Abstract: Rail side wear is the key factor influencing the service life of sharp curve rail in heavy haul railway. In order to study the development law of curve rail side wear in Shuohuang heavy haul railway, the development features of rail side wear under various curve radiuses are analyzed, a prediction model of curve rail side wear based on nonlinear autoregressive neural network (NARX neural network) is established, and the rail replacement strategy for Shuohuang heavy haul railway is proposed, based on the long-term data of rail side wear, line equipment, operation and repair. The results show that the curve rail side wear in Shuohuang heavy haul railway grows from slow to fast, following a two-phase development law. Rail side wear is affected by many factors, and the development law of rail side wear may vary even under the same curve radius and the same rail material. As for the curves with radius greater than 800m, side wear is not the dominant factor of rail replacement. The root mean square error and the correlation coefficient of the prediction model are 0.74 mm and 0.84 respectively, showing a strong correlation between the predicted value and the actual value. Finally, a rail replacement strategy based on the combination of state repair and periodic repair for Shuohuang heavy haul railway curve rail is proposed, so as to provide help for arranging rail replacement plans and ensuring proper use of the rail.

Key words: heavy haul railway, sharp curve, rail side wear, prediction model, rail replacement