Nonlinear Fault Detection and Isolation in a Three-Tank Heating System

Abstract

We consider the fault detection and isolation (FDI) problem for a nonlinear dynamic plant (the IFATIS Heating System Benchmark) affected by actuator and/or sensor faults. A general procedure is proposed for modeling faults of sensors that measure the state of a nonlinear system, so that each sensor fault is typically associated to a set of (always concurrent) fault inputs and the resulting dynamic equations are affine in the introduced fault inputs. This allows the application of recently developed nonlinear FDI techniques, slightly extended to cover the considered model structure. For the presented case study, assuming nonconcurrency of the faults in the hardware components, we describe in detail the modeling procedure, the synthesis of residual generators, and the design of a combinatorial logics that recovers perfect isolation. Simulation results are reported in the presence of input and measurement noise