Abstract:

Three-phase continuous magnetic brushing machines are used in many applications where reliability and error tolerance are important. In this study, for the simulation of the failure of a continuous magnetic brush-free straight stream (BLDC) engine with an external rotor, the stator phase circuit has been modeled by adding an additional parallel empedans to the shell. The failure empedance is necessary to calculate the circulation flow that causes a magnetic disturbance. The engine flow signature analysis (MCSA) takes into account the change in the current signs of the defective engine and analyses are carried out for the detection of defects in different intensities. The short circuit failure modeling is carried out with the final elements method (SEY) and the program used with the stator output signs has been obtained by applying Fast Fourier Conversion (FFT). Classifiers used to predict the formation and severity of short circuit defects by applying different classification methods such as the nearest neighbor (k-NN), multi-layer sensor artificial nerve networks (MLP), decision trees (RT). After comparing the accuracy of the defect detection with the two differences identity extraction, the appropriate identity data set has been selected and the short circuit error of the criteria presented has been detected. In the failure detection, the MLP classification method has a 80% success rate, but the k-NN and RT methods have achieved 100% success. In this case, k-NN and RT methods applied for failure detection appear to be quite successful.

Keywords: