Statistical Quality Control

Dr. Gyan Prakash

 SQC Categories
Statistical quality control (SQC)
 SQC encompasses:
 Statistical process control (SPC)
 Involves inspecting the output from a process
 Quality characteristics are measured and charted
 Helpful in identifying in-process variations
 Acceptance sampling
 used to randomly inspect a batch of goods
 Descriptive statistics
 e.g. the mean, standard deviation, and range
Sources of Variation
 Variation exists in all processes
 Variation can be:
 Assignable causes of variation
 Causes can be identified and eliminated
 e.g. poor employee training, damaged tool, machine needing repair

 Common or Random causes

 Random causes that we cannot identify
 Unavoidable
 e.g. differences in process variables like diameter, weight, service time, temperature
 Traditional Statistical Tools
 Descriptive Statistics include n

 The Mean- measure of central tendency of observations

x
i 1
i
x
n
 The Range- difference between largest/smallest observations in a set of data

 x 
n 2
 Standard Deviation measures the amount of data dispersion i  X
i 1
σ
n 1
 Distribution of Data shape
 Normal or bell shaped
 Skewed
Distribution of Data
 Normal distributions  Skewed distribution
 Setting Control Limits
Percentage of values under normal curve Control limits balance risks like Type I error
Area covered Right process judged wrong
100-99.74= 0.26

Variability
SPC Methods-Control Charts
 Control Charts
 Drawn with sample data plotted on a graph with specification of CL, UCL, and LCL

 Control chart for variables

 are used to monitor characteristics that can be measured, e.g. length, weight, time

 Control charts for attributes

 are used to monitor characteristics that have discrete values and can be counted,

e.g. % defective, number of flaws in a shirt, number of broken eggs in a box

Upper control limit

Central line

Lower control limit

Set of observations
Control Charts for Variables
 Use x-bar charts to monitor the changes in the mean of a process
(central tendencies)
 System can show acceptable central tendencies but unacceptable variability

 Use R-bar charts to monitor the dispersion or variability of the process

 System can show acceptable variability but unacceptable central tendencies
Control Charts for Variables
 x-bar and R-bar charts used together
 reveal different problems

Shift traced by x-bar

Shift traced by x-bar

Adapted from Reid and Sanders
Constructing a X-bar Chart
Example: A quality control inspector at the ABC soft drink company has
taken three samples with four observations each of the volume of
bottles filled. If the standard deviation of the bottling operation is 0.2
ounces, use the below data to develop control charts with limits of 3
standard deviations for the 16 oz. bottling operation.
Time 1 Time 2 Time 3 x1  x 2  ...x n σ
Observation 1 15.8 16.1 16.0
x , σx  Given σ
k n
Observation 2 16.0 16.0 15.9
where (k) is the # of sample means and (n)
Observation 3 15.8 15.8 15.9
Observation 4 15.9 15.9 15.8 is the # of observations in each sample
Sample means (X- 15.875 15.975 15.9 UCL x x  zσ x
bar)
Sample ranges (R) 0.2 0.3 0.2 LCL x x  zσ x
 Solution and Control Chart (x-bar)
 Center line (Mean of Mean: x- double bar)

15.875  15.975  15.9

x 15.92
3
 Control limits for ±3 σ limits:

 .2 
UCL x x  zσ x 15.92  3  16.22
 4
 .2 
LCL x x  zσ x 15.92  3  15.62
 4
X-Bar Control Chart
 Control Chart for Range (R)
Factors for three sigma control limits
Center Line and Control Limit
Sample Size Factor for x-Chart Factors for R-Chart
(n) A2 D3 D4
2 1.88 0.00 3.27
3 1.02 0.00 2.57
0.2  0.3  0.2
R .233 4 0.73 0.00 2.28
3 5 0.58 0.00 2.11
6 0.48 0.00 2.00
7 0.42 0.08 1.92
UCLR D4 R 2.28(.233) .53 8 0.37 0.14 1.86
9 0.34 0.18 1.82
LCLR D3 R 0.0(.233) 0.0 10 0.31 0.22 1.78
11 0.29 0.26 1.74
12 0.27 0.28 1.72
13 0.25 0.31 1.69
14 0.24 0.33 1.67
15 0.22 0.35 1.65

Adapted from Reid and Sanders

R-Bar Control Chart
X-bar Chart Using R-bar
 When sigma for the process distribution is not known

0.2  0.3  0.2

R .233
3

UCL x x  A 2 R 15.92  0.73.233 16.09

LCL x x  A 2 R 15.92  0.73.233 15.75

 Control Charts for Attributes
 Attributes are discrete events: yes/no, pass/fail

 Use P-Charts for quality characteristics that are discrete and involve
 Percent of fuse CFLs in a box of 48
 Percent of broken eggs in a carton

 Use C-Charts for discrete defects when there can be more than one defect
per unit
 Number of flaws or stains in a carpet sample cut from a production run
 Number of complaints per customer at a hotel
 Constructing a p-Control Chart
Example: A Production manager for a tire company has inspected the
number of defective tires in five random samples with 20 tires in each
sample. The table below shows the number of defective tires in each
sample of 20 tires. Solution:
Sample # Defective # Tires in each Proportion
Tires Sample Defective
1 3 20 .15 # Defectives 9
CL p   .09
2 2 20 .10 Total Inspected 100
3 1 20 .05 p(1  p ) (.09)(.91)
σp   0.64
4 2 20 .10 n 20
5 2 20 .05 UCLp p  z σ  .09  3(.064) .282
Total 9 100 .09 LCLp p  z σ  .09  3(.064)  .102 0
P- Control Chart
Constructing a c-Control Chart
Example: The number of weekly customer complaints are monitored
in a large hotel using a c-chart.
Week # of Complaints
1 3
2 2
3 3
# complaints 22
4 1 CL   2.2
# of samples 10
5 3
6 3 UCLc c  z c 2.2  3 2.2 6.65
7 2 LCLc c  z c 2.2  3 2.2  2.25 0
8 1
9 3
10 1
Total 22
C- Control Chart
Control Chart Patterns
UCL

Center
UCL

LCL
Center

Sample observations
consistently below the center line LCL

Sample observations
consistently above the center line

Adapted from Krajewski, Manoj K. Malhotra, Larry P. Ritzman

 Control Chart Patterns …
UCL

Center
UCL

LCL
Center

Sample observations
consistently increasing LCL

Sample observations
consistently decreasing

Adapted from Krajewski, Manoj K. Malhotra, Larry P. Ritzman

 Zones for Pattern Tests
UCL 3 sigma = x + A2R
Zone A
2 sigma = x + (A2R)

Zone B
1 sigma = x + (A2R)

Zone C
Process Mean of mean
average
Zone C
1 sigma = x - (A2R)

Zone B
2 sigma = x - (A2R)
Zone A
LCL 3 sigma = x - A2R

| | | | | | | | | | | | |
1 2 3 4 5 6 7 8 9 10 11 12 13
Sample number
 Normal Distribution
+/- 3 sigma most common choice of
confidence/control limits in quality control

Mean
-3 -2 -1 +1 +2 +3
68.26%
95.44%
99.74%
 = Standard deviation
Process Control Charting

0.8 2.2 3.6 0.8 2.2 3.6

After the first sample After the second sample

0.8 2.2 3.6

Fitting a normal distribution to the
histogram of sampled observations

0.8 2.2 3.6 0.8 2.2 3.6

By the end of the first day By the end of the second day

Adapted from Slack, Chambers and

Process Variability
X
AP AX
P

Not accurate but Accurate but not

precise precise

XX
AP AP
Accurate as well as
Neither accurate nor
precise
precise

Legend:
On/off target – accuracy: A
Scatter – precision: P
Adapted from Slack, Chambers and
Assignable Causes
Mean
Average

Shift in mean of sample

May be due to assignable causes

Grams
Assignable Causes…
Spread
Average

Shift in spread
May be due to assignable causes

Grams
Assignable Causes…
Shape
Average

One side shift

May be due to assignable causes

Grams
Control Charts
Assignable causes likely

UCL

Nominal

LCL

1 2 3
Samples
 Control Limit and Errors
Type I error:
Concluded that a process is out of control At 3-σ limits
But in reality it is because of randomness/not-assignable
Probability of a Type I error is 1 - 0.9997 = 0.0003 Making stringent control limits
i.e. moving away from average
4-σ Chances of Type I error decreases
3-σ
UCL
As long as the area encompassed by the control limits
is less than 100% of the area under the distribution,
there will be a probability of a Type I error
Process
average

LCL
Control Limit and Errors…
Type I error:
Concluded that a process is out of control 2-σ limits
But in reality it is because of randomness
Decreasing control limits
Probability of a Type I error is 1 - 0.9544 = 0.0456
i.e. Moving towards average
Chances of Type I error increases
3-σ limits
UCL
2-σ limits
As long as the area encompassed by the control limits
is less than 100% of the area under the distribution,
there will be a probability of a Type I error Process
average

LCL
Control Limit and Errors…
Type II error: 3-σ limits
Failed to detect a shift in average:
Probability of concluding that nothing Making stringent control limits
has changed i.e. moving away from average
Chances of Type II error increases

3-σ
UCL

Shift in process
average Process
average

LCL
Control Limit and Errors…
Type II error: 2-σ limits
Probability of concluding that nothing has changed
Moving towards average
i.e. by reducing control limits
Chances of Type II error decreases

UCL
2-σ

Shift in process
average Process
average

LCL
Process Capability
 Process capability
 Ability of a production process to meet or exceed pre-set specifications
 Measures the match between design variability and actual variability

 Tolerances
 Design specifications that reflect product requirements with variability
Process Capability…
Design
Specifications
Natural variation exceeds design
specifications

Process is not capable of meeting

specifications
Variation

Process
Design
Specifications

Design specifications and natural variation the same

Process is capable of meeting specifications most of

the time. No variation

Process
Process Capability …
Design
Specifications
Design specifications greater
than natural variation

Process is capable of always

conforming to specifications.

Process
Design
Specifications
Specifications greater than natural
variation, but process off center

Capable but some output will not

meet upper specification.

Process
Process of center
Process Capability
 Product Specifications
 Preset product or service dimensions: tolerances
 e.g. cold drink bottle fill might be 600 ml. ± 2 ml. (598-602 ml.)
 Process Capability – Cp and Cpk
 Involves evaluating process variability relative to preset product or service specifications
 Cp assumes that the process is centered in the specification range

specification width UTL  LTL UTL: Upper tolerance limit

Cp   LTL: Lower tolerance limit
process width 6σ

 Cpk when process is not cetered

 UTL  μ μ  LTL 
Cpk min , 
 3σ 3σ 
Interpretation: Cp Measures
Cp = 1 Cp ≤ 1
 Three possible ranges for Cp

 Cp = 1, process variability just meets

specifications

 Cp ≤ 1, process not capable of producing

within specifications

 Cp ≥ 1, process exceeds minimal

specifications
Cp ≥ 1
 Drawback: Cp assumes that the process is
centered on the specification range

 Cp=Cpk when process is centered

Interpretation: Cpk Measures

Cpk = negative number

Cpk = zero

Capable but some output will not

Cpk = between 0 and 1 meet upper/lower specification

Cpk = 1
Design specifications greater
than natural variation

Cpk > 1 Process is capable of always

conforming to specifications.

40
Choice of Control Limit: ±6 Sigma versus
± 3 Sigma
 Motorola coined - six-sigma, GE early adopter- DEMAIC (define, measure, analyse, improve, control)
 Six-sigma: quality benchmark
 Before design, marketing ensures product characteristics
 Operations ensures that product design characteristics can be met by controlling materials
and processes to 6σ levels
 Other functions like finance and accounting use 6σ concepts to control all of their processes
Process variation Process variation Process variation Process variation

LSL USL LSL USL LSL USL LSL USL

Natural variation in
process +/- 3 sigma
is half of specification

3 sigma process 4 sigma process 5 sigma process 6 sigma process

= 66800 defects/ million = 6200 defects/ million = 230 defects/ million = 3.4 defects/ million
Acceptance Sampling
 Process of randomly inspecting a certain number of items from a lot or
batch to decide whether to accept or reject

 Performed either before or after the process

 Sampling of material received from suppliers
 Sampling of finished components prior to assembly
 Sampling of finished items before shipment

 Used where inspection is expensive, volume is high, or inspection is

destructive
Acceptance Sampling Plans
 Criteria for acceptance or rejection of a lot is based on:
 Size of the lot (N)
 Size of the sample (n)
 Number of defects above which a lot will be rejected (c)
 Level of confidence we wish to attain
AQL, LTPD, Consumer’s Risk (α) &
Producer’s Risk (β)
 AQL (Acceptance Quality Limit) is the small % of defects that consumers are willing
to accept; order of 1-2%

 LTPD (Lot Tolerance Percent Defective) is the upper limit of the percentage of
defective items consumers are willing to tolerate

 Consumer’s Risk (α) is the chance of accepting a lot that contains a greater number
of defects than the LTPD limit; Type II error

 Producer’s risk (β) is the chance a lot containing an acceptable quality level will be
rejected; Type I error
OC Curve: 100% Inspection
Operating Characteristic represents discriminatory power
P (Accept Whole Shipment)

100%
Keep whole shipment Return whole shipment

0%
0 1 2 3 4 5 6 7 8 9 10
Cut-Off % Defective in Lot

45
OC Curve with Less than 100% Sampling
Sampling plan aims to discriminate between good and bad: sample size and acceptance number
P (Accept Whole Shipment)
Probability is not 100%: Risk of keeping bad shipment or
returning good one.
100% = producer’s risk for AQL, = Consumer’s risk
Acceptance quality level (AQL) , Lot tolerance percent defective (LTPD)
Keep whole
shipment
Return whole shipment

0%
0 1 2 3 4 5 6 7 8 9 10
Cut-Off % Defective in Lot
OC Curve Showing Risks
Sampling plan discriminate between good and bad: sample size and acceptance number

100
 = 0.05 producer’s risk for AQL (Probability of rejecting a high-quality lot)
95

75
Infinite slope
Probability of
Acceptance
Steep slope
50

Vary ‘n’ the number in sample and

‘c’ the acceptance number
25 Increasing n with constant c

= 0.10 Holding n and decreasing c

10 Percent
Consumer’s risk for 0 1 2 3 4 5 6 7 8 Defective
LTPD
Good AQL Indifference zone
LTPD Bad lots
(Probability of accepting a Lot tolerance percent defective
lots
low -quality lot)
Implications for Managers
Key decisions
How much and how often to inspect?
 Consider product cost and product volume

 Consider process stability

 Consider lot size

Where to inspect?
 Inbound materials

 Finished products

 Prior to costly processing

Which tools to use?

 Control charts are best used for in-process production

 Acceptance sampling is best used for inbound/outbound purposes

SQC in Services
 Difficult to measure the quality of a service
 Services - simultaneity, intangibility, heterogeneity, perishability

 Perceptions of quality are highly subjective

 Servqual Model: dimensions (tangibility, empathy, assurance, reliability, responsive)

 Expectation and perception

 A way to deal with service quality is to devise quantifiable measurements of the service
element
 Check-in time at a hotel

 Number of complaints received per month at a restaurant

 Number of telephone rings before a call is answered

 Acceptable control limits can be developed and charted

SPC to Service …
 Hospitals
 timeliness and quickness of care, staff responses to requests, accuracy of lab tests,

cleanliness, courtesy, accuracy of paperwork, speed of admittance and checkouts

 Stores
 waiting time to check out, frequency of out-of-stock items, quality of food items,

cleanliness, customer complaints, checkout register errors

 Airlines
 flight delays, lost luggage and luggage handling, waiting time at ticket counters and

check-in, agent and flight attendant courtesy, accurate flight information, passenger cabin
cleanliness and maintenance
SPC to Service …
 Restaurants
 waiting time for service, customer complaints, cleanliness, food quality, order accuracy,

employee courtesy

 Online Order Companies

 order accuracy, packaging, delivery time, order waiting time

 Insurance Companies
 billing accuracy, timeliness of claims processing, agent availability and response time
Thank you