Classification of Multiple Fault in an Automobile Engine Using Statistical and ANN Based Classifiers

The proposed system follows a model?based approach based on Digital Signal Processing and Artificial Neural Network. Fault Detection and Isolation (FDI) of an Automobile Engine’ have been carried out using acoustic signals which is captured from the engine. This method is based on parameter estimation, where a set of parameters is used to check the status of an engine and a model based approach is employed to generate several symptoms indicating the difference between faulty and non-faulty status. In this work, experimentation is carried out on Maruti Suzuki Alto Four Stroke (MSAFS) Engine. There are many more types of faults which may be developed because of wear and tear or lack of maintenance but, the database is generated only for three different types of faults and classification of the same is carried out. The signal normalization, signal conditioning, signal decompositions, analog to digital conversion and feature extraction were carried out by using the algorithm written in MATLAB R2010B. The paper describes the performance of statistical and Artificial Neural Network (ANN) based classifiers for individual and multiple faults and finally the optimal classifiers are proposed based on classification accuracy. It is observed from the experimental results that ANN) based classifiers are more appropriate than statistical classifiers. It is also observed that the magnitude of MSE is under permissible limits and percentage Average Classification Accuracy (% ACA) is also reasonable.