Application of artificial neural networks for transistor open-circuit fault diagnosis in three-phase rectifiers

This study deals with the transistor open-circuit fault diagnosis technique based on the grid current processing. In accordance with the proposed method, in the first stage, the defect of the power electronics converter is recognised. For this purpose, the zero current periods are registered in each converter phase circuits. The faulty transistors are identified calculating the average values of differences between predicted and measured phase currents. The novelty of the presented technique is an application of a neural network for the grid current prediction in the active rectifier. In fact, the transistor open-circuit faults do not affect the predicted grid currents immediately as soon as the transistor defects happen. Therefore, the differences between the predicted currents and the measured ones increase which are used for the faulty transistors identification. In the comparison to the switch open-circuit fault diagnostic techniques, which are known from the scientific literature survey, the method presented in this study is insensitive to load changes no matter a direction of the energy flow in the power conversion system.