Faulty measurement substitution and control reconfiguration by using a multivariate flow control loop

Abstract

A two-tank multivariate loop was designed and built to support research related to instrumentation and control, equipment and sensor monitoring. This test bed provides the framework necessary to investigate and test control strategies and fault detection methods applicable to sensors, equipment, and actuators, and was used to experimentally develop and demonstrate a fault-tolerant control strategy using six correlated variables in a single-tank configuration. This work shows the feasibility of using data-based empirical models to perform fault detection and substitute faulty measurements with predictions and to perform control reconfiguration in the presence of actuator failure in a real system. These experiments were particularly important because they offered the opportunity to prove that a system, such as the multivariate control loop, could survive degraded conditions, provided the empirical models used were accurate and representative of the process dynamics.