Ballasted Track Behaviour Induced by Absent Sleeper Support and its Detection Based on a Convolutional Neural Network Using Track Data

Abstract With development of the heavy-haul railway, increased axle load and traction weight bring a significant challenge for service performance safety maintenance railway track. Conducting defect recognition on concrete sleepers ballast using big data is vital. This paper focused detection absent sleeper support in ballasted track with an emphasis integration model-based data-driven methods. To this end, mathematical model consisting wagon, wheel–rail contact subsystems was first established to acquire necessary raw method, which wagon regarded as 47-degree-of-freedom multi-body subsystem, treated multi-layer discrete-elastic beam subsystem support. Then, architectural hierarchy three-layer convolutional neural network (TLCNN) developed, includes three layers two pooling layers, method reconstructing one-dimensional vertical displacement two-dimensional time–space matrix also proposed. Thirdly, verification carried out by comparing simulation experimental results illustrate accuracy reliability model, dynamic behaviour investigated. Lastly, TLCNN used train detect existence Results show that methods reliable effective approach The proposed can extract robust characteristics noisy environment. handle more complex tasks further improve performance, deeper CNN models larger sample sizes should be preferentially considered practical applications.