Feedforward Control Prof Ing Michele MICCIO Dip Ingegneria

Feedforward Control Prof. Ing. Michele MICCIO Dip. Ingegneria Industriale (Università di Salerno) o Prodal Scarl (Fisciano) o adapted from Romagnoli & Palazoglu’s Chapter 16: Model-Based Control see also Stephanopoulos, 1984 Chapter 21 § 21. 1 -4 rev. 3. 4 of April 30, 2019

Introduction to Model-Based Control In this course we consider the following control design techniques that explicitly use the process model: Ø Ø Delay Compensation (Smith Predictor) Inverse Response Compensation Feedforward control Model Predictive Control (MPC) Romagnoli & Palazoglu, “Introduction to Process Control “ 2

Introduction to Model-Based Control Definition of Model-Based Control Detailed Process Understanding Combination of detailed process understanding (advanced mathematical modeling) with the intelligent use of modern control systems (hardware, software and technology). Romagnoli & Palazoglu, “Introduction to Process Control “ Intelligent Use of Modern Control Systems Improved Profitability $ 3

Back to Feedback Control Ø Feedback control can never achieve perfect control of a chemical process. It reacts to the changes in the controlled variable after a deviation is detected in the output. Disturbance Set-point Controller Manipulated variable Process Controlled variable Sensor Romagnoli & Palazoglu, “Introduction to Process Control “ 4

Feedforward Control Ø A feedforward controller measures the disturbance directly and takes control action to compensate for its eventual impact on the output variable. Ø Feedforward controllers have theoretical potential for perfect control. Romagnoli & Palazoglu, “Introduction to Process Control “ 5

Feedforward Control Consider the following feedforward flow of information about disturbance … Disturbance Set point Controller output Feedforward Controller Final control element Manipulated variable Process Controlled variable the feedforward controller predicts the effect of disturbances Romagnoli & Palazoglu, “Introduction to Process Control “ 6

Feedforward Control Design We want to achieve the following control Therefore, in the Laplace domain: objective: y(t) = ysp(t) d(s) gd(s) Process m(s) gp(s) y(s) 1 We shall require perfect control : Romagnoli & Palazoglu, “Introduction to Process Control “ 7

Feedforward Control Design We introduce a suitable structure for the feedforward controller. Then, we further determine m(s) from the block algebra: 2 y sp g 1 ff disturbance measurement + − g md process vs. disturbance gd g 2 ff actual feedforward controller 3 d o g m f final control element g + + y p process Romagnoli & Palazoglu, “Introduction to Process Control “ 8

Feedforward Control Ø The feedforward control elements conventional controllers (P, PI or PID) are not Ø The feedforward controller: § needs the gff 1 block in order to make the set point comparable to the measured disturbance § depends on the knowledge of process and disturbance models § can be developed for more than one disturbance and for multiple controlled variables Romagnoli & Palazoglu, “Introduction to Process Control “ 9

Feedforward vs Feedback Feedforward - Advantages Ø Ø Ø Acts before disturbances affect the process Cannot cause instability Good for slow process dynamics Feedforward - Disadvantages Ø Ø Ø Must identify and measure ALL disturbances Fails for unmeasured disturbances Needs to have a reliable process dynamic model Fails for changes within the process No indication of control quality Romagnoli & Palazoglu, “Introduction to Process Control “ 10

Feedforward vs Feedback - Advantages Ø No disturbance measurements needed Ø Limited or even no process model needed Ø Can cope with changes within process Feedback - Disadvantages Ø Will always be some error Ø Poor for slow process dynamics, interaction, etc. Ø Instability is possible Romagnoli & Palazoglu, “Introduction to Process Control “ 11

Feedforward-Feedback Control Use a combination of Feedforward and Feedback control We expect that a combined feedforward-feedback control system will retain, The superior performance of a feedforward controller, and v The insensitivity of the feedback controller to uncertainties in model and inaccuracies in model parameters. v Romagnoli & Palazoglu, “Introduction to Process Control “ 12

Example 1: Process with dead time Consider the following process TFs: gmd and gf purely algebraic Ø Design a Feedforward Controller Ø Compare with PI feedback design Romagnoli & Palazoglu, “Introduction to Process Control “ 13

Example 1: process with dead time (disturbance rejection) FF Controller Feedback with a PI controller No modeling error with +20% error in gd gain (Kd = 1. 2) NB: Output signal = controlled variable different scales on axes ! Romagnoli & Palazoglu, “Introduction to Process Control “ 14

Example 2: Heat Exchanger Manipulated variable Tout, sp Steam Feedforward Controller T T FT Controlled variable Tout Disturbances Design a FF controller to compensate for variations in the feed flow rate and temperature. Romagnoli & Palazoglu, “Introduction to Process Control “ 15

Example 3 Distillation Column Design a FF controller to compensate for variations in feed composition and flow rate. Manipulated variable Disturbances C T Cout, sp Controlled variable FT FF Controller Romagnoli & Palazoglu, “Introduction to Process Control “ 16