Prediction of area of gray-spots flaw in alternate rail flash butt welded joint based on RBF neural network

On the basis of imported AMS60 alternate rail flash butt welding machine, the welding current, the welding voltage and the displacement of welding procedure experiment of U71Mn rail were acquired with high frequency.Eight weld quality characteristic values such as the percentage of the flashing time of the accelerated flashing stage, the percentage of the flashing time of low voltage Ⅱ and stable flash stage, the power input of weld, the flashed length of rail, the welding time, the short and broken circuit factor of low voltage Ⅱ and stable flash stage and the short and broken circuit factor of the accelerated flashing stage and upsed length, which had influence on the grey-spot flaw area in the alternate rail flash butt welded joint, were used as input data of radial basic function neural network the rail weld grey-spot flaw.The prediction model whose spread rate was 1.5 was built, and according to the TB/T1632-2005, the prediction accuracy of the model trained using 19 samples of 29 samples reached 100% using the other 10 samples.