Estimation of the Case Temperature of Insulated Gate Bipolar Temperatures in Induction Cooktops by Deep Neural Network

One of the most common causes of failures in inverters is the thermal stress acting on the Insulated Gate Bipolar Transistors (IGBT). This work develops a model to estimate the case temperature of IGBT, which is directly correlated with the thermal stress and can, therefore, be used for monitoring the IGBT degradation. Specifically, the model receives in input electric signals measured by sensors and produces as output the case temperature estimation. Considering the complexity of the input-output relationship and the large number of measured signals, a deep neural network is built by adding a classification layer at the top of a stacked auto-encoder. The model is successfully applied to inverters of industrial cooktops. The obtained results show that it provides more accurate estimates of the case temperature of IGBT than a wrapper feature selection method based on the combination of a Genetic algorithm and a neural network.