Objectives and Syllabus

Introduces a range of advanced process control methods based on multivariable & model predictive control strategies. Topics covered include model development; plant testing, time series models, non-linear models, applicability & model validation: multiple linear regression: identification; batch and recursive least squares estimators, instrumental variables, generalised and extended least squares: recursive estimation; initialisation, forgetting factors, windup, stability, convergence and consistency: Kalman filtering: adaptive & self-tuning control; controller structures, minimum variance, effect of time delay, adaptive PID, auto tuners: model based predictive control (MBPC); state space models and observers, internal model control (IMC), generalised predictive control (GPC): non-linear control; generic model control, globally linearising control.

Demonstrations and case studies are provided based upon proprietary MBPC packages. Practical classes using MATLAB & SIMULINK provide hands-on experience of regression, estimation, MBPC and filtering.

Module Details
Code: CME 8370 (formerly ACS 670)
Time Allocation: Lectures 40 hours
Assignments 40 hours
Private Study 70 hours
Prerequisites: Modelling and Simulation (CME 8380), Modern Control Systems Design (CME 8382)


7.5 credits


By report on assignment
By 1 x 2 hour examination
Advanced Process Automation


To develop a quantitative understanding of the various complex techniques that underpin advanced (and modern) process control strategies, and an appreciation of how and when to apply them.


  • To develop a quantitative understanding of the least squares based techniques for model identification and estimation.
  • To become familiar with the minimum variance methods as a basis for studying the techniques of self tuning and adaptive control.
  • To provide a basis for applying these techniques in an industrial context.
  • To develop an in-depth understanding of generalised predictive control (GPC) as a vehicle for explaining the principles of model predictive control (MPC).
  • To appreciate the functionality of commercially available packages for realising model predictive control.
  • To introduce some of the techniques of non-linear control.


Prerequisite for this module are the Modelling and Simulation (CME 8380) and Modern Control Systems Design (CME 8382) modules before doing this one.

It is desirable, but not essential, that students have completed (or have some familiarity with the material covered in) the Control Schemes and Strategies (CME 8376) module before doing this one.

Note that there is an interface with the Optimisation and Scheduling module (CME8490) in which the integration of MPC with real time optimisers (RTO) is covered.

Study Modes

This module will be of one week's full-time intensive study consisting of a variety of lectures, informal tutorials for problem solving and structured computer-based laboratory work.  It is followed by an assignment to be carried out in the student’s own time.


The time allocation for practical work provides for use of Matlab and Simulink for exercises on linear regression, data transformation, estimation, MPC and Kalman filtering.

Recommended Texts

  • Camacho E F & Bordons C,  Model Predictive Control,  Springer,  1999. 
  • Dutton K, Thompson S & Barraclough B,  The Art of Control Engineering,  Addison Wesley,  1997.
    Love J,  Process Automation Handbook,  Springer,  2007. 
  • Marlin T,  Process Control: Designing Processes and Control Systems for Dynamic Performance,  2nd Edition,  McGraw Hill,  2000. 
  • Ogunnaike B A & Ray W H,  Process Dynamics, Modelling and Control,  Oxford University Press,  1994.
    Seborg D, Edgar T, Mellichamp, D and Doyle F,  Process Dynamics and Control,  3rd Edition, Wiley,  2011.

Topics Included

Model types.  Role of the model in control.  Plant testing for model building.  Time series models: auto-regressive, moving average, with exogenous outputs (ARMAX), etc.  

Kalman filtering:  Estimator structure.  Luenberger observer and Kalman gain.  Sources of noise and filter formulation.  Discrete time Kalman filter design.  Riccati equation.  Realisation and implementation issues.  Worked example.

Identification by least squares estimation.  Batch and recursive least squares(RLS).  Instrumental variables.  Generalised (GLS) and extended least squares (ELS).  Setting the order.  Identification of time delays.  Initialisation.  Forgetting factors and co-variance resetting.  Stability and co-variance windup.  Convergence: quality, consistency and bias.  Model validation.  Practical applications and case studies.

Adaptive and self-tuning control:  Controller structures.  Implicit and explicit approaches to self-tuning.  Notation.  Cost functions and weighting factors.  Minimum va riance (MV) control.  Implementation and properties of MV control.  Generalised minimum variance (GMV) control.  Implementation and properties of GMV.  Offset and noise.  Effect of time delay.  Self tuning with extended horizon.  Adaptive PID methods.  Auto tuners.  Introduction to adaptive control.

Model predictive control:  Benefits of MPC.  Control and prediction horiz ons.  Generalised predictive control (GPC).  Prediction and use of Diophantine identity.  Controller output sequence.  Recursive implementation.  Extension to multivariable systems.  Constraint handling.  Models for MPC.  Overview of commercially available MPC packages such as dynamic matrix control (DMC).  Practical applications and case studies based upon Connoisseur (Invensys), Control MV (Perceptive) and RMPCT (Honeywell).
Non linear control:  Review of methods for non-linear control system design.  L/A control.  Generic model control (GMC).  Globally linearising control (GLC) and Lie derivatives.