Attention-based Convolutional Neural Networks for Diesel Fuel System Fault Diagnosis

Diesel fuel system is a significant part of the diesel engine, whose stable and reliable working state is the key guarantee for safety and efficiency of the whole system. It is essential to conduct health monitoring and fault diagnosis for the diesel fuel system based on intelligent technology. Recent years, deep learning has become an effective means to perform intelligent fault diagnosis for various mechanical system and convolution neural networks have achieved remarkable success in various applications. In order to achieve efficient and accurate fault diagnosis for diesel fuel system, an improved convolution neural network combined with attention mechanism is established in this paper. The additional attention coefficients are learned to provide importance bias for the extracted features so that the deep model can make the most of the key signature associated with the working condition and ignore the irrelevant intermediate features in the consideration of computational efficiency. Experiments have been carried out and the results have proved the model effectiveness and efficiency.