Amp Phase 60.63 © ETS G16 0° bb0 75 O60 0-000 3-8 12000 < 90:0 8:00 8:000, 8:9 8:60 07000 8: 00 0:0 G00 8.000 8:8 G60 0°000 8. 9:08 8:0 Sea 8-000, ©9 Geo 0:000 8: 0:00 Cc) 8:8 0:60 0:00 a0 Fig. 2-10. Single Freq. with 90 Degree Phase-Shifted Harmonic. 20.90 40:00 60.00 80:00” 100.00 vet.do sae te meee Ant ‘000 ‘200 8: ‘000 8: 8, 8, ‘000 700 fe ©, 8: 8, 8: 888883 Fig. 2-11. Single Freq. with a -90 Degree Phase-Shifted Harmonic. 45CHAPTER 2 Time and Frequency Analysis Techniques However, the lower peak is smaller and does not have a zero amplitude. The two signals do not cancel each other because the second harmonic only subtracts one half of the fundamental amplitude when it is 180 degrees out of phase. After seeing the effect of changing amplitude in Fig. 2-12, one can identify the effect of changing the amplitudes in other ways. Changing amplitude only affects the amplitude of the composite peak. It does not affect the number of peaks or the phase relationship of the composite. The signals in Figs. 2-7 through 2-12 illustrate the effects of changing the phase and amplitude. With SAP, the amplitude, phase, and harmonic content can be identified, and these phase Telationships between the fundamental and harmonics contain valuable diagnostic information. For example, if the spectra from a machine contain only a fundamental frequency generated by rotating speed and a second harmonic that is in phase with the fundamental, the shaft may be bent. If the second harmonic is out of phase with the fundamental, the machine may be loose. If the second harmonic of gearmesh frequency is out of phase with the fundamental, the gears have a backlash problem or are oscillating, If the second harmonic is in phase with the fundamental, the gears may be bottoming out. More examples and complete descriptions of problems will be addressed in later chapters. A good rule to remember is that if the source of the harmonic is tied to the source of the fundamental, such as a fixed, geared, or bolted coupling, the harmonic should be in phase. If the source of the harmonic is not tied to the source of the fundamental, the harmonic should be out of phase. This is also discussed in more detail later. The addition of a third harmonic will now be examined, along with the effects of changing the phase and amplitude. Changing the amplitude changes the amplitude of the individual peaks, as with two harmonics. Fig. 2-13 contains a signal with three harmonics. Three positive peaks per cycle are present, indicating the three harmonics. The amplitudes of the positive peaks above zero are equal to F, +F, + Fy 1.0 EU + 1.0 EU + 1.0 EU = 3.0 EU. The high peaks occur at the fundamental frequency of 60 Hz. The lower peaks occur at the third harmonic frequency of 180 Hz. The signal in Fig. 2-14 contains only the first and third harmonics. The negative portion of the signal is more pronounced because the second harmonic is not present to subtract out the third harmonic. In fact, one cycle of the third harmonic is in phase with the negative portion, causing the two signals to add. The amplitudes of the nogative and Positive portions are equal to F, + F;: 1.0 EU + 05 EU = 1.5 EU. When the first three harmonics are distinctive, misalignment is indicated. This fact was documented several years ago by VCI. However, until recently, the differences between the many time signals that can generate the same spectrum were not evaluated. These differences can now be explained because of the phase relationships of the harmonics. The phase relationships of the harmonics can significantly alter the time signal without affecting the frequency spectrum. There are an infinite number of phase and amplitude combinations possible. Figs. 2-15 through 2-20 show representative signals that have been identified on live data. For example, Fig. 2-15 is a result of a misaligned, rigid coupling. Misalignment is indicated because of the distinct first three harmonics, which must be in phase because of the rigid coupling. If a rigid coupling is not used, the phase of the 46CHAPTER 2 Time and Frequency Analysis Technique: Teo alee 0.22 -0.22 -0.66 ioe 10.00 100.60, a0. ‘to9, 20,00 nsec meee eu Tes 0.99 0.33 8:00 -0:93 -0.99 4 . Z -1.65 z 0.00 20:00 “40:00 40:00 80:00" 10:60 H:0.00 mace YL. nee nee freq (An! | Phase Freq fmol Phase Frea nol Phase. "53 1000 + "87S -BTSS. "000 3S «GTS oS BGa os o:300 ¢ 9:6 8.e0 B.e00 8:8 8:09. 8-000 a8 8.308 89 8:60 8.000 8:8 8:00 8:000 8:8 8:008 8:0 8°68 8:20 8:88:88 a:ba0 8 o:808 8:08:60. 8:000 8:88:00 G:000 80 Fig. 2-12. Single Frequency with a Lower Amplitude Harmonic. & ‘ I> sor ier, aye Phase Fea Freq tmp! Phase 0:0 0.00 9°00 0-000 0-0 i 8:0 b:00 8:00 0.000 8:0 8:0 B:00 9:00 8.000 8:0 0:00 0:000 8:00:00 9:00 9.000 6:0 8:00 0:00 Ce) 8:00 8.000 8:0 3 Fig, 2-13. Single Frequency with Two Harmonics. 47CHAPTER 2 Time ind Frequency Analysis Techniques Freq (np! | Phase Freq @mp] Phase Freq np sorog 1000 + 87S 888. ob0 376 = 8d. oa ‘9:00 0.000 8:8 8:00 8.008 8:0 8.00 8.000 48000 O:300 + 0:0 9:09 8:000 8:5 8:60 8°000 ‘0:00 9.000 8:8 8190 a:a00 8:8 = B:e0 8:e08, 8:00 8:000 8:8 8:00. 8000 8:0 8:00 8.000, Fig. 2-14. Single Frequency with Only Third Harmonic. a8 88 a8 eee a Freq mpl Phase Freq Freq fmol Phase 96°00 4.000 00 0-0 = 000 0.000 ‘0-0 $2000 2:00 8:0 O00 9:0 9:00 0-000 9:0 Y@0"00 ¥"o00 8:0 (O00 8:0 © G00 B:000 00 0:00 000 0:0 0.00, 9:0 0:00 0-000 9:0 8:00 8:000 8:0 (0:00 8:0 © 8:00 8.000 8:0 Fig. 2-15, Single Frequency with Two Harmonics. 48CHAPTER 2 Time and Frequency Analysis Techniques 0:00) 20.00 00 $0.00 30.00” 160.00 20.00 nsec” = Yet.00 EU. nsec eu ie “: 4aa9 $ ‘40-00 60.00 80.00" 100,00 ? is ? sessed SES88 Fig, 2-16. Single Frequency with Two Phase-Shifted Harmonics. 9.00 Fig. 2-17. Single Frequency with Two Phase-Shifted Harmonics, 49CHAPTER 2 Time and Frequency Analysis Techniques 888888 eeccon Fig. 218, Single Frequency with Two Phase-Shifted Harmonics, Freq Frea 8-0 © 0.00 9.000 006.00 3-0 90:8 9:00 8-000 8:0 880 8 8:0 30:0 0:00 8-000 6°0 = 9"00 0:00 oo i 0:0 © 0:00 9000 9:0 © 8:00 0-000 a0 0:00 8000 8:0 = 0'00 0:00 8:0 6700 8:00 0:0 Fig. 2-19. Single Frequency with Two Phase-Shifted Harmonics. 50CHAPTER 2 Time and Frequency Analysis Techniques Freq fmol Phase Freq mp1 Phase = Freq Arp! Phase sor00 1000 + "G;6 —-O'00 9,000 35 © G-05 grouy oo 130/00 O'a00 $ 9:0 © 8:00 8.000 8:0 8:00 8-000 8:0 ig0:00 8/300 ¢ 190:0 9.00 0.000 8:0 9:00 0.000 8:0 ‘0:00 9:000 3:08:08 8.000 8:09:00 9.000 8:0 0:00 8:00 8:0 8:00. 8000 8:0 B00 B00 6:0 Fig. 2-20. Single Frequency with Two Phase-Shifted Harmonics. Fig. 2-21, Data from Main Turbine Generator. 51nd Frequency Analysis Techniques harmonics will change, depending on the coupling. If the coupling is rigid, it acts in a linear manner. If the coupling is loose or nonrigid, it will appear nonlinear and out of phase. More examples with specific problems are discussed later. Fig. 2-21 contains a spectrum from a main turbine generator. The time signal contains the basic features of the signal in Fig. 2-20. Three harmonics are present and the third harmonic is 180 degrees out of phase with the fundamental. Note the signal amplitude is less on the positive half than the negative half. This is a form of distortion called truncation. Truncation occurs when the machine is more flexible in one direction than in the other direction. The first three harmonics in the frequency spectrum are also distinctive. These spectral lines are the result of the three positive- going peaks in the time signal. The low level harmonics in the frequency spectrum are a result of the distortion in the time signal, and are not true harmonics, The first three harmonics are true harmonics in this spectrum. The time domain signal is necessary to verify which harmonics are true and which harmonics are caused by distortion. In this case, the distortion is caused by various nonlinearities and is not required for analysis of this problem. However, this is not always the case. CLIPPING A signal is said to be clipped when a slight amount of the positive or negative signal is flattened, as in Fig. 2-22. The upper signal is an undistorted time signal. The lower signal is clipped at the bottom. Fig. 2-23 contains a time signal and a spectrum from a motor. The time signal is clipped. Such a signal can be generated when a machine goes against a stop in one direction and cannot move further in that direction for a small period of time. As the cycle continues, the machine moves away from the stop in a relatively linear manner. The signal is distorted because the time period for the negative and positive portion is not the same. Clipping is also a form of distortion. The frequency spectrum contains very little harmonic content, because in order for harmonic content to be generated, the signal distortion must be repeatable. In this case, the signal distortion is not repeatable. The nonrepeatable distortion is noise. The "skirts" on either side of the 29.6 Hz spectral line, predominately on the low side, contain the noise. The spectrum is up off the baseline to the left of the spectral line, which indicates looseness or random motion. ‘Square Wave A special case to note is a single frequency with only odd harmonics present, as in Fig. 2-24. The harmonics tend to cancel each other out, except for one positive and one negative peak per time period of the fundamental. The peaks have an amplitude equal to the sum of all the amplitudes added together. A special case of odd harmonics is shown in Fig. 2-25. Fig. 2-25 has only odd harmonics, and every other odd harmonic is 180 degrees out of phase. The resultant signal is a square wave. The amplitudes correspond to the amplitude of the fundamental divided by the harmonic number. 52CHAPTER 2 Time and Frequency Analysis Techniques aS V/ 0:00 °°" “400.00 FAL 0.00" "400.00 Stonal clipped at -0.40 Phase Freq fnpl Phase ‘o-3 © b.09. o000 od 8:5 8:00 0:00 0:0 8:3 6.00 G00 0:0 8:5 8:00 @:000 8:0 8:3 © 8!00 000 8:0 Fig. 2-22. Clipped Signal. IPs 2.95 177 > : — 0.59 : 0.59 . i i0 0.00 Hz Y=0.00 EU A/8= nsec. Hertz Fig. 2-23. 1800 RPM Motor Belt Driving a Fan. 53CHAPTER 2 Time and Frequency Analysis Techniques Fig. 2-24. Single Frequency with Only Odd Harmonics. Fig. 2-25. ‘Single Frequency with Only Odd Harmonics. 54CHAPTER 2 Time and Frequency Analysis Techniquos For example: Harmonic Amplitude 1 10.0/1=10.000 3 10.0/3=3.333, 5 10.0/5=2.000 Note that the signal is not exactly square, but has ripples. This is due to the fact that a limited number of odd harmonics are contained in the signal. A true square wave contains all odd harmonics, which cannot be truly simulated on a computer as a sum of the cosine functions. It may appear impossible for a piece of equipment to generate a square wave, but it is possible to generate a signal with square wave features. This can occur in a motor if the motor has a loose mount. If the mounting bolts are loose, the motor will tend to move up and down. If the motor moves up and is stopped by the mounting bolt, and then moves down and is stopped by the motor support, a square wave can be generated. If clipping occurs on both the top and bottom of a signal and the clipping is significant, the result will resemble a square wave. NATURAL FREQUENCIES A special case occurs when natural frequencies are excited. A discrete frequency is not Present in such cases because the natural or excited frequency is a band of frequencies. In this situation, a band of spectral lines occurs. ‘The bandwidth is determined by damping, and the difference frequency between the spectral lines is the source of excitation. A high degree of measurement accuracy is required to determine if a given frequency is an excited frequency (natural) or a generated frequency (events times speed). Amplitude modulation can occur between two generated frequencies, or an excited frequency can be modulated by the source of excitation, ic., a generated frequency. When a generated frequency equals an excited frequency, other tests and measurements must be employed to identify the excited and generated frequencies. These techniques will be discussed in following chapters. MULTIPLE FREQUENCIES - LINEAR SYSTEMS, High Frequency Riding a Low Frequency When two independent frequencies are present in a linear system, they cannot add together in amplitude or frequency. When this occurs, the two frequencies mix, and the high frequency will ride the low frequency, as in Fig. 2-26. At first glance, amplitude modulation appears to be present. However, close inspection indicates that the two envelopes enclosing the peaks of the time signal are in phase. This means when the positive peaks of the signal move upward, the negative peaks of the signal move upward. The time period for one cycle of the envelopes is 50 milliseconds or 20 Hz: Fei-_1 oon T 50 msecCHAPTER 2 Time and Frequency Analysis Techniques et 8000 Pa F020 788 1ke nsec Hertz i ‘2800080800 i 8 < a ae Freq amp Phase = Freq anol fmpt Phase 26°00 1000 ‘3°0 ©—800 0-000 ‘000 rd 10000 4/000 8:0 800 8.000 ‘800 8:0 ‘0:00 0.000 9:0 © 000 0-000 ‘800 9:0 8°00 "000 8:0 000 0"000 000 9:0 8:00 0.000 0:0 = 800 "000 ‘000 0:0 Fig, 226. High Frequency Riding a Low Frequency. which is the low frequency. Five cycles of the signal are contained in each cycle of the envelopes, and the time period for each cycle of the signal is 10 milliseconds or 100 Hz: 1 1 T 10 msec = 100 Hz which is the high frequency. The peak-to-peak amplitude of each cycle is two times the zero-to-peak amplitude of the 100 Hz frequency: 2 x 4.0 EU = 8.0 EU. The peak-to-peak amplitude of each envelope is two times the zero-to-peak amplitude of the 20 Hz frequency: 2 x 1.0 EU = 2.0 EU. The zero-to-peak value of the highest envelope peak is the amplitude of the 100 Hz frequency plus the amplitude of the 20 Hz frequency: 4 EU + 1 EU = 5 EU. The lowest point of the envelope is the amplitude of the 100 Hz frequency minus the amplitude of the 20 Hz frequency: 4.0 EU - 1.0 EU = 3.0 BU. Since harmonics are not present and the high frequency is riding the low frequency, there is not, a cause-and-effect relationship. The two signals are generated independently. It is important to note that a high frequency which is an exact multiple of a low frequency will cause the amplitude of the high frequency peaks to be the same in each period of the low frequency. A high frequency that is not a multiple will cause the amplitude of the high frequency peaks to vary during each period of the low frequency. MULTIPLE FREQUENCIES - NONLINEAR SYSTEMS, ‘Two or more independent frequencies can be mixed together if a machine has some form of nonlinearity or other problem. There are many forms and degrees of frequency mixing. Examples of amplitude modulation, sum and difference frequencies, puises, and frequency modulation are discussed in the following sections. 56CHAPTER 2 Time and Frequency Analysis Techniques Amplitude Modulation Amplitude modulation occurs when two frequencies are added together algebraically. Frequencies will not add in a machine that behaves in a linear manner. Therefore, a problem must exist before amplitude modulation can occur. There are several forms of amplitude modulation; one form is a beat. A beat occurs when the amplitudes of two frequencies are added together. Assume there are two frequencies of 29.6 Hz and 25.6 Hz, as in Fig. 2-27. ‘When the amplitudes of the two frequencies go into phase, they add together. Then, as, the two frequencies go out of phase, the amplitudes subtract until they are 180 degrees out of phase. The two frequencies continue to go into and out of phase, forming a time- varying amplitude signal called a beat. There are similarities between a high frequency riding a low frequency and amplitude modulation. The maximum zero-to-peak envelope amplitude is equal to the amplitude sum of F, + F; 0.6 EU + 0.15 EU = 0.75 EU = zero-to-peak envelope maximum ‘The amplitude difference of F, - F, is the minimum zero-to-peak envelope amplitude: 0.6 EU - 0.15 EU = 0.45 EU = zer0-t0-peak envelope minimum The time period for each cycle in the time signal is 33.9 milliseconds, which is the 29.6 Hz frequency. 920.00" "40060 eq mpi Phase ‘90 0-000 0-0 ‘00 0000 8:0 "00 0000 0:0 a ‘00 0000, 8:0 ‘000 8000 ‘bo 0000, 0:0 Fig. 2-27. Beat-Amplitude Modulation of Two Frequencies. 37CHAPTER 2 Time and Frequency Analysis Techniques 1 29.6 Hz = 0.0339 sec The envelope frequency, however, is not the 25.6 Hz frequency, as it was for a high frequency riding a low frequency. For amplitude modulation, the envelope frequency is the difference between the two frequencies: F, - F, = 29.6 Hz - 25.6 Hz = 4 Hz. Fig. 2-27 contains a complete cycle of the difference frequency of 4.0 Hz. The envelope occurs at a time period of 250 milliseconds: 1 40 Hz tT = 0.25 sec = 250 ms Amplitude modulation is the result of a cause-and-effect situation and the amount of modulation can be expressed as a percentage. The frequency spectrum identifies the percent modulation as the amplitude of the lower frequency divided by the amplitude of the higher frequency: 0.15 EU 0.60 EU” 0.25 or 25% The percent modulation in the time signal is expressed as the maximum envelope amplitude (M,), where M,= F, + F,= 0.6 + 0.15 = 0.75 minus the minimum envelope amplitude (M), where M,= F, - Fp = 0.8 - 0.15 = 0.45 divided by the maximum envelope amplitude plus the minimum envelope amplitude: M, - M, _ 0.75 - 0.45 . M,+M, 075+ 045 12 Some comparisons between a high frequency riding a low frequency and amplitude modulation are: 1. The envelopes at the top and bottom of the signal are in phase for a high frequency riding a low frequency. The envelopes are out of phase for amplitude modulation. 2 For a high frequency riding a low frequency, the zero-to-peak amy of the envelope is equal to the amplitude of the low frequency. For amplitude modulation, the zero-to-peak amplitude of the envelope is also equal to the amplitude of the low frequency. The zero-to-peak amplitude of the highest peak is equal to F, + F,, and the zero-to-peak amplitude of the lowest peak is equal to F, - F; for both a high frequency riding a low frequency and amplitude modulation. 3. In rotating machinery, a high frequency riding a low frequency there is 58CHAPTER 2 Time and Frequency Analysis Techniques no cause-and-effect relationship. There are two independent frequencies and two independent problems. In amplitude modulation, there is a cause-and-effect relationship because there are two or more frequencies and only one problem. 4. When a high frequency rides a low frequency, the reciprocal of the envelope time period is equal to the low frequency. In amplitude modulation, the reciprocal of the envelope time period equals the difference between frequency 1 and frequency 2 (F, - F). When two frequencies beat together, such as in Fig. 2-27, there is a cause-and-effect relationship. Therefore, two frequencies are present, but only one problem exists. The 29.6 Hz frequency is the carrier frequency and is the effect of the problem, i.., the motor is loose. The 25.6 Hz is the modulation frequency or modulator and is the speed of the fan driven by the motor. The only way the fan can shake the motor, in this case, is when the motor is loose. Looseness in the motor allows the motor to shake, and the two frequencies to beat. Amplitude modulation also occurs when two frequencies are not exact multiples. Fig. 2- 28 contains two frequencies with second harmonics. The two frequencies are almost the same, but not exact. The signal seems to be constant. However, Fig. 2-29 contains the same signal with a longer time period, and it is clear the signal is changing. Fig. 2-30 contains a full beat and amplitude modulation is evident. The difference between frequency 1 and frequency 2 is 0.2 Hz. Therefore, one beat takes 5 seconds to complete. As the frequencies get closer together, the time to complete one beat becomes longer. Therefore, it is extremely important to determine whether the frequency is a harmonic or is just close to the harmonic frequency. The beat in Fig. 2-30 Freq fmol Phase Freq anol Phase $0.60 1°000 ‘3 0:00 0-000 rd 130:00 1000 0:0 8:00 000 8:0 ‘$0.20 4:000 o0 8:00 9:000 8:0 120/40 1000, 00 800 0000 0:0 0,00 0.000 a0 0:00 0:00 0:0 Fig. 2-28. Two Similar Frequencies with Second Harmonic. 59CHAPTER 2 Time and Frequency Analysis Techniqui Fig. 2-29, Two Similar Frequencies with Second Harmonic. Heres 2 f | bb E 983 88 98 g 23 833 -2.64 $000 4000 2000 ‘3000 ‘4000 4.40 0.00 nse ca ine x: $88883 788888 388888 peeees 88893 Fig. 2-30. Beat of Two Similar Frequencies. 60CHAPTER 2 Time and Froquency Analysis Techniques is a common signal from a 3600 RPM motor that has an electrical problem. The motor speed is 60 Hz minus the slip frequency, when under a load. Line frequency is 60 Hz. In this case, the motor speed and line frequency must be examined with enough resolution to identify the true problem. SUM AND DIFFERENCE FREQUENCIES Another type of amplitude modulation occurs when one component is eccentric. A frequency of 100 Hz with a sum and difference frequency of 4 Hz is presented for analysis. The 100 Hz carrier frequency is contained in Fig. 2-31. One example of sum and difference frequencies is gear eccentricity. When one gear is eccentric or out-of- round, the amplitude of gearmesh frequency increases when the high place or places go into mesh. If the gear has only one high place, the signal amplitude will be higher once each revolution. In either case, amplitude modulation is caused by the eccentric gear. The associated spectra contain a spectral line at gearmesh frequency with sidebands of gear speed. If the gear has more than one high place, then the difference frequency between the gearmesh frequency and the sidebands is equal to the number of high places times the speed of the problem gear. If two high places are present, the difference frequency is two times gear speed. “Three high places would generate a difference frequency of three times gear speed, four high places would generate four times gear speed, ete. Assume the 100 Hz frequency is gearmesh frequency and gear speed is 1 Hz, if the gear has 100 teeth. If the gear has four high places or eccentricities, four times gear speed is 4 Hz. Fig. 2-32 shows the 100 Hz frequency with the sum frequency of 104 Hz: 100 Hz +4 Hz = 104 Hz. This signal has amplitude modulation. The carrier frequency is the gearmesh frequency of 100 Hz. The modulator or cause of the problem is the 4 Hz frequency of four times gear speed. The envelope frequency is 4 Hz and is noted in Fig. 2-32 as 250 milliseconds. One hundred cycles of gearmesh frequency occur each revolution of the gear. The modulation occurs at 25 cycles of gearmesh frequency or 4 times per revolution of the gear. The modulation is 10 percent. Fig. 2-33 contains the time period for one revolution of the gear. Four modulations can be observe 1 THz = 1 sec = 1,000 msec If the eccentric gear has not caused looseness, sidebands will occur at gearmesh frequency plus gear speed or multiples of gear speed. In other words, the sidebands will be on the high side of gearmesh frequency. The frequencies add, in this case, because the phase relationship between the carrier and the modulator is constant. As stated earlier, the machine is behaving in a linear manner. Fig. 2-34 contains the 4 Hz as a difference frequency of 96.0 Hz: 100 Fiz - 4 Hz = 96.0 Hz. The time signal appears similar to Fig. 2-32. When a gear or geared shaft system is loose, the looseness causes the modulator to subtract from the carrier because the two frequencies are out of phase. When the two frequencies are in phase, they add. Looseness causes an out-of-phase condition. Eccentricity is an in-phase condition. 61CHAPTER 2 Time and Frequency Analysis Techniques Fig. 2-31, 100 Hertz Carrier Frequency. Heres H | | a iNCHAPTER 2 Time and Frequency Analysis Techniques 00 ———160-00—200-06 Herts (00.06 400.00 00.00 ‘80:00 11000 0.55 ane Phase Fig. 2-33. One Revolution of Gear with Four Eccentricities. 0.00 1100:.00 260.00 ‘300.00 '400:00° "$00.00 50.00 nsec ¥20,55 EU ae He Freq fupl Phase mp1 Phase 10000 9-500 Oo 0-000 9. 96:00 B:080, 0:0 8000 8: ‘0:00 9:000 ° ‘9:000 9. 8:00 8.000 6 0:000 8: 8:00 0:000 ° 0:00 8; Fig. 2-34. 100 Hertz with 96 Hertz Difference Frequency. 63CHAPTER 2 Time and Frequency Analysis Techniques © ‘00 : : 0190 fone a —— 0.60 ©. ‘ srlctlongussneeeor ‘ 0.40 ‘ 0.20 ” - $ 0.00 Sc 00" a0 00 80-00 130707 165-00" "200-00 ‘X=0.00 Hz Y=0.00 EU 1K= nsec Hertz Bg O65*0. OB oy RCO AMOAA.EY , ovo RENE fh i 0.13 AKA i -o.t3 Hl I 0.13 Vy -0.0 EE | 0-88 500 ‘30060 40000" ‘500-00 0.00 nid ace Eres tog) Phase Freq _ampt phase Frag amp! Phase. 10038 oS STS oN ™S | OTRE ONG] ES Po a ne ee sem Geo 6 8.0 88 8:83 88 Sm oom 88 soo Sooo 68 Ba 8s 8B Be Bem E8 ove Seas) 88 Bes Sass) 88 we ote 88 Fig. 2-35. Sum and Difference Frequency with No Phase Shift. Fig. 2-36. Sum and Difference Frequency with Phase Shift. “4CHAPTER 2 Time and Frequency Analysis Techniques. Fig. 2-35 contains both the sum and difference frequencies. The beat is now more pronounced and the modulation is 20 percent: 06-04 02. 06+04 107% This occurs when both the 96.0 Hz and 104.0 Hz sidebands are in phase. Fig. 2-36 contains both the sum and difference frequencies, but the 96.0 Hz difference frequency is 180 degrees out of phase. In this case, the modulation is zero percent. Modulation on the high side, or a sum frequency, indicates a linear problem. Low side modulation, or a difference frequency, indicates a nonlinear problem. Therefore, because of the nonlinearity, the phase of the difference frequency will be changing. However, the sum. frequency will maintain its phase relationship. If gear eccentricity has caused looseness (nonlinearity) associated with the problem gear, sidebands can occur on both sides of gearmesh frequency. If looseness is the more severe problem, the amplitudes of the sidebands will be higher on the low side of gearmesh frequency. If eccentricity is the more severe problem, the amplitudes of the sidebands will be higher on the high side of gearmesh frequency. If looseness is the only problem, then the sidebands occur only on the low side of gearmesh frequency. For example, suppose a gear was mounted on a shaft rotating at 29.6 Hz, gearmesh frequency was 1280 Hz, and the bearings were loose in the housing. Gearmesh frequency would be present and one sideband could occur at 1160 Hz: 1280 Hz - (4 x 29.6 Hz) = 1160 Hz, Four times running speed is an indication of slippage, so 4 x 29.6 Hz is generated. The frequencies subtract when looseness is present because the phase relationship between the signals is not constant, which means the machine is acting in a nonlinear manner. This technology is not new. It has been used for years to identify moving targets with radar, and for space object identification to determine the shape as spherical, cylindrical, triangular, bell-shaped, etc. It has also been used to identify the way an object is moving, i.e, amount of spin, tumble, pitch and yaw. However, these principles had not been used for rotating machinery until the author applied the technology. The principles described for gearmesh frequency apply to other generated frequencies such as blade or vane pass frequencies, bearing frequencies, frequencies from multiple defects, and frequencies from bars or corrugations on press rolls. PULSES A pulse is caused by a hit or an impact. Pulses fall into one of three categories: empty pulses, frequency modulation, and amplitude modulation. Fig. 2-37 contains an empty pulse. This occurs when excited or generated frequencies are not present. It is called empty because it contains no generated or excited frequencies. A pulse is identified by a series of spectral lines. The repetition rate of the pulse is equal to the difference frequency between the spectral lines. The empty pulse has a low level spectral line at shaft speed, and the amplitude increases with each succeeding harmonic. Frequency modulation is a time-varying frequency. This frequency modulation can appear as a series of bursts or beats, as contained in Figs. 2-38, 2-39, and 2-40. Generated 65CHAPTER 2 Time and Frequency Analysis Techniques 0.22 | 0.53 i 0:420 0:120 02420 Fig. 237. Empty Pulse. zi Xe0.00 Hr ‘y=0.00 £0 $e iy 60:00 e282! F 38 a so @ Fig. 2-38. Generated or Excited Pulse. 66CHAPTER 2 Time and Frequency Analysis Techniques or excited pulses are usually caused by a once-per-revolution impact or excitation. Fig. 2-38 contains a series of spectral lines. The pulse repetition rate is equal to the difference frequency between the spectral lines, or 500 milliseconds. This would be similar to an impact once per revolution. Fig. 2-39 is a representation of a felt problem. The pulse or bad place in the felt is exciting a natural frequency of the loading diaphragm. The pulse repetition rate is 0.5 Hz, or 2 seconds, which is once per revolution of the felt. The signals in these figures contain characteristics of both amplitude and frequency modulation. Again, itis impossible to determine the character of the time signal from the frequency spectra. FREQUENCY MODULATION, Frequency modulation is a time-varying frequency, as opposed to amplitude modulation, which is a time-varying amplitude. The lower frequency is the carrier, and the higher frequency is the modulator. The modulator is normally an excited frequency, and the source of excitation is normally the speed of the rotating unit. Fig, 2-40 is an example of frequency modulation. The low frequency carrier is 10 Hz, the excited high frequency is 50 Hz, and the source of excitation is the 2 Hz difference frequency. The high frequency bursts occur at 2 Hz intervals. This could occur in a pump when pump speed is 2 Hz, and the impeller has 5 vanes. The 10 Hz carrier frequency is vane pass frequency. The problem or source of excitation is one impact per revolution of the impeller. If one vane hit the housing, the 50 Hz frequency could be the natural frequency excited by the impact. Frequency modulation can be a series of high frequency bursts similar to a pulse, or the high frequency can occur periodically with a low frequency. Since the frequency response 5 41/30 0.440, Fig. 2-39. Pulse Exciting Natural Frequency. 67CHAPTER 2 Time and Frequency Analysis Techniques Sp ge fern RR. 0100708 Heres Freq (Amp Phase Freq mp1 Phase Freq amp! Phase 38°00 0" 700 30 © s2"00 0" 300 ‘73 sa'00 0-400 ‘3 48:00 0.400 8:0 46:00 O!090 8:0 44:00 0.090 80 56.00 0.000 970 $000 0.060 9:0 60:00 0050 oo 42/00 0060 8:0 4000 8:0s0 8:0 “0:00 9000 oo 46:00 0400 8:0 “0:00 8000 8:0 000 0000 ca Fig. 2-40. Frequency Modulation. of an accelerometer is best at high frequencies, such problems may be best measured in acceleration. Frequency modulation occurs most often in impacts, such as defects on the inner race of cylindrical roller bearings, or when two shafts are rotating very close to each other. Frequency modulation can occur in screw compressors, vacuum pumps, and blowers when one shaft is bent enough to permit an impact once each revolution. One last comparison should be noted to clarify the differences between a high frequency riding a low frequency, amplitude modulation, and frequency modulation. 1. High frequency riding a low frequency - No looseness is present. High and low frequencies may be exact multiples of each other. No mixing of signals occurs. Changes in the phase have little or no effect. 2. Amplitude modulation - High frequency is the carrier; low frequency is the modulator. Signals go into and out of phase. 3. Frequency modulation - Low frequency is the carrier; high frequency is the modulator. CONCLUSION All machines obey the basic laws of physics. Therefore, vibration problems are repeat- able and can be identified. Both the time signal and frequency spectra must be analyzed to accurately identify problems. This was proven both empirically and theoretically. ‘The basics of time domain analysis were discussed and used to explain the frequency spectrum. Single frequencies and single frequencies with harmonics were explained to 68CHAPTER 2 Time and Frequency Analysis Techniques {identify the effects of amplitude changes and phase relationships. Using these basic principles, multiple frequencies were introduced to identify linear and nonlinear problems for more complex analysis. Multiple frequencies include high frequencies riding low frequencies, amplitude modulation, beats, pulses, and frequency modulation. It was also noted that if there is a cause-and-effect relationship in the signal, a problem exists in the equipment. The fundamental concepts in this chapter are the foundation for understanding and explaining the proceeding chapters. Using these fundamental concepts, all time and frequency relationships can be understood. However, with the many combinations of phase shifts and amplitude changes, each case needs to be analyzed to identify all the information contained in the vibration signal. Analysis can be complex and time consuming, but the use of SAP simplifies the analysis. The result is a complete and more accurate diagnosis of machinery vibration problems. Co)CHAPTER THRE! HARDWARE AND SOFTWARE REQUIRED FOR ACCURATE DIAGNOSTICS HARDWARE INTRODUCTION ‘The improvements in hardware, software, and technology for accurate diagnosis of machinery problems over the past twenty years is unparalleled in history. In 1973, when this author first became involved in diagnosing rotating machinery problems, some people were still using a piece of sharpened soapstone for balancing rotating equipment. Hand-held vibration meters were widely used to measure the vibration amplitude. These amplitude readings were recorded and plotted on graph paper to indicate trends. A variety of balancing instruments were also available. These instruments employed hand- tunable filters. Some frequencies could be identified, however, the accuracy was not very good because the filter "Q" was of low quality. Software was nonexistent, and hand-held calculators were being introduced at a cost of several hundred dollars. The technology for diagnostics was in initial stages of development. Some of the more sophisticated engineers were using an oscilloscope to view the time domain signal. Not very many problems were identified and equipment often failed. In the next few years several companies introduced real-time analyzers (RTA) that could produce a frequency domain spectra from the time domain signal using a Fast Fourier ‘Transform (FFT). This was a giant leap forward, however, some problems still existed. For example, these units had only 256 lines of resolution. A frequency translator or zoom was not available, and the time signal displayed was analog and could not be synchronized. Technical papers on how to identify defects in antifriction bearings and other problems were being published. On-line condition monitoring systems using proximity probes were being introduced. Around 1980, several manufacturers introduced new real-time analyzers. These new instruments had 400 lines of resolution, a frequency translator/zoom, and a digitized time signal. This improved equipment provided the capability for accurate diagnostics. However, the technology for accurate diagnostics was just emerging. These new analyzers with the improved frequency spectra caused some people to stop using the time domain signal. In retrospect this was a mistake, because now we know the time signal is required for diagnostics. ‘The personal computer (PC) was making its debut and the RTA could be controlled with a serial port. In 1980 the author introduced the first expert system. This system uscd a SD-345 real-time analyzer, an HP-85 desktop calculator, a Nagra IV scientific tape recorder, and software that could diagnose defects in antifriction bearings. The system was very accurate, however it was very slow. The HP-85 had 16k of memory and for $795 you could add another 16k memory module. These were limitations that required too many operator entries. This system required recording one data point for 1.5 minutes, and it took about 15 minutes to produce a fully annotated spectra. Today, we record four data points for 30 seconds and can produce almost three fully annotated spectra in one minute. ‘The 1980's were very dynamic in the predictive maintenance field. New and more powerful analyzers were introduced in rapid succession. Also, a lot of less powerful instruments such as shock pulse meters and data collectors were being introduced. The data collectors were the big success story of the 1980’s. These instruments were used with 1BM compatible computers and had sophisticated databases, with capabilities of up and down loading data, and trending with associated worker programs. These computerized predictive maintenance systems are widely used today even though the diagnostic ability does not match that of the real-time analyzers. The weak technology for diagnostics gaveCHAPTER 3 Hardware and Software Required for Accurate Diagnostics ise to several new concepts such as the Cepstrum, which is a spectrum of a spectrum. Spike energy, envelope detection, demodulators, and many new versions of the shock pulse meter. Each of these concepts/instruments provided varying degrees of success, however they did not provide the total solution. More powerful real-time analyzers were designed and built to perform specific tasks. If a new task or function was required, the RTA required modification. This caused some RTAS to be obsolete in a few short years. ‘The current state-of-the-art RTA is a multi-channel signal processor. This signal processor is controlled by a personal computer. The major advantage of this arrangement is that it uses the power of the PC to perform complex calculations, to manipulate data, and to store, recall, and display the data in a user friendly format. If additional tasks or functions are required, they can be provided with software changes or additions. This is more cost effective, and delays obsolescence. Today, we are seeing A to D converters for the new Personal Computer Memory Interface Adaptor (PCMCIA) slots on notebook computers. In the not too distant future, we may see an FFT signal processor on the PCMCIA cards. Arguments and discussions of the advantages and disadvantages of the data collectors and real-time analyzers may continue for some time. It is also true the data collectors of today are very much improved from the ones originally introduced. With the trend toward smaller, faster, and more powerful, the data collectors may someday achieve the power of a real-time analyzer. However, it will still be difficult to get the engineer or technician to collect enough data at the equipment location for accurate diagnosis because of the hostile environments. Fig. 3-1. Personal Computer. Personal Computer Fig. 3-1 contains a photo of the current state-of-the-art personal computer. The PC’s of today are a marvel over those produced two or three years ago. Today, RAM of up to 16 megabytes is common. Hard disk drives with 500 megabytes are the standard. The capability to add stand-alone disk drives provides virtually unlimited disk storage. The use of cache systems and DSP chips with a clock speed of 66 MHz, processes data at a very high speed. The introduction of the new Pentium chip will make parallel processing, and multi-tasking in the personal computer a reality. 72CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Fig. 3-2. Tektronix 2630. Real-time Analyzer The real-time analyzer has enjoyed significant improvements as well. The current state- of-the-art system employs an instrument with a multi-channel signal processor, and a signal generator output. Fig. 3-2 contains a photo of the Tektronix TEK 2630, 4-channel analyzer. These units are controlled by a PC, and the PC screen is used to display the data. The operating software is stored in the computer. If a new requirement is identified, new software can be developed to satisfy the requirement. This can often be accomplished without a hardware change. The following features are required for a real- time analyzer: 1. Anti-aliasing protection. This feature prevents false signals from being presented in the data. 2. A flat frequency response from DC to 20 KHz. Operator selectable frequency range from 5 Hz to 20 KHz in a1, 2, 5 sequence. A. zoom/frequency translator with a user-selectable center frequency and bandwidth. 5. Variable lines of resolution of up to 1600. The lines of resolution determine the measurement accuracy. Frequency range divided F, by lines of resolution equals measurement accuracy. My = 7 A real-time rate of at least 10 KHz. This means the analyzer can perform an FFT on the signal in about the same time it takes to fill the memory period on the 10 KHz range. 7. Sampling rate. Physics requires the data to be sampled at 2.56 times the highest frequency. This sample rate prevents the loss of data, and a higher sampling rate does not improve the data.CHAPTER 3 Hardware and Software Required for Accurate Diagnostics. 8 A good time domain signal. The digitized time domain signal is required because of stability. As the machine zero motion point is zero on the time signal, phase relationship between the various signals can be observed. Truncation, peak amplitude of each cycle, amplitude modulation, frequency modulation, pulses, and beats are also observable in the time domain signal. The use of the time domain signal requires some explanation. The amount of time displayed is determined by the frequency range used. The length of time displayed is the reciprocal of the measurement accuracy. For example, on the 200 Hz range using 400 lines of resolution, the measurement accuracy is 20 = 05 Hz. The frequency 1 = 2.sec. This means on measurement accuracy is 0.5 Hz. The reciprocal of 05 is 05 the 200 Hz frequency range, two seconds of time data would be displayed. This time period is also called the memory period. If you are using a 200 Hz frequency range, frequencies above 200 Fiz will not be observed in time data displayed. The measurement accuracy in frequency is improved by using lower frequency range and or using more lines of resolution. Measurement accuracy is improved in the time domain by using less time. Less time is obtained by using a higher frequency. Let’s look at a couple of examples using 400 lines of resolution, which is a 1024 data point transform. = 1.95 ms. This means that the data contains 1024 data points and that each of the data points are 1.95 ms apart. Time could then only be measured accurately within 1.95 ms. This may not carry much meaning for some readers. The following explanation may be helpful. Suppose we wanted to measure a frequency in the time signal of 29.6 Hz. The time period for one i § ~ 0.0897838 sec Then 0.0397838 sec - 33.76378 ms 33.78378 ms and “1.95 ms measurement accuracy is based on data points, and 33.78378 ms does not fit evenly between any two data points, We cannot measure fractions of data points, so we must measure either 17 or 18 data points. Therefore 17 x 1.95 = 33.15 ms, and 18 x 1.95 = 35.1 ms are the theoretical measurements. 4. 30.18 Hz and —1 ———— 1 ___ . 99.49 Hz “9.03315 sec 0.0951 seo ~ 2849 He ‘Therefore 29.6 Hz would be measured as 30.16 Hz in the time signal. Using the above example, 2 sec = 2000 ms, and ze cycle is = 17.32 This means that 33.78378 ms cannot be measured because To measure 29.6 Hz in the frequency spectrum the process would be similar. We will use 200 Hz the 200 Hz fi ind 400 lines of resolution, -200 #2 _ 9.5 yz This 1e same frequency range ai ines of resolutior f Iz This is the distance between each data point in the frequency spectrum. m8 PE ~ 592 This 5 He means that 29.6 Hz cannot be measured exactly. We would have to measure using 59 or 60 data points. Each data point is 0.5 Hz so that 59 data points would give you 59 x 0.5 29.5 Hz, and 60 data points would give you 60 x 0.5 Hz = 30 Hz. In this case the 29.6 Hz signal would be measured as 29.5 Hz. In the above example, the 29.6 Hz frequency is measured more accurately in the frequency domain. 74CHAPTER 3 Hardware and Software Required for Accurat Diagnostics Now suppose we have the same frequency, 29.6 Hz, on a 2000 Hz frequency range. Then the measurement accuracy of the frequency spectrum would be 20 He = 5 He. Since 20.6 Hz 5 He could be theoretically measured as 6 x 5 Hz = 30 Hz on a 2000 Hz range in the frequency spectrum. In the time signal, the measurement accuracy would be ~ 5.92 Hz, the closest number of data points is 6. This means that 29.6 Hz Now each data point in the time signal is only 0.195 ms apart. To measure 29.6 Hz: 1 33.78378 ms = 003378378 sec - 33.78378 ————— = 172.97 208 He 8 38 ™ 0.95 ms The closest number of data points is 173. This means that 29.6 Hz or 33.78378 ms could 1 be theoretically measured as 173 x 0.195 ms = 33.735 = \eoretically as 173 x ms ms or Eas = 296 He on a 2000 Hz range in the time signal. In this case the 29.6 Hz could be measured more accurately in the time signal. The frequency range was increased from 200 Hz to 2000 Hz, which improved the measurement accuracy in the time domain by using less time. By contrast, the measurement accuracy in the frequency spectrum decreased from 0.5 Hz to 5 Hz with the increased frequency range. This supports the earlier statement that measurement accuracy is improved by using a lower frequency range in the frequency domain and a higher frequency range in the time domain. The screen on the personal computer, data collector, or real-time analyzer also affects measurement accuracy. The data points on the screen or display are called pixels. The industry standard for the horizontal axis is 600 pixels. Regardless of the lines of resolution in frequency, or the data points in time, the display will plot 600 points across the screen and connect each point with a line. Increasing the lines of resolution above 600 has little effect on measurement accuracy unless an expansion or zoom is used. 9. The three types of averaging required are: arithmetic mean averaging, peak hold, and synchronous time averaging. ‘The arithmetic mean average is equal to the summation of the x’s over N. M, = We This is the type of averaging used most, however, it is not often understood. This type of averaging takes the average power contained in a frequency during one memory period for each average. If eight (8) averages are taken, then the average power in each frequency in each memory period is added together eight times and the sum is divided by eight. Since averaging enhances the signal, sixteen (16) averages may be required to ensure all frequencies are obtained. However the enhancement obtained by the last eight averages is small, and most people agree that eight averages is enough for most applications. If the signal contains a frequency during the complete memory period, and each memory period contains the same frequency, then the result of the averaging is satisfactory. However, if the memory period contains the frequency for a short time period, pulses for example, then the energy can be understated by a factor of five or more. Some examples aCHAPTER 3 Hardware and Software Required for Accurate Diagnostics are pulses generated by a broken tooth on a gear, some inner race defects, and defects on rollers in antifriction bearings. In these cases, the amplitude must be obtained from the time signal and/or peak hold averaging. Peak hold averaging takes the peak amplitude for each frequency in the memory period and writes it to the screen. Then, each time the amplitude of a frequency is higher, the screen is updated with the higher amplitude. This type of averaging is used as indicated above, for coast down data, when determining speed variations of machines, and bump tests. Synchronous time averaging employs a trigger from some source to start filling the memory period at exactly the same point each time. Often some event or pulse in the time signal can be used as a trigger. This type of averaging employs a comparator method and all frequencies that are in phase with the previous memory period are retained. All frequencies that are out of phase with the previous memory period are canceled according to the phase difference. Since different frequencies go into and out of phase with each other, the frequencies could be in phase part of each memory period. However, the frequencies are out of phase most of the time and will eventually average out to a point of little or no concern. This is why the synchronous to nonsynchronous data is improved by the square root of N. Therefore SNR « ¥NV, when N equals the number of averages. DATA COLLECTION Accurate diagnosis of rotating machinery problems requires three or more frequency spectra and one or more time domain signals. The route /database of some data collectors allows only one frequency range per point. The slow speed (500 Hz real-time rate) of some data collectors and unpleasant work environment of some machines make it difficult to get a human to collect enough data for accurate diagnosis. Fig, 3-3. Sony Digital Audio Tape Recorder. The best way to collect the data is to collect an uninterrupted time signal at the equipment location and then process and analyze the data in a better environment. The amount of data required depends upon machine speed and the type of processing involved. For most machines about 25 or 30 seconds of data is enough. On slower machines, and when synchronous time averaging is required, several minutes of the time 7CHAPTER 3. Hardware and Software Required for Accurate Diagnostics signal and triggers may be required. Currently, the best way to collect this data is with a tape recorder. The best recorder is the Sony PC-204. Fig. 33 contains a photo of this scientific quality, digital audio tape recorder (DAT). It can record four (4) data points in 25-30 seconds or longer if needed. If a 120 minute tape is used, and the recording time is 30 seconds, then 120 min x 60 s/m = 7200 sec + 30 = 240 data points. These data points are recorded per channel, or 240 x 4 = 960 data points can be recorded on one 120 minute tape, if data is recorded on all four channels. If a 180-minute tape is used, the numbers would be 360 data points per channel, and 4 x 360 = 1440 points per tape. Administre tasks for route information, equipment identification, and data identification numbers are simplified by controlling the DAT with a pen-based, hand-held computer. These portable, battery-powered instruments can collect more data faster than other equipment on the market today. Since the hand-held computer is a full-functioning personal computer (PC), the operator can go off route, change routes, or create a new route at any time. The 10 or 20 megabyte plug-in disk installed in the PCMCIA slot provides ample storage for all routes and data points in most plants or mills. Fig. 34 contains a photo of a hand-held computer. These instruments with control software make a fast and efficient data collection system. The hand-held computer has a minimum of two megabytes of RAM, which is adequate for most programs. This revolutionary data collection system allows the operator to record horizontal and axial data on the drive and driven unit in about 30 seconds. This is enough data on most motors and pumps, etc. Fig. 3-4. GRID PalmPad Computer, ‘The processing and analyzing of the volume of data required for accurate diagnostics is a big workload if it were done manually, However, the host PC can now control the DAT, real-time analyzer, and the printer. This automated process is still time consuming, however the human interface is reduced to making hardware connections, starting the software, and changing the tape when required. The computer time for processing is offset by a factor of more than 10 to 1 by the results of accurate diagnosis of machinery problems The technology contained in this book has made possible the development of Expert software modules to diagnose problems from the data. This truly Expert software makes specific diagnosis and assigns a priority of 1, 2, or 3. Such as "Defect on the outer race priority 3” or "Defect on the inner race priority 2", etc. Currently other expert systems do 7CHAPTER 3 Hardware and Software Required for Accurate Diagnostics not make specific diagnosis. Instead they provide a list of possible problems with a percentage listed by each problem. The workload of the analyst using good expert software is reduced to reviewing output products for accuracy. Printers The various types of laser jet and ink jet printers have the best quality and are widely used for many operations. When these printers are used for graphics, the quality is still 00d, however, it takes a long time to down load the graphics from the computer to the printer. Fig. 3-5. Dot-matrix Printer. ‘The dot-matrix printer is the best choice for large volumes of printed graphics such as frequency spectra and time signals. Fig. 3-5 contains a photo of a dot-matrix printer. ‘These printers are fairly fast and can print a fully annotated spectra in about 20 seconds. The quality of these prints is quite good. Transducers ‘The electro-mechanical device used to convert mechanical motion into an electronic signal is called a transducer. These transducers can be divided into five categories. Each category contains one or more transducers for the various required measurements. The five categories are: 1. _ Displacement transducers measure how far something is moving, normally in thousandths of one inch or mils. The displacement transducer is unique because it measures relative motion, ie. how much one component is moving relative to another. 2. Velocity transducers measure how fast a component is moving in inches per second (IPS) of velocity. ‘The measurements are made in tenths of an inch per second. Most measurements are less than one (1) inch per second. 3 The accelerometer measures the rate of change per time period. 2BCHAPTER 3 Hardware and Software Required for Accurate Diagnostics ‘The units of measurement are in g’s of acceleration. Low frequencies normally have low g levels, or a fraction of one g. Higher frequencies can have levels of several g's. 4, Pressure transducers measure pressure fluctuations. The selection of pressure transducers is based on static pressure, amplitude of the fluctuations, and the frequency of the fluctuations. 5. Microphones measure audible signals in the audio range from 20 Hz to 20 KHz, The major problem with microphones is they are not very discriminating. They pick up all audible signals in the audio range. However, the fact remains that the only way to pick up an audible signal is with a microphone. There are many applications for the microphone and it should be used more often. For example, the only way to measure sound in a room is to use the microphone. NOTE: Please refer to the frequency response curves in Fig. 1-27 when selecting the correct transducer for the job. Displacement Transducers The two basic types are noncontacting and contacting. Both types require firm mounting. A bolted installation is required for permanent installation. However, a heavy duty magnetic base and a Starrett flex-o-post are satisfactory for most portable measurements. The frequency response varies by type and manufacturer, therefore, you must review the manufacturers’ specifications to obtain frequency response, transducer sensitivity, etc. Generally speaking, the displacement transducers should be used to measure most low frequencies, below 10 Hz, and all relative motion measurements. Fig. 3-6. Noncontacting Displacement Transducer. Noncontacting displacement transducers require a power supply, a modulator/ demodulator, normally called a driver, and a volt meter. Fig. 3-6 contains a photo of a portable noncontacting displacement transducer. The transducer has a small wound wire 79Ind Software Required for Accurate Diagnostics coil at the end. When the tip of the transducer is placed near the target, about .040", the relative motion between the component supporting the transducer and the target can be measured. The driver generates a high frequency. When the frequency passes through the coil, a magnetic field is generated. The motion between the mount and target intersects the magnetic field. This movement is measured when the driver demodulates the signal. If the target is moving slowly, the depth of the finish or other metal disfigurations can be measured. However, at turbine speeds those items are not measured because the frequency response at such high frequencies is not very good. This means the relative motion between a turbine housing and shaft across the coupling can be measured, even when the shaft has teeth marks left from turning the shaft with a pipe wrench. The total runout or the amount of bend in the shaft can also be measured. If a keyway is under the transducer, a pulse will be generated. The data on the shaft can still bbe used if enough attenuation is used to prevent saturation by the pulse during data acquisition. After the data is collected, enough attenuation must be removed in order to measure the shaft runout. Saturation of the pulse does not affect the data after the data is collected. This measurement may not be as accurate, because of the amplitude scale, as when a keyway is not present. However, the data is still useful. Please note these measurements must be made in the time domain, and shaft speeds should be about 6000 RPM or below. The noncontacting displacement transducer has been used for years in high-speed turbo machinery. These permanently installed transducers with the associated vibration instruments are used for on-line condition monitoring. These systems may alert by turning on a light or ringing a bell. If the vibration becomes serious enough, the instrumentation may shut the machine down. When the proximity transducer is used in high speed, all specifications concerning shaft runout, ete. are much more critical. For example, 0.5 mils on a machine that operates at 5,000 RPM is about 0.13 IPS; however, 0.5 mils at 10,000 RPM is about 0.23 IPS, and 0.5 mils at 20,000 RPM is about 0.5 IPS. The specifications for transducer installation in turbo machinery are found in the API specification 670. In slow speed applications, 0.5 mils is of little consequence. 1g Displacement Transducer. Fig. 3-7. Contacting Figure 3-7 contains a photo of the contacting displacement transducer. This transducer must make contact with the target. The contacting portion is a machined rounded surface. The tip can be removed and has the same thread size as the tips on a dial indicator. This permits replacement of the tip with a variety of "shoes" o "rollers." This transducer uses the Linear variable differential transformer (LVDT) principle. This 80CHAPTER 3 Hardware and Software Required for Accurate Diagnostics transducer came out of the gauging industry, however, it is very effective in measuring low frequencies from DC to about 200 Hertz. The measurement range and sensitivity varies. However, if the range is + 200 mils, the normal sensitivity is 25 mV/Mil. NOTE: ‘THE USER SHOULD ALWAYS REFER TO THE MANUFACTURERS’ SPECIFICATIONS FOR THE EQUIPMENT BEING USED. Dial indicators have been used for many years and will continue to be used for a long time. The usefulness of this tool is widely recognized. The development of the contacting and noncontacting displacement transducers establishes the dawning of a new era. The new era is the electronic dial indicator. If engineers, mechanics, millwrights, and technicians would start using the electronic dial indicator while the machine is in operation, to the same extent dial indicators are used on non-operating machines, untold millions of dollars could be saved in lost production and wasted manhours and materials. Fig. 3-8. Velocity Transducer. Velocity Transducers These transducers are voltage generators and do not require an external power source. Fig. 3-8 contains a photo of a velocity transducer. Most velocity transducers employ a permanent magnet mounted on a stud. A coil of wire surrounds the magnet and the coil is supported by leaf springs. When mounted on a machine, the motion causes the wire coil to move through or intersect the magnetic field. You may recall from basic physics that any time a conductor is passed through a magnetic field, a voltage is produced. The coil is wound to produce an output proportional to the movement - normally four or five hundred millivolts per inch per second. Since calibration is affected by wire size and the number of tums in the coil, calibration is determined during manufacturing, and recalibration is not required. The rugged construction and very low failure rate make this. transducer the best choice for data collection with a portable instrument or a permanent installation for on-line systems. Before a final decision is made for any transducer, consideration must be given to temperature tolerance and frequency response. Once again, the manufacturers’ specifications must be consulted. Generally speaking, the frequency response rolls off sharply below 10 Hz. This does not mean frequencies below 10 Hz cannot be measured. Quite the contrary, frequencies down to about 2 Hz can be ‘measured, however, the amplitude is understated for frequencies below 10 Hz. The upper frequency range is to about 2 KHz. However, this author has measured frequencies well above 2,000 Hz with a velocity transducer. Accurate measurements 81CHAPTER 3 Hardware and Softw Required for Accurate Diagnostics below 10 Hz should be made with a displacement transducer. Accurate measurements above 2,000 Hz should be made with an accelerometer. Accelerometers Three different accelerometers are required in most plants. Fig. 3-9 contains a photo of an accelerometer. A low frequency accelerometer is required to measure the low frequencies that cannot be measured with a displacement transducer, for example, low frequency vibrations in buildings, the earth, and some radar antennas. ‘The frequency response of these accelerometers is from DC to about 500 Hz. Since low frequencies generate low g amplitudes, the sensitivity is often 500 mV/g or more. A medium range accelerometer is also required for various measurements. The frequency response is often from about 2 Hz to about 10 KHz. However, the sensitivity may vary 15 or 20% on the high end. The sensitivity of these accelerometers is often 100 mV/g. High frequency accelerometers are also required in most plants. These accelerometers can measure frequencies up to 100 KHz, however the amplitude accuracy may be + 20%. The transducer sensitivity is often 25 mV/g or lower because high frequencies generate high g levels. The mounting of the high frequency accelerometer is quite critical. The transducer should be screwed down, the mounting surface should be machined, and a couplant should be used between the transducer and machined surface. Fig. 3-9. Accelerometer Kit. Most accelerometers are based on the piezoelectric crystal principle. A. piezoelectric crystal responds to a mechanical motion by generating an electric voltage. The voltage amplitude is quite low, in the micro (4) volt area (1 x 10“). The piezoelectric crystal responds to an electrical voltage with a mechanical vibration. These crystals are used in this manner also for delay lines in some radar applications. In most applications, the frequency response of a crystal is determined by size. The larger the crystal, the lower the frequency. The smaller the crystal, the higher the frequency. Since the output of a crystal is so low, the signal must be amplified in, or very close to, the accelerometer. If this amplification is not accomplished, the signal may get lost in the electronic noise. When the accelerometer is used, it should be hard mounted, ie. either glued or screwed down. A large segment of industry is collecting data with hand-held accelerometers, or accelerometers with a magnetic base. Such practice should be discouraged because some 82ind Software Required for Accurate Diagnostics frequencies can be over or understated by as much as 12 dB, (about 4 times the actual). Also, if you use the zoom feature, the resulting frequency window may contain noise rather than discrete frequencies. The reason data collected with a velocity transducer using a magnetic base are acceptable, is because the velocity transducer is more forgiving. For example, a slight amount of movement caused by the mounting of the velocity transducer may not be observable with the unaided eye. This is true because first, the sensitivity is high, about 400 mV/IPS; second, because of the roll off of the frequency response at the low end. Also, looseness is a low frequency phenomena. By contrast, all motion of the accelerometer is amplified. Pressure Transducers Pressure transducers measure pressure fluctuations. Fig. 3-10 contains a photo of a pressure transducer. They are similar to accelerometers, except the output is calibrated in pounds per square inch (psi). This is why the amplitude and frequency of the fluctuations must be considered before selecting a pressure transducer. These transducers measure pressure fluctuations of gases and liquids. Such measurements are often required for accurate diagnosis of problems in pumps and compressors. These transducers must be mounted in a pipe or fitting, near the discharge of the unit, where the pressure fluctuations can be felt by the transducer. In other words, the gases or liquid must contact the end of the transducer. Fig. 3-10. Pressure Transducers. Microphones ‘A microphone is required to measure audible sound, or the audio range from 20 Hz to 20 KHz. Fig. 3-11 contains a photo of a microphone. The microphone is the only instrument that can measure sound. ‘The units of measurement are dB above 1 mV. A wide range of microphones is available. The prices vary anywhere from less than $100 to several thousand dollars. The discussion of sound measurement here applies only to determining the frequency of sound for noise abatement and equipment diagnostics. The measurement of sound to satisfy OSHA requirements is beyond the scope of this book. The measurement of sound can be accomplished with a microphone. The microphone will pick up all sound in the audio range. When this data is processed on the RTA a frequency spectra is produced. 83CHAPTER 3 Hardware and Softwar Required for Accurate Diagnostics Fig. 3-11, Microphone. Once-Per-Revolution Markers When balancing rotating machinery and performing synchronous time averaging and other tasks, a trigger or a once-per-revolution marker is required. There is no single best instrument for this task. Often, more than one instrument is required, depending on the variety of circumstances encountered. The two basic types of instruments that generate a trigger are magnetic and optical. There is a wide variety in each type. The discussion of each type is beyond the scope of this book. If the once-per-revolution marker is engineered for a pickup and installed on a rotor, then a wide variety of instruments can be used. If you have to trigger from a shaft or ‘component, your best option may be the noncontacting displacement transducer installed cover the keyway. Warning: Before attempts to install are made, the shaft should be viewed with a strobe light to ascertain that a key is not present. If the trigger must be obtained from color contrasts on slow speed equipment, the fiber optic pickup may be the best choice. See Fig. 3-12. A good example is the trade line on a felt. If the green light fiber optic is used, a brown mark on a tan background can be identified. Fig. 3-12. Once-per-revolution Marker.CHAPTER 3. Hardware and Software Required for Accurate Diagnostics Fig. 3-13. Flex-o-post_and Heavy Duty Magnetic Base. A laser tachometer is also quite useful and has a broad application. However, it may not be useful in every situation. Also as with other once-per-revolution markers, mounting sometimes poses a challenge. The Starrett heavy duty magnetic base and flex-o-post is one of the best mounts for portable units. Sce Fig. 3-13, When using this type of mounting, the unit should be tied off so the unit will not fall into the machine if it falls or is knocked off. The best mounting method is to drill and tap a hole, then screw the flex-o-post onto the machine. When the flex-0-post is removed, a pipe plug could be used to fill the hole. Multiplexer Multiplexers, sometimes called switching units, are available in two basic types. The first type is a master or control unit that controls other multiplexers. The second type is the slave unit that takes instruction from the master unit. Either unit can have manual control options and /or computer control. Some examples of computer control include 16- bit parallel, IEEE 488, RS-232, etc. Depending on the configuration, these units can control ‘over 8000 data points. A typical installation would have installed transducers similar to Figure 3-14. The cable is a shielded, twisted pair and depending upon the environment the cable should withstand high temperature. Coaxial cables are normally required for high frequency applications. The other end of the cable is connected to a slave unit. The slave unit has many inputs as in Figure 3-15 and normally 3 or 4 outputs. A status feedback should be employed to identify broken wires, failed transducers, and inoperative circuits. The computer tells the master unit to access a particular slave unit, to obtain data from the desired point. After data is collected, the next point is selected. This routine continues until data from all points have been taken, processed, and analyzed. The entire process starts over again. Figure 3-16 contains a photo of a typical master unit. On-line systems using installed transducers and computer-controlled multiplexers with expert software provide identification of defects soon after they occur. The safety aspect 85CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Fig. 3-15. Rear View of a Slave Unit Fig. 3-16. Typical Multiplexer Master Unit. 86CHAPTER 3. Hardware and Software Required for Accurate Diagnostics is very important because personnel spend less time in hostile environments. Product is increased and the return on investment is often 600 percent in the first year. This author is convinced that on-line systems will be the preferred method in the near future. Gauss Meter A small gauss meter should be used to measure the magnetism in bearings, shafts, or other items that may contain small amounts of magnetism. See Fig. 3-17. This gauss meter measures + 20 gauss. The plus and minus indicate the north and south poles of the magnet. The strength of magnetism measured is a function of the strength of magnetism in the piece and the distance between the meter and the piece. Magnetism can be measured through paper, wood, etc. Therefore, the magnetism in a bearing can be measured before the bearing is unpackaged. If a bearing contains magnetism on receipt, it should be rejected. Magnetism all by itself may not be harmful. However, if the magnetism is strong enough to attract tiny metal particles, the particles could cause a premature failure. If a bearing has too much magnetism from the manufacturer, it should be rejected. Magnetism caused by induction heaters may not cause a problem unless the bearing does not have seals or shields. Bearings used in circulating oil systems should not contain magnetism or magnetized spots. Fig. 3-17. Gauss Meter. SOFTWARE INTRODUCTION The development of the personal computer created the demand for new software. The new electronic technology made possible the development of digital equipment for vibration measurements. Most plants were measuring vibration amplitude with hand- held meters, and plotting the amplitude trends manually. During this period, the predictive maintenance industry was divided into two groups. One group designed and developed data collectors and software programs for displaying, storing, retrieving, and trending data. The data collectors and software programs used with the personal computer satisfied the current demand. This new computerized predictive maintenance 87CHAPTER 3 Hardw: 1nd Software Required for Accurate Diagnostics programs were and still are a huge success. These hardware and software systems permit the collection, display, and trending of huge volumes of data. ‘The original systems had limited diagnostic capabilities. However, some improvements have been made and these improvements continue. The efforts of the other group were toward developing new technology for diagnosing problems. This book is one result of that effort. This group also believed time domain signals should be recorded for processing in a controlled environment. The supporting facts are that three or more frequency spectra/zooms and one or more time signals are required for accurate diagnosis. Since all signals may not appear in every memory period, several memory periods may be viewed before one captures the desired signal. ‘Also, a person may not be able to stay at the equipment location long enough to process and analyze the data because of temperature and safety. The collection of four measurement points in 30 seconds greatly simplifies and speeds up the data collection process. It is the best and fastest way to collect data. This task is made possible with a DAT recorder and computer control during data collection and processing. ‘The analysis of this volume of data is simplified with Expert software that performs specific diagnosis and assigns priorities. A considerable amount of software is required to support this effort. All software manufacturers claim their software is user friendly, easy-to-use, etc. Software should meet two tests in addition to being easy to use, fast, etc. The first test is one of utility. The utility test defines whether the job done by the software really needs to be done. For example, some software is developed to perform a function, regardless if the function should be performed. Such software is often in search of a problem to solve. ‘The second test is one of efficiency, proficiency, and accuracy. Efficiency has to do with conserving memory, task speed, and human interface. The working memory and storage memory should be allocated to improve efficiency and fast access. Task speed is a function of the hardware used and software design. This is accomplished by using state- of-the-art hardware, and reducing user entries by storing required data for program access. Also when a user entry requires changes of other parameters, the program should automatically change these other parameters. The human interface is concerned with the operator. The operator should not have to go off screen, except when entering or exiting the program. This is accomplished by pull-down menus. When the operator wants to change a setup, the menus are pulled down on the screen, the changes are made, and the menu is then put away until needed again. The function key, "FI," should provide the required on-screen help messages for each program function. The operator should not be required to make duplicate entries. For example, the operator should not be required to enter the 1D. number of the data to be stored, if the data contains the 1.D. number. The best software is developed by a highly skilled engineer who knows the details of what should be done, and how the task should be done. To achieve maximum results the required hardware and technology must be supported by software. This chapter presents some of the software available to perform advanced vibration analysis. This software system, called Machine Doctor (MACHDOC), contains various software packages. The first group of programs can be used as stand-alone programs to support the technology in this book. These programs are also integrated with the MACHDOC system by using the function keys, and on-screen display of results. This Diagnostic Toolbox software consists of Signal Analysis Program (SAP), Vibration 88CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Calculator, Resonance and Deflection Calculator (RADC), Bearings Program, Gears Program, and Roll Ratio/Rusch Chart Program. These programs can be used on any personal computer that uses MS-DOS 5.0 or higher. These programs calculate and display frequencies and time signals to show the operator how the signals add, subtract, and mix. They also convert between engineering units, and calculate generated and excited frequencies and improper roll ratios. These programs aid the analyst in diagnosing the problem and determining the cause of failure. Group 2 of the software is hardware dependent and can be used only with a personal computer that uses MS-DOS 5.0 or higher and the Tektronix 2630 two-channel or four- channel analyzer. This software is the MACHDOC Base system with the following options: Polar Plot, Time Plot, Balancing, Diagnostics data base, diagnostic modules, and the Roll Quality Assurance Program. Group 3 of this software is hardware dependent and requires the hardware and software listed in Group 2. It contains the Digital Option using the hand-held computer and DAT recorder for data collection and processing. The data collection routes are built into the host PC. These routes are then downloaded onto the disk of the hand-held computer. Except for major changes and additions, this is a one-time operation because the plug-in disk in the hand-held computer can hold all routes in most plants. The hand-held computer displays the routes, controls the DAT, and keeps track of the equipment LD. numbers and related tape ID. numbers. After the data is collected each time, the equipment LD. and associated tape LD. are downloaded into the host PC. The DAT is connected to the host PC and the RTA. The control software then processes the data on user-selected frequency ranges and time periods. All processed data is stored and can be printed during processing or at a later time. If the diagnostic modules are installed, this expert software can diagnose problems, assign priorities, and print listings of the diagnosed problems. Group 4 of the software is also hardware dependent and requires the hardware and software listed in Group 2, plus the real-time option using installed transducers connected toa multiplexer. The control software takes data from each installed transducer. After the data is processed, analyzed, stored, and the problem diagnosed, the next transducer is accessed and the procedure is repeated. The collected data is stored. The overall RMS value and the RMS value of each diagnosed problem can be plotted as a trend over a user-defined period in the form of a history plot. This system alerts and alarms from diagnosed problems, not preestablished amplitude levels. ‘These various hardware and software groups and options have been engineered and designed to satisfy the needs for accurate diagnosis of rotating machinery problems in small as well as large plants. They can be used to troubleshoot and diagnose on-machine problems, to collect and analyze data from routes, and to make on-line diagnostics with expert software. This system uses any transducer, and can be integrated with other on- line monitoring systems using currently installed transducers. ‘The major economic advantage of this system is that the base system can be connected to the on-line system. at night, on weekends, and on holidays for or-line diagnostics. Then during the day, the base system can be taken off line and used for test and diagnostics or used with the digital system for processing, analysis, and diagnostics of route data. 89CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Group 1, Toolbox Software The Signal Analysis Program (SAP) SAP is a versatile program designed to investigate /teach /demonstrate simple signals, or sinewaves, harmonic content, the effects of a phase difference between the signals, the results of truncation, amplitude modulation, pulses, etc. Fig. 3-18 contains the display of a screen from the SAP program. The top screen contains the frequency spectrum, or can be changed to display the time signals before the signals are added together, as in the top screen of Fig. 3-19. ‘The bottom screen display in Fig. 3-18 contains the composite time signal and the frequency amplitude and phase of two or more frequencies. The Signal Analysis Program can be used as a tutorial for the user to learn how signals add, subtract, truncate, modulate, beat, or otherwise mix. The program can also be used as a diagnostic tool. For example, the frequencies and amplitude of the spectrum from a machine can be entered. Then the phase of the signals can be adjusted until the time signals from SAP is similar to the time signal from the machine. The phase of the signals can be obtained from the SAP screen. Remember, the frequency spectra are a result of averaging, and the time signal is one memory period. This explains why the result in the SAP time signal may be a little different from the equipment time signal. Vibration Calculator Program The vibration calculator program displays displacement, velocity, and acceleration as a function of speed for pure imbalance. The program also calculates and presents the various conversions between displacement, velocity, and acceleration. It also calculates transducer sensitivity, engineering units, and dB when any two items are entered. See ver some aneLysts PrOgRA 2.08 ts Ly ‘ : oa 0.17 i out : 0.08 ed B00 | z ee 20-00 oo "B00. 0.00 We 0.0018 ee mae tee Bias ¥ os | ‘= W all v.08 ‘ 0.46 : VY Solas ed : 060 a ay ‘8606 "30000 10.00 neat? veg. ai? BHR nace POSURE maoale ees crag sirel"hase” Freq roi Mase Freq mpi Phase s5SS oMGS | "SS STSS TUB] MSS BTS GSES "SS aS Sito 8:8 8:83 9883 8 eh dg Bg ee a ee ee ey es a ee a Fig. 3-18. Signal Analysis Program Display. 90and Software Required for Accurate Diagnostics. UCI SIGNAL ANALYSIS PROGRAM 2.08 tes 0.55 0.33 outs 0.4 0.39 “oss : 0 60° ab ob X-0.00 mess” WoO. S810 ase ws, bre 3 0.48 0.16 £ “0.16 ENS -0.40 4 -o1e2 : 0:00 80:00 X-0.00 wesc” yo0. FSIS Erote Signat Freq’ fing! Phase Freq Anpt 26°83 oftB «3S 8°53 o°Ga8 0-00 G:000 “8:8 8:00 8900 8:00 9'000 = 88 8:80 8.900 8:00 9:00 8:8 8:88. 8:803 0:00 G:000 88 0:00 0.008 Fig. 3-19. SAP Time Signal Component. cr UIBRATION CALCULATOR 2.00 mun ane UIBCALC GLOBAL LABEL Frequency (Hertz>: 16.25 Displacenent : 0.84 Velocity (IPS): 0.45 Acceleration (G): 0.12 Sensitivity (nv/EU>: 375.00 Uibr. Level CEU: 1.00 9:00 oe: sta IPs a 0:90 0.30 0.72 0.24 o.sa i 0.48 oo 0.26 0.12 0.18 0.06 0.00 © oe agg 8-00 =0.00 Hz yso.00 Tes Hert? x=0.00 Hz Fig. 3-20. Vibration Calculator Program Main Display. 1CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Fig. 3-20. When actual startup or coast down data are compared to the theoretical curves, it is easy to identify imbalance. Resonance and Deflection Calculator (RADC) This program calculates resonant frequencies, mode shapes, shear, moment, slope, and deflection for beams, shafts, and rotors. Fig. 3-21 contains the main screen display. The program stores various cross sections, mass type, and weights. This keeps user entries toa minimum. Fig. 3-22 contains typical loading conventions and cross section types. From the operator entries, the static state data in Fig. 3-23 are generated. This display contains the shear, moment, slope, and deflection of the modeled piece. A deflection list is also generated. This list contains the amount of deflection at various points along the piece. The program calculates natural frequencies and displays the mode shapes, as in Fig. 3-24. The program permits the user to model proposed fixes to determine if the fix will solve the problem, and if new problems will arise as a result of the fix. This program improves diagnostic accuracy because the resonant frequencies can be calculated. Bearing Calculation Program The Bearing Calculation Program contains over 4500 bearings and provides for user updates. The program locates bearings and calculates bearing defect frequencies. The main program display is shown in Fig. 3-25. ‘The program features include: speed entry, or speed calculation based on machine speed and roll diameter; manufacturer file selection; bearing location; bearing editing and adding with option to calculate fundamental frequencies based on ball diameter, pitch UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 Section Number: 2 Section Length: 12.00 Material Type: carbon steel Cross Section Type: I-bean Concentrated Hass: 496.00 Wo: 6.00 HT: 6.00 TH: 0.35 Inertial Disk Dianeter: 0.00 Spring Stiffness: 0.00 Mass Eccentricity: 0.00 Static Force: 0.00 Harmonic Force: 0.00 Static Torque: 0.00 Harmonie Torque: 0.00 Distributed Load (Left): 41.00 Distributed Load (Right): 41.00 UNITS: English (inches, pounds > VIBRATION TYPE: in-plane bending LEFT END CONDITION: fixed RIGHT END CONDITION: free LABEL: RADC MODEL NUMBER OF DATA POINTS: 200 Fig. 3-21, RADC Main Screen Display. 92CHAPTER 3. Hardware and Software Required for Accurate Diagnostics VOI RESONANCE AND DEFLECTION CALCULATOR 2.00 Section Nunber: 2 Sect Typical Bean or Shaft Section Cross Sect ions applied force distributed load | oOo bo Cross > sprit solid bar hollow box teat I H eae tbe Hobean visti solid shaft hollow pipe UNIT: «press any key to continue> ine = : te seotin : Caanfeeeteoes ementan erasers Tod Fig. 3-22. RADC Program Loading Conventions and Cross Section Types. Setfenones ealtenennn emanates onto ‘Shear 12:00 24-00, 36-00 “48-00 60.00 ar ae — Za a aes 5's sr foo Deflection 36/00 48-00 ‘60]00 Fig, 323, RADC Static State 93CHAPTER 3 Hardware and Software Required for Accurate Diagnostics UCI RESONANCE AND DEFLECTION CALCULATOR 2.00 NATURAL FREQUENCIES & HODE SHAPES REPORT RADE MODEL. S Meturel Frequencies (Hz>: 59.94, 401.57. 1465. 2278, + NORMALIZED MODE SHAPE Wi Natural Frequency: 59.34 Hz 2a'00 35760 28105 "00 Fig. 3-24. RADC Natural Frequencies and Mode Shape. Aaa east File selection Import DOs file Speed File : SKF - Bearings Nunber Contact 1D OF Balls finale’ PFI FIFE wnSF 22252 19 10 900 0.430 6.980 2ezs2cr 18 10 370 0.420634 2ezseca 8 10 950 0.490 6.820 222608 8 so 360 014201400 22264cA ie ‘0 360 0.420 6.360 2227208 18 40 820 0.430 6.880 220000 3 40 760 0.400 4.040 223030 44 105.750 81280 O.ata S260 zest0c 44 40 5700 8.300 0.410040 2asite 44 405.730 9.270 0.410 S400 2eatze 44 10 5.720 81200 Bato S440 2eat90 44 10 5.760 8.240.410 5.200 2estac 44 10 51720 81200 aio. 5.280, zeaise Ss 40 ©6170 60 aio. Baa. 223t6c "4 $0 $.730 8.270 0.410 5.200 Label : GLOBAL BEARING LABEL Bearing : 22300 PFO = 157.200 300 Speed = 90.000 Contact Angle = 10 FIF. = 12.000 200 File : TORRINGTON Bearinos Bearing 22900.” 40023 BPFO = 157.500 470 Contact Angie = 0 RIF) = “42/320 ‘oa0 Fiselp E*Exit prosran Fig. 3-25. Bearing Calculation Program Main Screen Display. 94CHAPTER 3 Hardware and Software Required for Accurate Diagnostics diameter, contact angle, and number of balls; contact angle adjustment; and file importing to allow new bearing files to be created. The bearings are separated into files by manufacturer. The manufacturers include: Fafnir, FAG/Stamford, Link Belt, MRC, SKF, Timken, Torrington, and others. Eight harmonics are calculated and listed for the outer and inner race frequencies. Contact angle from ball pass frequency of the outer race (BPFO) or the ball pass frequency of the inner race (BPFI) is calculated to identify bearings in a thrust load and the angle of the thrust. Cross reference is made for different manufacturers. The program calculates two times ball spin frequency (BSF). The fundamental train frequency (FTF) with inner or outer race rotating is also calculated. Gears Program The Gears Program simulates two gears in mesh. Fig. 3-26 displays the gears with simulation completed. ‘The number of teeth on each gear, the gear identification, and the speed of one gear or the gearmesh frequency is entered. From this data the other speeds and frequencies are calculated. This includes gearmesh frequency and harmonics, fractional gearmesh frequencies, hunting tooth frequency, common and uncommon factors, and ratio. When the diameter of a gear is entered, the pitch-line velocity is calculated and the AGMA gear quality number is listed. The program also features a pull-down window for calculating gearmesh frequencies for planetary and sun gears, A visual display is provided on how an eccentric gear with an improper ratio can cause accelerated wear on certain teeth. The reduction in gear life expectancy is calculated based on the common factor. Gear 1: GEAR ONE Gears Global Label Teeth: 55. OF: 11 Connon Factor’ 9 Speed: 5.72 Ratie: 4.22 Dia: 42.25 Vel: 3000 GH: 345.00. TF: acta: 8-11 FINISHED 45 G~63.00 2/5 GH=126.00 3/5 Gr=189.00 4/5 o=252.00 42 cate157.50 1 1/2 GM=472.50 2 GH=690.00 3 GH=845.00 Fig. 3-26. Gears Program Display of Finished Simulation. 95CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Paper Makers Inc, Roll 1: Roll 2: SBBRS 33/7 33/7 33/7 33/7 82/11 52/11 82/11 1/15 1/15 85 / 18 85 / 18 90 / 19 90 / 19 109 / 23 109 / 23 109 / 23 318 / 25 118 / 25 118 / 25 123 / 26 412.00 to 412.70 inches #3 Paper Machine (Yankee Dryer) 87.50 inches (Pressure Roll) 87.00 to ‘BAR_INTERVAL ROLL 1 12.48 412.00 12.49 412.17 12.49 412.33 12.50 412.50 7,92 412.10 7.93 412.36 7.93 412.62 5.80 412.15 5.81 412.51 4.85 412.25 4.85 412.67 4.58 412.20 4.58 412.65 3.78 412.47 3.78 412.56 3.79 412.66 3.49 412.24 3.49 412.41 3.50 412.57 3.35 412.05 ROLL 2 87.40 87.43 87.47 87.50 87.18 87.23 87.29 87.08 87.15 87.30 87.39 87.02 87.12 87.03 87.05 87.08 87.34 87.37 87.41 87.10 Hig. 3-27. Improper Roll Ratio Table. 96CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Roll Ratio Program and Rusch Chart Special problems are presented when two or more rolls are operated in nip. These problems also affect other items going through the nip such asa felt and product. These problems are similar to gears in that an improper ratio exists if the circumference of the rolls and/or felt have a common factor. When a common factor is present and either roll or felt has an eccentricity, the rolls and felts can get bars/corrugations/streaks. These irregularities can cause problems in the product and associated equipment failure. The distance between the bars is normally greater than the width of the nip before these problems occur. In an effort to solve these problems the roll ratio program was developed. The maximum and minimum circumference of two components is entered. Then all size combinations of the two components that will cause barring are entered. Figure 3-27 contains a sample of the improper roll ratio table. The table lists the number of bars that will occur on each roll, the distance between the bars, and the size of each roll that can cause the bars. This table is quite large and difficult to use. To simplify the volume of data, the Rusch Chart was developed. This chart carries the name of the design engineer, Terry L. Rusch. Figure 3-28 contains the Rusch Chart. The circumference of one roll is plotted on the y- axis and the circumference of the other roll or component is plotted on the x-axis. The lines indicate where bars or improper roll ratios can occur. These lines are actually sunburst lines. The lines appear parallel because only a small segment is displayed. The blank spaces where lines do not occur indicate roll combinations where bars should not occur. The box inset can be moved around, made bigger or smaller, and then expanded to fill the screen. This provides better resolution. Block range; tom lett: 97.59,412.31 | bottom right: 07.21. 412.59 Sursor'foestion™” “"y.so, 412.50 ""* Bar interval” 12°50 87.70 . 97.50] 3 97.10] 27.001 ‘2-00 “412.70 Roll #1 Lats Bara Roll #2: 7 Gers Roll #1: 99 Mme. 242.59 9.00 4.00 Conoany Ident it teat ton. Fisteip RUSCH CHART Fig. 3-28. Rusch Chart Display from Roll Ratio Program. 97CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Group 2, Machine Doctor (MACHDOC) This software is the vibration analysis platform for the processing of all data. It operates the base system. The digital and real-time systems are also designed to operate with the base system, and integration consists of installing additional software. ‘The hardware required is a personal computer, the Tektronix 2630, two or four-channel analyzer, and a printer. Other software options include Polar Plot, Time Plot, Balancing, Diagnostic Database, and Expert Diagnostic Modules. MACHDOC This software provides the graphics for displaying the frequency spectrum and time signal. The operator can use a single display as shown in Figure 3-29, or a double display as in Figure 3-30. Up to four displays can be selected. Hardware setup is accomplished with pull-down menus, These menus are accessed with the ALT key and the first letter of each word across the top of the display. In Figure 3-29, the triggering menu is displayed. Some of the major features of the software are: 1. Cursor, with harmonics, and difference cursors in frequency and time 2. Single key labeling of up to 50 peaks in frequency and time 3. Diagnostic label storage and recall 4. Horizontal and vertical on-screen labeling 5. On-screen conversion and display of time and frequency 6. Zoom and expansion of frequencies. Expansion only of the time signal. Center frequency and bandwidth are user selectable 7. Toolbox, Polar Plot, Time Plot, and DOS accessible from function keys. When bearings or gears are accessed, the calculated frequencies are displayed on the screen upon return to the MACHDOC screen 8. Primary and secondary equipment identification labels of up to 25 characters each 9. Scaling, printing, get data or go, etc. are accessed by the first letter of the function. i.e. ’S’ for scale, ‘P’ for printing, ‘G’ for go, etc. 10. Function key, "FI" help available at each function level 11, Single key storage of data, and data recall Some of the above features are displayed in Figure 3-30. In addition to storage and recall of data, the data can be recalled and plotted /printed as stacked spectra. Figure 3-31 contains a sample display. Some people call this a waterfall or cascade plot, however, it is not. A true waterfall or cascade plot is a larger scries of 98CHAPTER 3. Hardware and Software Required for Accurate Diagnostles Averaging Displays Frequency Gride Processing " oro 0.56 ; +: “ 4 Egret i 0.42 EBennnsoadh i 0.20 i a ate on eg isle § { mu A a oo weg 0 Baas tes ce 8 shover Supply Puno Savi Ont date tstops FESS cena window seiset ise Fig. 3-29. MACHDOC Display. ae ee i “a0 60300700 Herts Shover Supply Pum 34+11 Fig, 3-30. MACHDOC Display with Labeling,CHAPTER 3 Hardware and Software Required for Accurate Diagnostics Displevs Frequency Grids Processing Setup ‘Triggering (06/09/1991 09:14:04 o6or/i991 12:36:23 os/agsi991 ia:z0iie Ft i ' 05/23/1991 i1:05/43 20.00.42 Y=0.0085 IPS 1/K2 ewe Here Fr=0, 996 2H DRIVE 2PBGBrA Averages taken (62 Get data [stop] 2250 Frm FisHelp F2:Beer F $00" M1 Phase Adjustnent: “3.78 eo [Traneducer ere? | "25 nU/ Mil Tranechicer Polarity Positive Ver POLAR PLOT 2.10) Fig. 3-33. Polar Plot Display. 101CHAPTER 3 Hardware and Software Required for Accurate Diagnostics er TIRE PLOT 2.02 “pbs ea 56: stONaL. - CHAN'S ‘6:00 63.08 4397-97 iets sesvea ata l82 Xm0.00 nese -Y=-D.35 TPS Me He TIME poNeIN oava ~ eat DATA Frequency: S000 Tine span: 2.24 sec Frana: tC 71 5.00 wsec/in Mi: MAAKEML 0.00 neac. 2: MARK (0.00 newe Fig. 3-34. Time Plot Display with Trigger. Time Plot The Time Plot Program is used with the high-speed parallel interface option in the ‘Tektronix 2630 analyzer. This option permits the writing of an uninterrupted time signal and trigger to the hard disk in the computer. The only limitation is the space available on the hard disk. Figure 3-34 contains the Time Plot screen display. The recorded data can then be printed with user selectable resolution with two computer generated timing marks. This time signal is required to analyze the eccentricity of each gear, hunting tooth frequency, and other long time events that cannot be observed on one screen of the time domain signal. Figure 3-35 contains a small sample of the Time Plot output with trigger and timing marks. When long time periods are used with a lot of resolution, it may take several hours to print the time signal. Balancing Imbalance may not be the most prolific diagnosis made concerning rotating machinery. Untold thousands of dollars are wasted annually by trying to solve vibration problems by balancing when imbalance is not the problem. However, if a rotor has imbalance, it must be balanced. All rotors should be balanced in two planes before installation. ‘The balance tolerance for rigid rotor in Figure 4-1 should be used as a reference. Installed rotors sometimes require single plane balancing, sometimes called trim balancing. Single plane, or trim balancing, is often selected for installed rotors because it is quicker, easier, and requires less downtime. Sometimes a two plane balancing is required on installed rotors. 102CHAPTER 3. Hardware and Software Required for Accurate Diagnostics. ote file: TMEDATATZ0 9/17/1991 12:38:10 TIME DOMAIN DATA jOPR “Sicnat” ‘ot MARKER OF MARKERS i Fig. 3-35. Time Plot Sample Output with Trigger and Timing Marks. When balancing is accomplished using a computer, all information such as whether rotor operates above or below first critical, phase lag and balance sensitivity can be stored and recalled for future balancing of the same piece. Figure 3-36 contains the single plane balancing window and Figure 3-37 contains the two plane balancing window. The computer calculates the angles, weight, and weight location. The operator installs/ connects the equipment and moves the weights. Diagnostic Database The diagnostic database contains the internal geometry and speed for each data point. From this information, the computer calculates the generated frequencies. It is used with the diagnostic modules for accurate diagnosis of problems with the expert software. A tree structure is used so all data can be displayed on one screen. Figure 3-38 contains the main screen display. The database is arranged by machine or group, section, and measurement point. The diagnostic information is listed for each point. This information contains required items such as transducer sensitivity and type, speed, diameter. Several bearings and/or gears can be entered for each point. Also the number of vanes on impellers, blades on fans, etc. can also be entered. The Add, Delete, Edit, List, Mode, Path, Replicate, and Tach can be accessed with the ALT key and the first letter of each word. When more than one measurement point contains the same information, the multiple entries can be made as a single entry. For example, TR 1-6 and WRR 1-8 in Figure 3-38 represent Table Roll 1 through 6, and Wire Return Roll 1 through 8. When different sections have the same points, the replicate feature can be used to duplicate the sections. These advanced design features reduce the time required to build the database. For example, if 3 hours are used to build a database with 103CHAPTER 3. Hardware and Software Required for Accurate Diagnostics Orig anol: Trial Wt: Trial anpl: Balance Ht= AAD fossa -5.50 X=0.00 nsec vss. 0.00 45.00 80:00 33.69 Ed 42/29/1999 08: 10:01 Fig. 5-36. Single Plane Balancing. Ma 42/29/4999 08: 18:01 5.50 ~ Phase Orig _Ano1: Orig Ane: 40-00 3.30 Tr fl NP: Te Aap) FP 45.00 Tr fol FP: Tr Amp NP: 5.00 Phase Trial Wi Trial ut: 0.00 Balance Wi Balance Ht= 2.30 14.96. 1.10 | 0.00 4.40 |. -3.30 BO 270° Near Plane Far Plane lense Fig. 3-37. Two Plane Balancing. 104CHAPTER 3. Hardware and Softwar Required for Accurate Diagnostics Add Delete Edit List Hode Path Replicate Tach Machine/Group Sect ion Point/Roll mostics Information PD 1DR BELBONDS ‘Type: Roll are APR BELBONDZ Dianeter: 38,50” zen 20R BELBONDS Pri SKF 22326¢ ent 2PR BELBONDS Trans: 375 ws IPS A POWER 30k BELBONDS Tach: PHT FPH want SPR BREAST uinnz 40n couck SDR Dendy oR FLovacs 18 FLovacz cALST FLovaca FOURD LUNPBR SMPR TR-6 TACH umRt: ur Fig. 3-38. The Diagnostics Database Main Display. Delete _EAdit List Mode Path Replicate Tach Point/Roll Mux. Channel Listing BELBONDL Chand ILBOMDZ Chan? 2 arn FOURD BELBONDS Chan sreD FOURD HLBOND4 ch spe TOURD BELBONDS Chan APeD FOURD BREAST Chan sreD TOURD coucn Chan arp FOURD DANDY chan APD TOURD rLovact Chan Are FOURD FLovacz Chan APD FOURD TLovac3 oh sre FOURD Lunn Chan = 1PHD FOURD T™. chan: sPeD FOURD Te Chan seep FOURD ™3 Chan = sPMD FOURD ™4 Chan APD FOURD 7s Chan sPeD TOURD 776 Chan = ir) ——— Fours ——— wnrt = ——— Chan APH FOURD RR. Gran Fistelp Fig. 3-39. The Diagnostics Database Real-time Option with Multiplexer Listing Display. 105CHAPTER 3 Hardware and Software Required for Accurate Diagnostics MACHDOC, 30 hours may be required on a less sophisticated system. The mode selection permits access to the multiplexer listing in Figure 3-39. This listing is used with the real-time option and lists the multiplexer channel for each point. The route listing used with the digital option is also accessed through ALT-Mode. The listing in Figure 3-40 displays measurement point, tape ID. number, recorder channel, machine speed, length of recorded data in seconds, and the date the recording was made. This information is required for the computer to process the recorded data. For large plants with many databases, the path function is used to access the various databases. The list selection permits access to the various reports. To obtain these reports, the diagnostic modules must be used on stored data, recorded data, or real-time data. ‘The priority listing contains Priority 1, 2, and 3 problems by machine group, section, and point. Figure 3-41 contains a sample report. ‘A comparison report is also available. This report compares the priorities between any two user-selected periods. ‘A sample of the History Report is contained in Figure 3-42. This report plots the trends for the overall RMS and the RMS value of each diagnosed problem, along with the machine speed. The time period is user selected. This report identifies when the defect occurred and indicates the deterioration rate. Add Delete Edit List Mode Path Replicate Tach Mach/Grp Section Pt /Roll Ibe Chan Speed Time Date. aD 208 DRI. e101 wse 3S 3/81/93 2D 2S DRIZ wer 2 1258 3593/81/93 mo ee mos a aD 20S DRI4 wl 4 125835 BBLS 2D 2s DRIS 121 125835 381/93 ao 2s DRIG 1622 1258-35 83/81/93 ao 2s DRI? 18203 1258-35 3/81/93 ao abs DRIa 1824 1258 3583/81/93 2 20s DRIS 16301 1258 3583/81/93 2D 2s DRza 1832 1258-35 83/81/93 aD 308 DR2t 1833 1258 3543/81/93 ao 30S DR2z 10304 1258-35 83/81/93 a DS DRZ3 18401 ise 3583/81/93 ao as DR24 1040 2 125835 3/1/93 aD 30S DR2S 10403 1258-35 3/1/93 ao as R26 10404 1258-35 31/93 Fictelo File: TAPEL Fig. 3-40. The Diagnostics Database Digital Option with Route Listing Display. 106CHAPTER 3. Hardwar ind Software Required for Accurate Diagnostics UCI PRIORITY LISTING 4M LOOSE-PRI1, OUTER-PRI3(@.08), INNER-PRI3: 1D-(9/9): OVERALL PRIZ LOOSE-PRI1, CAGE-PRI3; 1D-(3/4); OVERALL-PRI1 LOOSE-PRI1, OUTER-PRI3, INNER-PRIZ, ROLLER-PRI3; 1D-(29/43);OVERALL-PRI1. “press any key to advance to next section? Fig. 3-41. The Diagnostics Database Sample Screen Listing of Priority Report. ANT 10S. ORS eee 0:20) Fig. 3-42. History Report 107CHAPTER 3 Hardware and Software Required for Accurate Diagnostics The diagnostic modules perform the actual diagnostics. Since all industrial plants do not have the same equipment, the diagnostic software has been divided into four (4) broad categories. These categories or modules diagnose problems in antifriction bearings, gears, rolls in nip, and general machinery condition such as imbalance, misalignment, ete. The diagnostic database is required for the diagnostic modules. The modules are the Expert software for diagnostics. The computer diagnostics can be performed on stored data, taped data, and real-time data from the multiplexer. The major difference between these diagnostic modules and other expert systems is: These modules diagnose specific and multiple problems, and assign a priority for each problem and an overall priority. These modules alert and alarm on actual diagnosed problems. ‘Figure 3-43 contains an analyzed spectra from MACHDOC. The data in the upper right corner identify the various problems, frequency, and RMS value. The letter ‘L’ is looseness, 'O’ is outer race defect, ‘’ is inner race defect, 'C’ is the cage frequency, and 'R’ is roller speed. These defect codes are also printed on each spectral line. The diagnosis from the expert software indicates a looseness, inner race, and roller problem at a priority 3. The outer race and overall is a priority 1. The ID-(27/37) indicates the software identified 27 of a total of 37 spectral lines. A visual observation of the data in Figure 3-12 indicates all spectral lines were identified. The human eye may not see all frequencies because of the vertical scaling. However, the computer can "see" those frequencies. Where the diagnostics were tested, a qualified engineer analyzed the data. Then the computer analyzed the same data. An analysis of the results indicated the computer found all the problems the engineer found plus several more low level problems. These are the results when a qualified diagnostic engineer programs the Averaging Displays Frequency Grids Processing Setup Triggering ies Oras 0. z 0.00. 4s O02 TPS AK meee Hert faez80 208 AMT 108 ons LOOsE-PRIa, QUTER-PRIL, 11M 10-¢29799; foe Fig. 3-43, MACHDOC Display after Diagnostics Have Been Performed. 108CHAPTER 3. Hardware and Software Required for Accurate Diagnostics computer to diagnose problems using the same engineering logic. These results are quite impressive. The reason the computer found more problems than the engineer, is that the engineer could not see all spectral lines because of scaling levels. The computer could identify those spectral lines. Roll Quality Assurance Program When the plant is operating rolls that are in nip, considerable savings can be realized by the improvement of roll management and roll quality. When rolls are operated in nip, ie. one roll is rotating against another roll, all parameters must be held to a closer tolerance. For example, imbalance and runout should be less than 1 mil (0.001"). Most materials take a set. Some rolls may be measured round while in the grinder and be a problem when installed. When rolls with a set are installed and operated under load, the hard and/or soft places could appear as bars or corrugations from before the grind. In such cases a repeat failure could occur. Roll roundness should be measured under operating conditions before the roll is removed. This can be accomplished by relative motion measurements between the rolls and/or dynamic measurement of steel rolls. Synchronous time averaging should be accomplished in both cases. Roll roundness and both targeted and actual crown should be measured before and after the grind. The roll roundness should be measured again when the roll is installed, and before the roll is removed. If the above measurements were made, and the Roll Ratio/Rusch Chart in Table 7-1 and 7-2 were used, much would be learned about the cause of roll failures, transfer function, and what should be done to prevent roll failures. Some of the new roll grinding machines are equipped with the instrumentation to measure roll roundness while in the roll grinder. For those that must continue to use what they have, the Roll Management program may be helpful. This program operates with the above hardware. While the roll ig in the grinder, measurements are made with two displacement transducers mounted 180 degrees apart, in order to subtract out way and bed sag. The roll is divided into a user-determined number of segments and measurements are made at each segment. The following reports are generated: The Roll Diameter Report lists average, minimum, and maximum diameter at each roll segment and location. The Polar Report displays the roll shape and runout for each segment and location. The Crown Report displays the targeted crown and the actual crown. ‘The program measures circular, sine, and straight crowns. ‘The Surface Report is also available. The computer presents the round surface as a flat matrix. The display can be rotated and viewed from different angles. 109CHAPTER 3 Hardware and Software Required for Accurate Diagnostics A database is provided for storing the reports for later retrieval and comparison. Group 3 The digital option is a method to collect a large volume of data in a short time period and the resulting computerized processing, of the data. ‘The measurement/data collection routes can be built in the host PC and then downloaded to a hand-held computer. Figure 3-44 contains a hand-held computer screen of a route or portion thereof on a paper machine. This software is operated by touching the computer pen to the screen and/or writing on the screen. The hand-held computer controls the DAT recorder and keeps track of the tape recorder channel, and tape .D. number when each point is recorded. When the route is completed, the measurement point, DAT channel and I.D. number are downloaded from the hand-held to the personal computer, The DAT is then connected to the personal computer. The operator installs the tape and starts the program, and the computer processes the data. The computer can also analyze the data and diagnose the problems, if the diagnostic database and diagnostic modules are installed. There is also a subroutine that permits the analyst to screen large volumes of data for serious problems. When the computer is processing data unattended and an error occurs, the error is written to a specific file. When the analyst returns, the error file can be listed and the analyst can make necessary corrections. Group 4 This software uses the multiplexer listing from the database and controls the multiplexer for processing and analyzing data from permanently installed transducers. 1b: 20, 205, ons, crs Tesetoe: 488, (ear) 2MD_205, 0R16 BW 2 Speed: 15,8,8, 20, 20 0.n3,7 a Roc Time: 8,3,5, ano, 205 oni, a Tozn200 Lares. eel 2MD 208 DAIT 2MD 2S DRI2 2MD 20S DRI3 2MD 20S DRI4 2MD 208 DRI6 2MD 25 DRI7 2MD 208 DRI8 2D 208 DAIS || 2m 20s pr20. 2MD 2DS DR21 2MD 208 DR22 |,2MD 205 DR23 Fig. 3-44. PalmPAD Screen of Displayed Route. 110CHAPTER FOUR: ACCURATE EVALUATION OF MACHINERY CONDITION INTRODUCTION This chapter describes the procedures for accurate evaluation of machinery condition. Diagnosis of bearings, gears, and rolls are covered in other chapters in this book. Accurate evaluation of machinery condition is more difficult because: 1. The same frequency spectra can be produced from various time domain signals. This is because the time signal contains phase and the various forms of signal distortion. 2. Some frequencies are acceptable below certain levels and other frequencies are not acceptable at all within prescribed calibration levels. Some frequencies can be caused by more than one problem. For example, the fundamental or one times RPM can be caused by looseness, imbalance, bent shaft, misalignment, resonance, loading, pump starvation, open iron/broken rotor bars, and soft foot/casing distortion. The fundamental can also be measured in a machine coupled to another machine that has one of the above problems. 4. Accurate analysis of one frequency often depends upon the presence of one or more other frequencies. The importance of the frequency domain spectra in various frequency ranges and the ability to zoom in on various frequency windows are stressed. The requirements for the time domain signal in various time ranges and the understanding of the phase between various signals are also emphasized. As pointed out in Chapter Two, some valuable data is lost when converting from the time domain to the frequency domain. This is why the same or similar frequency spectra are presented for very different problems, and why the time domain signal is required for accurate diagnostics. In fact, several time periods in various time ranges are required. This is true because all events may not be occurring in a given time period. High frequencies are not visible if the time period is too long, and low frequencies are not visible if the time period is too short. Problems that can occur in all machines, such as imbalance, bent shaft, misalignment, looseness, etc., are discussed separately. Problems unique to certain machine groups, such as motors, pumps, fans, turbines, compressors, etc., are discussed with that group. The importance of calibration and adjusting sensitivity for speed is also discussed. THEORY, ‘The measured frequency tells what is wrong with a machine, and amplitude identifies the relative severity. Frequencies do not lie, although they can be misunderstood, measured improperly, or interpreted incorrectly. Amplitude, on the other hand, can be over or understated. Amplitude is not understood by most people and is often misinterpreted. For these reasons, amplitude, all by itself, is of little value and should not be relied upon. For example, a small fan could vibrate at levels up to 1 IPS for years and not fail. On the other hand, a bearing with a cracked inner race could fail within days with an amplitude of 0.02 IPS.CHAPTER 4 Accurate Evaluation of Machinery Condition ‘Most of the vibration severity charts were designed for imbalance only. However, over the years these charts have been used and applied to all machinery problems. The net result is that machines fail and the vibration level does not go up, in fact it often goes down. Consequently, people think it is a catastrophic failure. If diagnostics had been performed along with or instead of condition monitoring, some warning would have been obtained. Fig. 4-1 is the balance tolerance nomograph which contains a good balancing specification for rigid rotors during shop balancing. When using this graph, place one end of a straightedge on the rotor weight scale, and place the other end on the rotor operational speed scale. Then read the allowable gram inches of residual imbalance from the intersection of the straightedge and the center scale. This graph is based on ISO 1940, and ANSI $2.19 - 1975 with some refinements. Fig. 4-2 contains the operating machinery imbalance guide. This chart is a good guide for general severity of imbalance only. We do not know the origin of this chart. However, it may be a derivative of the Rathbone Chart. Various versions have been published in several places. In some circles it is referred to as the IRD Chart. Regardless of who deserves the credit, it is a good chart for imbalance severity and should not be used for other problems. Some frequencies are permitted at acceptable levels in some machines. Some examples are imbalance (1 X RPM), gearmesh frequency, vane pass frequency, etc. These frequencies should not contain harmonics (except for 1 X RPM in fluid film bearings), modulation, or truncation. In other words, a "good" machine could have the above discrete frequencies at amplitudes of less than 0.2 IPS. Other frequencies are not permitted at any level within the prescribed calibration levels. Some examples of these frequencies are oil whirl, all bearing frequencies, pulses (except in cutters, etc.), noise, and fractional discrete frequencies. Most real-time analyzers have enough sensitivity to measure the balls or rollers passing under the transducer. If the calibration standards are not observed, or if auto-ranging is used, bearing frequencies could be observed without the presence of a significant defect. In Chapter Two, we discussed how frequencies add, subtract, modulate, and truncate based on physics in an electronic environment. Sometimes rotating machines can behave in a very similar manner. A good example is a beat: If a 1776 RPM motor was belt- driving a 1500 RPM fan, and the motor and fan were isolated from each other, the motor would be vibrating at a frequency of 29.6 Hz, and the fan would be vibrating at a frequency of 25 Hz. This is true when there is no mechanism to allow the two frequencies to beat. If the motor was mounted on the fan pedestal, then the two frequencies could beat. The flexibility of the fan pedestal is the mechanism that permits the two frequencies to beat. If the fan and motor mounting were isolated and the motor was loose, the beat would be felt on the motor and not on the fan. If the fan was loose, the beat would be felt on the fan and not on the motor. In both cases, the looseness is the mechanism that permits the beat. In these examples, the nonlinearity (looseness and flexibility in this case) permitted the frequencies to beat. In other cases, two frequencies can be present, generated by the same machine, and they will not beat. An example would be imbalance and vane pass frequency. In such cases, a spectral line is present at rotating speed, and another spectral line is present at vane pass frequency. In this case, we have two frequencies and two problems (imbalance and a vane pass frequency problem). The time signal is that of a high frequency riding a low frequency. On the other hand, if the impeller was loose on the shaft, the spectral line at vane pass frequency could be modulated by unit speed. Modulations of speed could appear below and above 112CHAPTER 4 Accurate Evaluation of Machinery Condition Allowable residual unbelance (gm.—in.) 50005 055 4000-4 14 3600 4 J 3000 z | For two plone & i - 400-4 balancing use one~ i half the allowable ¢ | 4 per plane. 4 5 10 g & 0 0 a u a a w Q 5 oe u a max — 2500 Les ROTOR WEIGHT, Fig. 4-1. Balance Tolerance Nomograph for Rigid Rotors. 113CHAPTER 4 Accurate Evaluation of Machinery Condition RPM ooo'es c0o'0z 10.00) 8.00 6.00 4.00 1.00 0.60 30 g °. DY3d-O1-wwad ~ SIN ~ yosd~"08/u1 0.10 0.08 0.08 0.04 oz = INaNaOvIESIa NOLLYUBIA 0.008 0.008 0.004 0.001 Fig. £2, Imbalance Severity Chart. 14CHAPTER 4 Accurate Evaluation of Machinery Condition vane pass frequency. However, more energy may be contained in the sidebands below vane pass frequency. Sidebands below the discrete frequency are an out-of-phase condition or looseness, and sidebands above are an in-phase condition or eccentricity. In some looseness problems, sidebands can occur on the high side when eccentricity is not present. This occurs because looseness goes in and out of phase. Remember, if frequencies are out of phase, they subtract, if they are in phase, they add. In either case, amplitude modulation occurs, In other cases, the various frequencies are not generated and then added together. In these cases, only the resultant time signal is produced. Remember, a perfect sinusoid contains only a discrete frequency. The only way to distort the sinusoid is to add some harmonic content. The distorted signal can take many forms, depending upon the harmonic content and the phase relationship of the harmonics. This signal is normally the discrete or fundamental frequency along with various harmonics and their phase relationships. Since phase is not presented or is discarded after the FFT is performed, the time signal must be used to obtain phase for accurate diagnosis. If your analysis does not include the time signal, your diagnosis could contain up to a 20% error rate. CALIBRATION Some frequencies at acceptable amplitudes are found in “good” machines. Some examples are imbalance, vane pass frequency, gearmesh frequency, etc. Other frequencies are not acceptable at any level within prescribed calibration standards. Some examples are the various frequencies generated by antifriction bearings, pulses, bars, or corrugations on rolls, etc. In addition, spectra taken from "good" machines should not be modulated by shaft speed or any multiple of shaft speed. However, the data from most machines contain some sidebands. The reason for this paradox is that it may not be economical to fix or repair all identified problems. Most of the problems in determining what is good or bad can be eliminated by strict adherence to calibration standards and adjusting sensitivity to speed. This writer has used the speed-adjusted calibration standards for twenty years with outstanding success. See Table 4-1. Table 41. Speed-adjusted Calibration Standards. eed ROTATING MACHINE PAPER MACHINE ‘SPEED IN RPM ‘SPEED IN FPM 1,000 and above 2,400 and above 0 0.0.70 IPS 500 - 1,000 1,700 - 2,400 0 to 0.35 IPS 200 - 500 1,000 - 1,700 0 to 0.18 IPS Below 200 Below 1,000 0 to 0.10 IPS ed Another valuable utility of Table 4-1 is that problems of the same severity appear at about the same percentage of full scale for all speeds. For example, an imbalance problem on a 1200 RPM fan at 0.35 inches per second (IPS), or half scale of zero to 0.7 IPS, is just as bad as 0.05 IPS, or half scale on a 150 RPM fan. Of course, the latter fan would not require balancing. However, if a similar situation occurred with looseness or bearing defects, a fix would be required. Therefore, automatic ranging should be used only in 15CHAPTER 4 Accurate Evaluation of Machinery Condition Tare cases. FREQUENCIES GENERATED The fundamental frequency or one times RPM can be caused by imbalance. When a rotor has one heavy place, the weight of the heavy place, distance from the shaft center line to the center of the heavy place, and angular rate of velocity or speed all combine to create a forcing function or vector at some magnitude and direction. If the rotor has more than one heavy place, the weight will be the vector sum of all the heavy places. As the heavy place rotates with the rotor, a frequency at the speed of the rotor is generated. Several conclusions should be made from this simple example: 1, The industry standard for measuring machine speed is revolution per minute (RPM) or feet per minute (FPM). 2 The industry standard for measuring frequency is cycles per second (CPS), now called Hertz, to honor the man who developed the frequency theory. 3. Frequency identifies the problem. If the discrete frequency is not present, it will always manifest itself as a sum or a difference frequency. 4. Amplitude identifies relative problem severity. Amplitude can be overstated by looseness and resonance. Amplitude can be understated by mass, rigidity, and damping. Amplitude can also be overstated or understated by the way most instruments measure amplitude. Harmonic content and signal distortion are other good ways to determine problem severity. All generated frequencies are equal to events times speed. For example, half-shaft speed is generated when an event is occurring every other revolution, as in some looseness problems. If a rotor has imbalance, the heavy spot rotates at the same speed as the rotor. When two events are occurring each revolution, as with looseness, bent shaft, or two high places on a roll, the fundamental and second harmonic are generated. If three events are occurring each revolution, as with misalignment, the fundamental, second, and third harmonic are generated. It also follows that: VP = Sx Ny Blade pass frequency: where VP = vane pass frequency CASES S speed Gearmesh frequency: Ny = Number of vanes on the impeller GM*SxN, Some frequencies vary. For example, frequency is equal to the number of events times speed. If a rotor becomes loose or improperly adjusted, it could hit or strike the housing or some other object. The frequency generated can vary, depending on how many times it hits each revolution. The frequency will be the number of hits times speed. 116CHAPTER 4 Accurate Evaluation of Machinery Condition Note that vane pass, blade pass, and gearmesh frequency are all a specific harmonic of rotor speed. In some cases, it may be difficult to distinguish between a harmonic and another generated frequency, such as vane pass frequency, ball pass frequency, ete. In other cases, frequencies such as ball pass frequency inner race (BPFI) or ball pass frequency outer race (BPFO) may be close to a harmonic of rotating speed, or vane pass or ball pass frequency. In such cases, a frequency translator or zoom feature and the time domain signal are required for differentiation between the various frequencies. Accurate analysis also requires determining the phase relationships between the fundamental and the harmonics. DATA COLLECTION Data should be collected in the load zone with due respect to flexibility. For example, imbalance is normally a radial-loaded problem. However, imbalance in some cantilevered fans can cause an axial load. Looseness can be either radial or axial-loaded, depending on where the machine is loose and how the machine is installed. Horizontal and vertical offset misalignment cause a radial load. Angular offset misalignment causes an axial load. Monitoring of small simple machines can often be accomplished by collecting data in the horizontal and axial directions on the drive and driven units across the coupling. Depending on time, workload, etc., one monitoring point at 45° between horizontal and axial may suffice for monitoring. However, data must be taken in the horizontal, vertical, and axial directions when trouble-shooting and establishing baseline data for new installations, and when new equipment is added to the program. Most failures occur on the drive and driven end of machines. Therefore, data must be taken on both sides of the coupling. Data should also be taken on the "off" end of the drive and driven unit in some cases. For example, data should be taken on the off end when the rotor is between two bearings. The location of where the data is taken also depends on what you want to measure. If a motor has the same bearing on both ends and one bearing is defective, often the location of the defective bearing can be identified by taking data from both ends. If the problem is a radial-loaded problem, then data must be taken in the radial direction. Imbalance is a good example. If the problem is an axial-loaded problem, then data must be taken in the axial direction. Antifriction bearings and angular misalignment are good examples of axial-loaded problems. When trouble-shooting, data should be taken in the horizontal, vertical, and axial directions on the drive and off end of each machine in the complete machine train. For routine monitoring, it may be adequate to take data in the horizontal direction on the drive and driven end of each machine. For large motors, fans, and other equipment, it may be desirable to monitor both ends. Note: These data collection procedures apply to equipment that doesn’t have permanently installed monitors. Different rules apply to high-speed turbo machinery. The appropriate specifications should be reviewed in such cases. 117