Explicit Multi-model Predictive Control for fault accomodation.

Abstract

Technological development has enabled industrial process have a high performance in terms of reliability, dependability, maintainability and other safety, operation and environment requirements. In cases that the process is critical in these requirements, the occurrence of a fault could make it impracticable to reach them. As a solution to this problem, research of different strategies for fault-tolerant control systems have been developed in order to guarantee that, once the occurrence of a fault is given, the main objectives would be achieved. The control strategy known as Model Predictive Control (MPC) has been widely used in industry and also in fault-tolerant approaches. But, the MPC is limited by the high computational demand required at the real-time optimization problem. Predictive control strategies in which the optimization problem is solved a priori have been proposed to reduce the computational effort. This paper has as objective to study and implement strategies MPC whose control problem is solved a priori. Multi-model approaches will be investigated in order to satisfy some performance indices that guarantee the correct system operation given a fault presence. Only abrupt faults occurrences that will be diagnosed by an appropriated system will be considered. The results will be compared to the state-of-the-art and evaluated through computational simulations and, as a case study, an implementation in the Quanser company's kit available in the computer control laboratory at ITA.