Sie sind auf Seite 1von 3

Project 2

Saturday, December 7, 2013 12:16 PM

Brian Sit

a) Determine Process a. GP(s) in standard form per reactor

i. For two reactors with similar dynamics; ii. CV CA2 MV CA0 DV T Set point CSP(s)

Y'sp
%TO

Error(s) %TO

Signal C(s)
%CO

CA0(s)
g/m3

GA2(s)
g/m3

CA2(s)
g/m3

b.

c. Substitute values in

Gm(s), Feedback signal %TO

CA2(s) g/m3

b) Fill in the block diagram above with all input and output labels as well as units for each line. c) Open Loop Transfer function = G OL =

d) MatLab Root Locus Diagram

a.

b. rlocfind(sys) KCU = 333 U = 1.47 rad/s

e) Tuning a PID controller i. IMC Method 1) Using Skogestad's Second Order Approximation a) b) c) IMC using case B i. ii. iii.
4.1458

Homework Page 1

ii.

1) Tyreus-Luyben a) Kc = 333*.45=149.85 b) i = 4.27*2.2=9.394 c) D = 4.27/6.3=0.6778 2) Ziegler-Nichols a) Kc = 333*.6=199.8 b) i = 4.27/2=2.135 c) D = 4.27/8=0.5338 Method KC iii. IMC TL ZN I D 0.6778 0.5338 13.433 16.5834 4.1458 149.85 9.394 199.8 2.135

Skogestad's for second order

&

iv. ZN is the most aggressive since the gain is so high and the time constants are the lowest 1) IMC is the least Aggressive since the gains are lower and the s are higher f) Block Diagram

a.

g) Proportional Only

a.

b. P only controllers do have more oscillations and offset versus PID however, the ultimate frequency leads to infinite oscillat ions and this does not. The true ultimate frequency should be around 460, and this difference versus 333 is possible from the interpolation between points used by matl ab in the graphical approach.

i.

1) In the SISO Tool, the graphical approach gives different values and the pole location around 1.47 when using rlocfind(sys,[1.47]) gives a value of 447. These varying values and the location of the vertexes on the plot indicate that tighter samplings would give more accurate results.

h)

a. IMC seems to be the best option. The overshoot and oscillations are reduced. It also most closely follows the pattern of a cr itically dampened function which are ideal. Model Rising Time (s) Overshoot Settling Time (s)
b. IMC 0 0 12

TL ZN

2.3 2.2

1.5 2.6

15 29

Homework Page 2

b.

IMC TL ZN PID

0 2.3 2.2 K 333*.5=166.5

0 1.5 2.6 i

12 15 29 D

i) For ZN With P, PI, and PID P PI PID

a.

333*.45=149.85 4.27/1.2=3.5583 333*.6=199.8 4.27/2=2.135 4.27/8=0.5338

b.

c. PID gives the best results. PID gains enhancements over PI otherwise a more complicated process like PID would not be needed if PI was better. In this case, the settling time and oscillations are reduced using a PID over PI or P j) Tuning a. Table PID ZN PID K i D 4.27/8=0.5338 Filter 333*.6=199.8 4.27/2=2.135

Tuned (14s response) 4.723

4.723/0.38=12.43 4.723*6.88=32.50 0.387

b. These values are different but they also give different trends on the graph. There is also a filter which changes things. The ZN method with its high gain has a lot of oscillations versus the tuned method. The lower gain from the tuned method also has higher values.

c.

i. The IMC values seem to be the best, the TL is next then the Tuned then the ZN method which is the most aggressive. The Tuned method does not seem to see the Sensor and that is why it may have provided what seemed to be good tuning but without accounting for the rest of the loop. 1) I would go with IMC tuning however, with a bit more play, perhaps the Tuned method could be tuned correctly and provide even better results.

Homework Page 3