High-Speed Train Health Assessment Based on Degradation Stages and Fault Classification by using Dual Task LSTM with Attention Mechanism

Health and Safety Assessment is critical for high-speed trains to maintain their operation continuously. Health assessment based on degradation levels to predict RUL and fault diagnosis are two key task events to decide the end of life for bearing or decide maintenance actions. However, a gap exists as these tasks are predicted separately. This study proposes a dual-task LSTM model with an attention mechanism to predict RUL based on degradation levels and diagnose the fault classification. This enables more accurate fault assessment that helps reduce maintenance costs. The performance of the proposed model demonstrates and leads over other state-of-the-art methods and proves an excellent accuracy of its new application with the potential to decide the end of the Life of Bearing based on degradation levels with fault classification.