Implementation of Multilayer Perceptrons for the Detection of Broken Rotor Bar Faults in an Induction Machine

Induction machines are mostly known for their simplicity and robustness and used as a drive for many industrial applications, converting electrical to mechanical energy. As with all electrical machines, induction machines are prone to degradation over time requiring expensive repairs and maintenance. Maintaining the induction in working conditions helps to prevent the unwanted effects caused by the faults. One of the faults that can significantly impact the performance of the induction machine is the broken rotor bar faults. The operation of an induction machine with broken rotor bar faults induces the line current above and below the fundamental frequency termed as sidebands. This paper employs the multilayer perceptron to detect the broken rotor bar faults using the position and magnitude of the sidebands. An experimental dataset is obtained consisting of the position and magnitude of the sidebands and corresponding machine state. The dataset is randomly divided into two parts: 75% for training and 25% for testing purpose. The performance of the model is then evaluated using a confusion matrix.