A Novel Deep Parallel Time-Series Relation Network for Fault Diagnosis

Currently, deep learning-based methods are widely used in the fault diagnosis of time-series data for their high precision. However, application traditional learning models is limited by calculational efficiency and poor interpretation ability. To address problems, a model named parallel relation network (DPTRN) proposed this article. There three main advantages DPTRN. First, our time relationship unit can perform feature extraction on each node sample simultaneously; therefore, DPTRN performs way improves computing significantly. Second, improving absolute position embedding, novel decoupling embedding be directly applied learn contextual information. Third, has an obvious advantage interpretability compared with models. Applying four datasets, we achieved higher performance much lower cost, which indicates effectiveness, efficiency, model.