Fault Diagnosis, Prognosis and Self-Reconfiguration for Nonlinear Dynamic Systems Using Soft Computing Techniques

Diagnosis and prognosis are processes of assessment of a system's health. The former is an assessment about the current health of a system based on observed symptoms, while the latter is an assessment of the future health. System reconfiguration or accommodation is essential once faults are detected and identified. This paper presents model-free fault diagnosis, prognosis and self-reconfiguration using soft computing techniques (i.e. neural networks, fuzzy logic and Taguchi methods). A three-tank nonlinear dynamic system was chosen to demonstrate the presented techniques. The diagnosis detects and classifies a fault, and also determines the degree of each fault. The prognosis estimates the remaining life at any time. While a fault cannot be repaired by itself, the faulty system can be quickly reconfigured to avoid reaching the catastrophic level. A fast self-reconfiguration technique by means of instant optimization and fuzzy inference is presented.