DETECTION AND ISOLATION OF SENSOR FAULTS IN NONLINEAR SYSTEMS: A CASE

Abstract: In control systems, the most natural way to model failures of hardware components is introducing additional inputs (fault inputs) that equal the difference between actual and assumed values of the ‘faulty ’ quantities. However, when applied to faults affecting the measure of the state of a nonlinear system, this natural modeling procedure results, in general, in non-standard models that are nonlinear in the fault inputs, whereas an affine dependence is desired to exploit the available fault detection and isolation techniques. Through the analysis of a case study (the IFATIS Heating System Benchmark), we illustrate how to alternatively model the failure of a sensor, in particular, by a set of (always concurrent) fault inputs, so that the standard, affine in the inputs, form of nonlinear systems is kept. With the model in this form, the recently developed idea of Fault Set Detection and Isolation (FSDI, (Mattone and De Luca, 2003b)) can be applied to detect sensor failures under the hypothesis of their non-concurrency