Robust multivariable tuning methods

The rapid growth in the complexity of modern process plants both in terms of material flow and energy exchange have substantially increased the number of feedback control loops for maintaining desired production conditions and product quality. Traditionally, PID controllers are used in large numbers in all industries as a single-loop controller. The controllers come in many different forms and are often packaged as standard products. The popularity of PID controllers is due to their functional simplicity, which allows process engineers to commission them in a simple and straightforward manner. The most attractive practical property of PID is the guaranteed property of the ‘Integral’ action to eliminate constant set-point error and disturbance offsets. In practice, most industrial processes have multiple numbers of inputs and outputs. The inputs and outputs are arranged or paired to minimise the interaction between the control loops. Static or dynamic decoupling is also used to minimise the loop interactions and make the system diagonally dominant. Multiloop PID or non-interacting controllers can then be used to control MIMO processes. Koivo and Tanttu [10] gave an early survey of multivariable PID tuning techniques. When the process interactions are modest, a diagonal PID controller structure is often adequate. This type of structure is simple and easy to understand. While there are a large number of methods to tune scalar PID controllers, the numbers of multiloop PID tuning methods are few. Attempts have been made to extend some SISO methods to multivariable systems, and some new methods have been developed in recent years. This chapter reviews and compares some of the important multivariable tuning methods in terms of their stability and performance robustness.