Multi-scale Convolutional Neural Networks Based on Self-attention And Residual Network for Industrial Equipment Fault Diagnosis

Texts in industrial production contain massive amounts of faulty textual information, as traditional convolutional neural networks (CNN) have difficulties in extracting sufficient valid information in the multi-classification task of texts. A multi-scale CNN composite model based on a self-attention and an Inception residual connection module is proposed. The method preprocesses unstructured text, obtains a text vector from the input text by word embedding, extracts feature at more scales using the CNN and Inception modules, and assigns weights to the feature vectors through the self-attention so that the model can pay more attention to the key information in the text, thus improving the performance of the model in text classification tasks. Taking the industrial equipment fault data of an automotive company as an example, the proposed method outperforms the traditional model in terms of precision, recall and f1-measure evaluation indexes, and provides a new idea for the research of industrial production equipment fault diagnosis.