Analysis and Modeling for the Real-Time Condition Evaluating of MOSFET Power Device Using Adaptive Neuro-Fuzzy Inference System

Aging has been generally regarded as one of the principal root causes of power device failure, due to it makes the performance of power device degradative, which will directly influence the electrical–thermal performances of the power device. Thus, it is essential to assess the condition to improve the operating reliability of power device. In this paper, a finite-element analysis (FEA) model of MOSFET is built for simulation. Then, the main impact of failure on the overall characteristics of the MOSFET power device will be discussed and analyzed based on the FEA simulation model of MOSFET. In addition, some suitable feature parameters that can indicate the condition of MOSFET are selected and a methodology is proposed for on-line evaluating the real-time condition without intruding the power device by recognizing the aging rate. In this method, MOSFET is deemed as a whole system considering only external feature parameters, and all feature parameters are classified as the inputs, while the aging rate is considered as the output. First, the feature parameters are extracted by especial measurement circuits, a model is built for evaluating the condition of MOSFET using the adaptive neuro-fuzzy inference system, and a reasonable evaluation criterion is established. Then, a case of practical application study for the monitoring method is illustrated based on the evaluation model. Finally, the real-time condition can be monitored, and the condition grade of aging can be evaluated so that the operator can take some measures to maintain the power device by means of this method. The results of a practical application validated the effectiveness of the proposed methodology.