Fault detection and isolation for an experimental internal combustion engine via fuzzy identification

Abstract

Abstruct-Certain engine faults can be detected and isolated by examining the pattern of deviations of engine signals from their nominal unfailed values. In this brief paper, we show how to construct a fuzzy identifier to estimate the engine signals necessary to calculate the deviation from nominal engine behavior, so that we may determine if the engine has certain actuator and sensor "calibration faults." We compare the fuzzy identifier to a nonlinear ARMAX technique and provide experimental results showing the effectiveness of our fuzzy identification based failure detection and identification strategy.

Fault diagnosis of automobile hydraulic brake system using statistical features and support vector machines

“…The commonly used techniques for the selection of features are the principal component analysis (PCA) [5], genetic algorithm (GA) [6,7] and decision tree (DT) [8]. The principal component analysis is a method that reduces the dimensionality of data by performing a covariance analysis between factors.…”

Hybrid Genetic-Fuzzy Systems for improved performance in Residual-Based Fault Detection

“…These system models are then able to serve as fault-free reference situations, based on which residuals can be calculated, pointing to deviations between expected behavior (the predictions from the models) and the real behavior (the recorded measurements). The usage of data-driven fuzzy systems as system models in the context of FD has been proposed, for instance, in [22] [15] and in our preliminary works in [21] [20]. The main strengths of data-driven fuzzy systems are their good approximation capabilities (proved universal approximators [2]), while at the same time they offer an easy interpretation of the model components by readable rules [16].…”

Feature-based fault detection of industrial gas turbines using neural networks

“…Compressor inlet pressure 1 measurement 0-1300 mbar abs a 8 Compressor outlet pressure 2 measurement-redundant 0-30 bar a 9 Compressor inlet temperature 3 RTD 0-400 • C a 10 Compressor outlet temperature 2 RTD 0-350…”

“…easier to interpret. They also have inherent abilities to deal with imprecise or noisy data, therefore making them suitable for model-based fault diagnosis [8][9][10][11][12][13].…”