Jixie chuandong (Jan 2016)
Study on the Bearing Fault Diagnosis based on Feature Selection and Probabilistic Neural Network
Abstract
To improve the aero- engine fault diagnosis accuracy grade,by using the DET and PNN classification techniques,a bearing fault diagnosis technique based on feature selection and PNN is put forward.Firstly,the bearing fault test data are extracted to form the multi- domain fault diagnosis feature set composed of 14 time- domain features and 13 frequency- domain features. Secondly,to increase classification efficiency and reduce the influence on classification result from coupling characters between features,the feature selection technique based on DET is applied to obtain feature parameters which can be classified easily. On this basis,the PNN technique is applied to carry on research of bearing fault diagnosis. The bearing simulation fault experiment data is applied for verification,the results prove that compared with diagnosis techniques of BP neural network and support vector machines,the PNN is higher in the respect of diagnosis accuracy grade. Meanwhile,the efficiency and accuracy grade of diagnosis are further improved for the reason of employing feature selection technique.
