23-Chem-A6 Process Dynamics and Control · December 2018
Question 3 of 8: IMC Design for a First-Order-Plus-Dead-Time Process
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Notes on this paper
Reference texts: D. E. Seborg, T. F. Edgar, D. A. Mellichamp & F. J. Doyle III, Process Dynamics and Control (4th ed., Wiley) — Laplace-domain modelling, linearisation, transfer functions, feedback stability, frequency response and IMC design; G. Stephanopoulos, Chemical Process Control: An Introduction to Theory and Practice (Prentice Hall) — Routh test, Nyquist criterion, dead-time systems and state-space models; D. R. Coughanowr & S. E. LeBlanc, Process Systems Analysis and Control (3rd ed., McGraw-Hill) — first-order tank/thermal dynamics, step/impulse response and block-diagram algebra. Standard control conventions (deviation variables; unity valve/sensor gains unless stated) are used throughout. This sitting is an open-book paper of eight problems of which any five constitute a complete exam; all eight are solved here.
Problem 3: IMC Design for a First-Order-Plus-Dead-Time Process (20%)
Find. (1) $G_c^{*}$, the classical equivalent $G_c$, and whether it is PID; (2) $\delta C(t)$ for a unit set-point step.
Problem 3: IMC structure. The controller $G_c^{*}$ acts on the error between set point and the model-mismatch signal $C-\tilde C$; with a perfect model $\tilde G=G$ the inner signal vanishes and $C/R=G_c^{*}G$.
Approach. Factor $G$ into its invertible and non-invertible (dead-time) parts, invert only the invertible part and append the filter to get $G_c^{*}$, convert to the classical loop via $G_c=G_c^{*}/(1-G_c^{*}G)$, and use the perfect-model identity $C/R=G_c^{*}G$ for the response.
Factor the model. Split $G=G_-G_+$ into invertible $G_-=\dfrac{5}{10s+1}$ and all-pass/dead-time $G_+=e^{-2s}$ (a delay cannot be inverted causally).
(1) IMC controller. Invert $G_-$ and add a first-order filter $f=\dfrac{1}{\tau_Cs+1}$: $$\boxed{G_c^{*}=G_-^{-1}f=\frac{10s+1}{5}\cdot\frac{1}{20s+1}=\frac{10s+1}{5(20s+1)}.}$$
(1) Classical equivalent. Using $G_c^{*}G=\dfrac{10s+1}{5(20s+1)}\cdot\dfrac{5e^{-2s}}{10s+1}=\dfrac{e^{-2s}}{20s+1}$, $$G_c=\frac{G_c^{*}}{1-G_c^{*}G}=\frac{\dfrac{10s+1}{5(20s+1)}}{1-\dfrac{e^{-2s}}{20s+1}}=\boxed{\frac{10s+1}{5\,(20s+1-e^{-2s})}.}$$
Is $G_c$ PID? No. The $e^{-2s}$ in the denominator makes $G_c$ a transcendental (dead-time-compensating, Smith-predictor-like) controller, not a rational $K_c(1+1/\tau_Is+\tau_Ds)$. Only if the delay were replaced by a Padé rational approximation would $G_c$ reduce to (approximately) PID form.
(2) Perfect-model set-point response. With $\tilde G=G$ the IMC closed loop gives $\dfrac{C}{R}=G_c^{*}G=\dfrac{e^{-2s}}{20s+1}$. For a unit step $R=1/s$: $$\delta C(s)=\frac{e^{-2s}}{s(20s+1)}\ \Rightarrow\ \boxed{\delta C(t)=\Big(1-e^{-(t-2)/20}\Big)\,\mathbf 1(t\ge 2).}$$ The controlled variable stays at zero for the $2$ s of dead time, then rises as a first-order curve with time constant $\tau_C=20$ s toward $1$ — perfect (offset-free) set-point tracking, delayed only by the unavoidable dead time.
Quantity
Result
(1) $G_c^{*}$
$(10s+1)/[5(20s+1)]$
(1) $G_c$ (classical)
$(10s+1)/[5(20s+1-e^{-2s})]$
(1) PID form?
No — dead time in denominator
(2) $\delta C(t)$
$(1-e^{-(t-2)/20})\mathbf 1(t\ge2)$, final value $1$