SURFACE SUBSIDENCE PREDICTION IN MINED-OUT BASED ON GENETIC-ALGORITHM-OPTIMIZED BP NETWORK

A new method for training the artificial neural network is presented.In this method,the genetic algorithm(GA),a general-purpose global search algorithm is used to train the network with updating the weights to minimize the error between the network output and the desired output.Then the back-propagation(BP)algorithm is used to further train the artificial neural network.The method is used to speed up the convergence and improve the performance.As an example,the method was used to predict surface subsidence in mined-out.An neural network prediction model was established.A lot of practical surveying data sets are collected to train the neural network.Several sets are used to predict the surface subsidence by the neural network.Results show that the neural network prediction model is of high convergent speed and good prediction precision,so the model offers a useful approach for surface subsidence prediction in mined-out.