Performance Degradation Estimation of High-Speed Train Bogie Based on 1D-ConvLSTM Time-Distributed Convolutional...

Area of Science:

• Railway Engineering
• Artificial Intelligence
• Machine Learning

Background:

• High-speed train bogie performance degradation is critical for operational safety and passenger comfort.
• Early detection of bogie degradation is necessary to prevent failures and ensure reliable train operation.

Purpose of the Study:

• To propose a novel deep learning model, the 1D-ConvLSTM time-distributed convolutional neural network (CLTD-CNN), for estimating high-speed train bogie performance degradation.
• To validate the effectiveness and superiority of the CLTD-CNN model using experimental data from the CRH380A high-speed train.

Main Methods:

• Development of an end-to-end encoder-decoder CLTD-CNN architecture incorporating time-distributed 1D-CNN and 1D-ConvLSTM layers.
• Introduction of an auxiliary training component and a specialized input format to enhance the learning of performance degradation trends.
• Testing the model on a CRH380A high-speed train under various performance degradation states.

Main Results:

• The proposed CLTD-CNN model demonstrated superior performance in estimating bogie performance degradation.
• The model achieved the smallest estimation error compared to existing methods.
• Experimental validation confirmed the model's effectiveness across different performance states.

Conclusions:

• The CLTD-CNN model provides an effective, end-to-end solution for estimating high-speed train bogie performance degradation without requiring expert knowledge.
• The developed method significantly improves the accuracy of early fault detection in train bogies.
• This advancement contributes to enhanced safety and reliability in high-speed rail operations.