Decentralized control for multi-input multi-output integrating processes

This thesis presents a decentralized PI/PID controller design scheme for Multi-Input Multi-Output (MIMO) integrating processes with time delays. A new loop pairing criterion is proposed for multi-loop control of MIMO integrating processes, based on a new interaction measurement method called relative effective gain array (REGA) / relative normalized gain array (RNGA). Compared with its counterparts such as RGA or DRGA, this novel loop pairing technique considers not only the steady-state information but also the transient information of the process. It also comprehensively reflects the dynamic interactions among individual loops without requiring specification of the controller type and with a relatively low computational load. Based on the loop pairing results of a certain multivariable process, its Equivalent Transfer Function (ETF) matrix is then derived for the design of decentralized control systems. The ETF of a process in a certain loop represents the approximation of its transfer function when other loops are closed, and reflects the dynamic changes of the process when other loops are closed. A decentralized controller is then designed based on the ETF matrix. This design scheme guarantees the stability, causality, properness and compatibility of the overall control configuration for MIMO integrating processes. The REGA/RNGA-based design method is simple, straightforward and easy to understand for field control engineers. A few industrial MIMO integrating processes are presented to demonstrate the advantages of the method in terms of simplicity in design and effectiveness in control system performance.MASTER OF ENGINEERING (EEE