In many applications, the process equipment is characterized as having multiple
manipulated inputs. This type of equipment configuration is often referred to as a multiple input/multiple output process. The process control consists of multiple PID blocks where a process output and process input pair are associated with each PID. The challenge in designing and commissioning controls for this type of process is interaction, that is, each manipulated input may impact multiple controlled parameters. In general, this interaction may impact the control of two or more loops. A cycle may be established in which the loops fight each other. In such cases, it is important to pair the controlled and manipulated parameters to minimize the interaction between the control loops. Such interaction can significantly degrade control and, in some cases, can cause sustained oscillation.
During system commissioning, the degree of interaction associated with two or more loops may not be immediately obvious. For example, during startup, all loops may be initially placed in Manual. When one loop is commissioned, then no interaction will be observed since the interacting loops are in Manual. Only when you are attempting to tune a second loop with the first loop in Automatic will the interaction between these loops become obvious.
The textbook solution for addressing interactive processes is to install a de-coupler between the PID and Analog output blocks as illustrated below.
A change in the output of each PID may then impact one or more manipulated parameters with the net effect being that only one controlled parameter is impacted by the change. In theory, a decoupling network may be used to address interactive processes. However, in practice decoupling networks are seldom used because of the complexity involved in commissioning and maintaining the proper values of the gains used in the decoupling network. In actual practice, the fighting between interactive loops is most often addressed by simply detuning one of the control loops. The valve associated with the detuned loop will change very slowly. Thus, the two loops will tend not to interact. If there is an operational requirement for both controlled parameters to be tightly controlled, then as address in Chapter 14 of Control Loop Foundation – Batch and Continuous Processes, one approach to effective control of interactive loops is the use of model predictive control as illustrated below.
The structure of model predictive control addresses process interaction without making compromises in control performance.