Research on Proxy DNN Diagnostic Model for Wet Clutch Brake Fault of Hot Die Forging Press Based on Knowledge Distillation

Hot forging presses are widely used in automotive, aerospace and other industries. However, due to the complex working mechanism of the wet clutch brake system of the hot forging press, fault diagnosis is still challenging. The traditional model-driven fault diagnosis method is difficult to meet the requirements of real-time and accuracy. In this paper, a knowledge distillation agent deep neural network model is proposed for fault detection of wet clutch brake system of hot forging press. The model first uses a regularized DNN architecture for feature extraction and fault diagnosis, and then applies knowledge distillation to transfer knowledge from a large teacher model to a lighter-weight surrogate model, thereby reducing computational load and improving inference speed. The experimental results show that the KDS model achieves similar accuracy to the traditional DNN model, but the inference speed is improved by 37.84 %, and the model parameters are reduced by 72.79 %. The model significantly improves the real-time and computational efficiency of fault diagnosis of hot forging press, and has a good industrial application prospect.