关键词:
Sensitivity analysis
摘要:
Subway spatial route shape significantly affects rail wear. If effects of spatial route shape on rail wear can be considered during the route selection stage to optimize route plan, rail wear can be reduced at source. However, feasible route schemes have infinitely many, and a large number of schemes need to be designed for comparative analysis during route selection stage. The existing mode to establish simulation model and then calculate rail wear can consume a lot of time to significantly affect design efficiency, and long calculation time is difficult to accept. Here, an artificial neural network was introduced to explore correlation between subway spatial route shape and rail wear, and realize efficient and accurate prediction of rail wear under different subway spatial route shapes. Firstly, a rail wear calculation model was established based on the prototype of subway A-type car, and effects of different spatial route shapes, parameters on wears of inner and outer rails were analyzed, and sample datasets were established. Furthermore, based on the analysis results, a neural network model was established to predict rail wear considering circular curve radius, circular curve length, transition curve length, slope, slope algebraic difference and superelevation, and a model structure optimization method based on Sobol analysis was proposed. Finally, through sample data training, a mapping relation between subway spatial route shape and rail wear was established. The study results showed that for inner and outer rails, the same route shape parameters have different effects;for inner rail wear, the main affecting parameters are transition curve length, circular curve radius and superelevation;for outer rail wear, circular curve radius is the main affecting parameter, and its affecting degree is much higher than other parameters;rail wear can be correctly predicted according to subway spatial route shape parameters, and prediction accuracies of inner and outer rail