Prediction of railway subgrade subsidence based on geological mining conditions

In order to ensure the safe operation of the special railway line in Guozhuang coal mine, the steady-state prediction of the railway subgrade subsidence caused by working face mining was carried out from the perspective of geological mining conditions.Firstly, a GA-BP neural network model between geological mining conditions and the predicted parameters of probability integral was established.After training using the collected data from the surface mobile observatory station,the accuracy and reliability of the model were tested from four aspects：network error analysis, data fitting degree, test results and generalization performance.The results showed thatthe convergence rate of the network is fast, the average relative error of the model is less than 3%, the fitting degree of the data is more than 0.8, and the generalization performance index is more than 0.8. The model has high prediction accuracy and good prediction ability.Secondly, based on the parameters of geological mining conditions of each working face in the study area, the model is used to calculate the predicted parameters of mining subsidence.By comparing the estimated mining subsidence results of S3-13 working face with the measured data, it is found that:the sum of squares of the differences between the two values is 6.21×105, and the median error was 104.38 mm, which is 2.60% of the maximum subsidence value of the observation point. The high accuracy of the prediction results indicated that the obtained probability integral predicted parameters have certain reliability.Finally, the final subsidence contour line and subsidence curve of the railway subgrade are estimated through the steady-state prediction. It is estimated that the maximum subsidence value along the railway line is 4 261 mm, and two sections will form a subsidence basin larger than 4 000 mm. The railway will be severely deformed and damaged.The research results can provide a theoretical basis for the mining of the subsequent working face and the prediction and maintenance of the dynamic deformation of the railway.