Plot

of inputs and outputs can be seen as on Figure 1.

From plots we can see that measured value has very small noise. Also we can not that there are changes can be observed in measured value due to changes in set point or disturbance. From plots we can representations of set point and disturbance signals are as below: !" = 1 10 + 2 20 + 30 = 50 2 70 ! = !"#$%&& !" + !"#$%&'()*+ ! = !!"#$%&& + !!"#$%&'()*+

Figure 1. !" , and ! vs plots

where is a unit step function. We need to identify process and disturbance transfer functions or in other words determine !"#$%&& and !"#$%&'()*+ of this process. We see that measured signals from = 0 to = 50 is a response to changes in set point and from = 50 to = 100 is a response to changes in disturbance. Also, we can see that responses are first order or overdamped second order with or without delays. We will use graphical analyses to identify transfer functions !"#$%&& and !"#$%&'()*+ .

 Lets make a closer look at time interval = [0,10]. On Figure 2 we can see !" and ! in the same plot for time interval = [0,12]. We can see that step response of ! can be modeled as of a first order system with time delay. So lets assume that !"#$%&& is in the form as on equation below. !"#$%&& = !!! + 1

Figure 2. !" and ! vs plots for = [0,12]

Lets determine parameters , and from plot. We know that set point changes from 0 to 1 at = 1 and back to 0 at = 10. To determine we need to find steady state value of ! in = [1,10] interval. If we try to ignore noise, we can see that in interval = [6,10] that ! can be assumed as steady state. So with assumption that noise is completely random, average of ! in interval = [6,10] will give us steady state value of ! . ! 0.020905 So gain of process transfer function is then = !! 0.0209.
!"

We can try to guess ! without noise graphically as on Figure 3, where dashed line was drawn above noisy signal. From Figure 3 we can see that ! start to increase at time approximately equal to 1.6 seconds. Since set point changes at = 1, it means that delay is = 0.6. 2

Figure 3. Guessing ! without noise for = [0,6] The last step is to determine time constant . We can estimate from a step response plot using value of at which response is 63.2% complete. So, we need to find time when ! becomes greater than 0.632 = 0.01321. We can use plot on Figure 4, where only guess from Figure 3 and a lines helping in estimating are drawn.

Figure 4. Guessing for without noise for = [0,6] From Figure 4 we see that ! reaches its 63.2% value at = 2.74. So, = and !"#$%!"#$%&& is as below 1 0.88 2.74 1.6

!"#$%!"#$%&& = Lets now find !"#$%!"#$%&&

0.0209 !!.!! 0.88 + 1

!"#$%!"#$%&& = !"#$%!"#$%&& !" !" = !" !! !!"! !!"! !!"! = +2 +

!"!"#!"#$%&& = = !"#$%!"#$%&&

0.0209 !! !!"! !!"! !!"! !!.!! +2 + 0.88 + 1

0.0209 !!.!! !!".!! + 2 !!".!! + !!".!! 0.88 + 1

= 0.0209

1 ! !.!! 1.6 1 !
!!!".! !.!!

!!!.!

!!!".! !.!!

10.6

+ 2 2 !

20.6 + 1 !

!!!".! !.!!

30.6

Plot of !"#$%!"#$%&& can be seen on Figure 5. It can be noted that after = 50 we have disturbance change, thats why ! differs for > 50.

Figure 5. !"#$%!"#$%&& and ! for = [0,55] To find steady state value of !"#$%!"#$%&& we eliminate all exponential terms as below.

!"#$%!"#$%&&!! = lim !"#$%!"#$%&& = 0.0209 1 1 + 2 + 1 = 0.0627

Now lets find transfer function of disturbance. From Figure 1 we can see that change in disturbance causes significant change in ! . Also from the same plot we see that there is no change in disturbance in interval = [50,70] and ! can be assumed steady state during = [60,70]. To find steady state value of ! in interval = [60,70] we take average in order to eliminate noise, assuming that noise is random. ! 0.2726 = ! !"#$%!"#$%&&!! = 0.2099

On Figure 6 we can see ! for time interval = [49,59], which is a response in change in disturbance at = 50. We can see that noise is undistinguishable on this plot. Curved nature of response for time interval = [50,50.4] points that system is a second order and may contain very small or no delay. Lets first assume that there is no time delay and use method by Smith and then with trial and error try to estimate all parameters. For method by Smith we need times at which response is 20% or 60% complete. So, we need to find times when ! becomes greater than 0.2 + !"#$%!"#$%&&!! = 0.1047 and 0.6 + !"#$%!"#$%&&!! = 0.1886. These lines and corresponding to them !"% and !"% can be seen as on Figure 7. We can estimate times as !"% = 50.42 and !"% = 51.13.

Figure 6. ! for = [49,59]

 ! Assuming that there is no time delay then !" = 0.42 and !" = 1.13 and ratio !!" = 0.372.
!"

Figure 7. Estimating !"% and !"% for = [50,52]

Figure 7.7 from textbook can be seen on Figure 8.

From Figure 8 we can read 1.3, !" 2.5 having 0.452. From this analysis !"#$%&'()*+ can be estimated as !"#$%&'()*+ = = ! ! + 2 + 1 ! + 1 (! + 1)

Figure 8. Smiths method relationship of and to !" and !"

where ! ! = ! and ! + ! = 2 . So, ! ! 0.204, ! + ! 1.18 and ! 0.97, ! 0.21. By trial and error around these parameters we found best estimate for ! = 1 and ! = 0.2. Plot of !"#$%!"#$%&'()*+ and ! can be seen on Figure 9.

System can be described as in equation below ! =

Figure 9. !"#$%!"#$%&'()*+ and ! for = [49,54]

0.0209 !!.!! 0.2099 !" + 0.88 + 1 + 1 0.2 + 1

and the overall !"#$% and ! can be seen on Figure 10.

Figure 10. !"#$% and !