Skip to Main Content
At present, there are many fault conditioning systems for electrical/mechanical systems due to flexibility in data collection for training and testing as entire data can be collected from one or two machines. Only constraint is different machines from which data for training and testing are collected should be of same behavior specifications. But in the case of automobiles, data for training has to be collected from many automobile engines as fault signals will vary in its characteristics from one to another. The acoustic signal produced by a motorbike engine is important information of fault diagnosis in any automobile. Previous approaches have been tried out in silent conditions. Here, acoustic signals are collected in noisy conditions and required preprocessing is done using wavelets to remove the effect of noise. In this paper, hidden markov model and neural network are used for back-end classification and their performance is compared.