You are on page 1of 19

Lecture 11 - Processes with Deadtime, Internal Model Control

• • • • • • Processes with deadtime Model-reference control Deadtime compensation: Dahlin controller IMC Youla parametrization of all stabilizing controllers Nonlinear IMC
– Receding Horizon - MPC - Lecture 14

EE392m - Spring 2005 Gorinevsky

Control Engineering

11-1

Processes with Deadtime • Examples: transport deadtime in paper.Spring 2005 Gorinevsky Control Engineering 11-2 . mining. oil • Deadtime = transportation time EE392m .

Processes with Deadtime • Example: transport deadtime in food processing EE392m .Spring 2005 Gorinevsky Control Engineering 11-3 .

Processes with Deadtime • Example: resource allocation in computing Computing Tasks Difference Equation Modeling Resource Queues Resource Desired Performance Feedback Control EE392m .Spring 2005 Gorinevsky Control Engineering 11-4 .

PI CONTROLLER: k P = 0.3.6 0.2 0 • Can we do better? PC – Make = z −d 1 + PC – Deadbeat controller −d 0 5 10 15 20 DEADBEAT CONTROL 25 30 1 0.8 0.8 0.2 continuous time discrete time 1 0.2 0 0 5 10 15 20 25 30 z PC = 1 − z −d EE392m .Spring 2005 Gorinevsky 1 C= 1 − z −d u ( t ) = u ( t − d ) + e( t ) 11-5 Control Engineering .4 0. k I = 0.Control of process with deadtime • PI control of a deadtime process yd PLANT: P = z -5 - C P y P = e − sTD P=z −d .4 0.6 0.

Model-reference Control • Deadbeat control has bad robustness. at least as large as in P(z). deadtime • More general model-reference control approach – make the closed-loop transfer function as desired P( z )C ( z ) = Q( z) 1 + P( z )C ( z ) 1 Q( z) C( z) = ⋅ P( z ) 1 − Q ( z ) Q ( z ) is the reference model for the closed loop C P y y d • Works if Q(z) includes a deadtime.r. EE392m .Spring 2005 Gorinevsky Control Engineering 11-6 . Then C(z) comes out causal. especially w.t.

. + a z )u ( t ) = (b z +bz + .. + b z )e ( t ) 14 4 4 4244 44 3 14 4 4 4 4 424 44 4 44 3 1 −1 N −N 0 M −N 1 M − N −1 B(z) N −N A( z ) EE392m .Spring 2005 Gorinevsky Control Engineering 11-7 ... + a N b0 z M − N + b1 z M − N −1 + ....Causal Transfer Function B ( z ) b0 z M + b1 z M −1 + .. + bN C (z) = = N A( z ) z + a1 z N −1 + .... + bN z − N = 1 + a1 z −1 + .. + a N z − N • Causal implementation requires that N ≥ M ( 1 + a z + .

λ .Spring 2005 Gorinevsky • Single tuning parameter: α .a. generic first order response P( z ) = z −1 1 − bz with deadtime 1 − α −d st order+deadtime Q( z) = z • reference model: 1 1 − αz −1 1 − bz −1 1−α C( z) = ⋅ g (1 − b) 1 − αz −1 − (1 − α ) z −d • Dahlin’s controller EE392m .tuned controller Control Engineering 11-8 . 1967 P( z ) 1 − Q ( z ) g (1 − b) −d • plant.tuned controller a.k. 1 Q( z) C( z) = ⋅ • Dahlin’s controller.Dahlin’s Controller • Eric Dahlin worked for IBM in San Jose (?) then for Measurex in Cupertino.

5TD Open-loop 50 60 CONTROL STEP RESPONSE 1.5 deadtime.Spring 2005 Gorinevsky Control Engineering 11-9 .5 1 0.5-2. CLOSED-LOOP STEP RESPONSE WITH DAHLIN CONTROLLER • Industrial tuning guidelines: Closed loop time constant = 1. • Direct use of the identified model parameters.2 0 0 10 20 30 40 Ta=2.8 0. 1 0.Honeywell-Measurex.4 0.5 0 0 10 20 30 40 50 60 EE392m .5TD Ta=1.Dahlin’s Controller • Dahlin’s controller is broadly used through paper industry in supervisory control loops . 60%.6 0.

Spring 2005 Gorinevsky Q C= 1 − QP0 Reference model: T = QP0 Filter Q Control Engineering Internal model: P0 11-10 . some use in process industry e P P0 e = r − ( y − P0u ) u = Qe • continuous time s • discrete time z EE392m .IMC General controller design approach.Internal Model Control .

IMC and Youla parametrization reference yd disturbance d Q C= 1 − QP0 Q= C P - y output e error u control d→y yd → y d →u • Sensitivities S = 1 − QP0 T = QP0 Su = Q C • If Q is stable. then S. T. and the loop are stable 1 + CP0 • If the loop is stable. This is called Youla parameterization • Youla parameterization is valid for unstable systems as well EE392m .Spring 2005 Gorinevsky Control Engineering 11-11 . then Q is stable • Choosing various stable Q parameterizes all stabilizing controllers.

Q-loopshaping • Systematic controller design: select Q to achieve the controller design tradeoffs • The approach used in modern advanced control design: H2/H∞.Spring 2005 Gorinevsky Inversion 11-12 Control Engineering . H∞ loopshaping • Q-based loopshaping: Loopshaping −1 S = 1 − QP0 S << 1 ⇒ Q ≈ (P0 ) • in band • Recall system inversion EE392m . LMI.

F≈ T. −1 S << 1 ⇒ Q ≈ (P0 ) −1 • in band • out of band T << 1 ⇒ QP0 << 1 • For a minimum phase plant T = QP0 = F F= 1 (1 + λs )n S = 1 − QP0 = 1 − F • F is called IMC filter. reference model for the output • Lambda-tuned IMC EE392m .Spring 2005 Gorinevsky Control Engineering 11-13 .Q-loopshaping • Loopshaping S = 1 − QP0 T = QP0 Q = P0† = F (P0 ) .

Spring 2005 Gorinevsky Control Engineering 11-14 .Lecture 14 EE392m .IMC extensions • Multivariable processes • Nonlinear process IMC • Multivariable predictive control .

Spring 2005 Gorinevsky Control Engineering 11-15 .Nonlinear process IMC • Can be used for nonlinear processes – linear Q – nonlinear model N – linearized model L e EE392m .

need to change the model • actuators/sensors off • add-on equipment EE392m . • Not used commonly in process control (except Dahlin controller) – detailed analytical models are difficult to obtain – field support and maintenance • process changes.Spring 2005 Gorinevsky Control Engineering 11-16 .Industrial applications of IMC • Multivariable processes with complex dynamics • Demonstrated and implemented in process control by academics and research groups in very large corporations.

v ) + G ( x .Dynamic inversion in flight control & = F ( x .Space Station Lifeboat Reference model: 1 des & v= v s ⎡ LCV ⎤ v = ⎢ MCV ⎥ ⎢ ⎥ ⎢ ⎣ NCV ⎥ ⎦ EE392m .Spring 2005 Gorinevsky Control Engineering 11-17 . v )u v & des − F ) u = G −1 ( v • Honeywell MACH • Dale Enns X-38 .

Dynamic inversion in flight control • • • • NASA JSC study for X-38 Actuator allocation to get desired forces/moments Reference model (filter): vehicle handling and pilot ‘feel’ Formal robust design/analysis (µ-analysis etc) EE392m .Spring 2005 Gorinevsky Control Engineering 11-18 .

Summary • Dahlin controller is used in practice – easy to understand and apply • IMC is not really used much – maintenance and support issues – is used in form of MPC – Lecture 14 • Youla parameterization is used as a basis of modern advanced control design methods. • Dynamic inversion is used for high-performance control of air and space vehicles – this was presented for breadth.Spring 2005 Gorinevsky Control Engineering 11-19 . – Industrial use is very limited. the basic concept is simple – need to know more of advanced control theory to apply in practice EE392m .