Classification of Fault Detection Methods

Fault detection and isolation are a prerequisite of fault diagnosis necessitating system observability and proper signal measurement and processing in the presence of noise. The thus obtained information then is to be processed in real time by a decision support software system that takes into account that sensors themselves can be faulty. According to Venkatasubramanian et al. [14-16], fault detection and diagnosis methods may be classified into 

Model-based fault detection methods make use of a model to obtain the a priori knowledge needed for fault diagnosis. The development of a model can be a challenging task. Modelling assumptions, decisions on which effects shall be taken into account and how to model them clearly affect the quality of diagnosis results. Model-based FDI depends on the model accuracy. However, once an appropriate model has been developed it allows to obtain information about the expected system behaviour and to compare it with measurements of the real system behaviour. The diagnosis techniques and the diagnosis precision depend on the kind of model used. 

Qualitative models consider cause and effect relationships, express component malfunctions in a qualitative manner and link them with deviations in measurement data. (Qualitative fault diagnosis is based on fault tree analysis [15, 21], or 

Quantitative models describing the dynamic behaviour of a system are often developed from first principles and establish mathematical relations between system inputs and outputs. The computed outputs of a behavioural model, however, will never perfectly match measured outputs of the real system even if no fault has occurred for the following reasons. 

Once a quantitative system model is available, different methods can be used for the generation of a fault indicator or a residual as a primary step in FDI. These methods are either based on observers or a bank of observers [26, 27], on parity relations [28, 29], or on analytical redundancy relations [30, 31], or on parameter estimation [32, 33]. In case a fault has occurred in the system, the time evolution of some residuals must deviate distinctly from that during normal healthy system operation. 

---