Fault-tolerant predictive control based on linear parameter varying scheme for industrial processes

Abstract

Background: Process safety is a major concern in the researchers community, both in the past and today. However, the hardware complexity and the involved non-linear dynamics of industrial processes could lead to unsatisfactory behavior of traditional control methods.Methods: To cope with these issues, this paper presents a model-based strategy for fault tolerance in non-linear chemical processes. Specifically, an observer-based fault detection and diagnosis scheme was implemented, which generates early and detailed fault information. Therefore, this valuable data was used to compensate the effects induced by actuator and sensor faults throughout the use of an integrated optimization-based identification and model predictive control technique, which allowed to track a reference even in the presence of faults.Significant Findings: This method reinforces the inherent robustness against faults of linear parameter varying predictive controllers. Moreover, the observers convergence and the controller stability were guaranteed in terms of linear matrix inequalities problems. A simulation based on a typical chemical industrial process, the highly non-linear continuous stirred tank reactor shows that the proposed method can achieve satisfactory performance in fault tolerance.