Control and Fault-Handling Subject to Asynchronous Measurements

Abstract

This chapter addresses the problem of control and fault-handling subject to asynchronous measurements and data losses. First, an approach for handling sensor data losses via Lyapunov-based model predictive control is developed. Specifically, in this control scheme, when feedback is lost due to sensor data losses, the actuators implement the last optimal input trajectory evaluated by the controller. This control scheme allows for an explicit characterization of the stability region and guarantees practical stability in the absence of sensor data losses. Application of the control scheme to a continuous crystallization process subject to sensor malfunctions is presented to illustrate the robustness of the control scheme when the process is subject to measurement unavailability, asynchronous sampling and parametric model uncertainties. Next, an integrated fault detection, isolation, and fault-tolerant control framework is applied to a polyethylene reactor system where several process measurements are not available synchronously. First, an FDI scheme that employs model-based techniques is designed that allows for the isolation of the faults. This scheme employs model-based FDI filters in addition to observers that estimate the fault-free evolution of the asynchronously measured states during times when they are unmeasured. The FDI scheme provides detection and isolation for a fault where the fault entered into the differential equation of only synchronously measured states, and grouping of faults where the fault entered into the differential equation of any asynchronously measured state.