Abstract: In order to evaluate rock burst risk degree, a new evaluation method based on the model of particle swarm optimization (PSO) algorithm and BP neural net is proposed. The method employs the technology of BP neural network to establish a regression model and takes advantage of PSO algorithm to optimize the parameters of the network, namely, the weights and the threshold values. All the data used in the model are from existing rock burst database. As a result, the BP neural network cannot fall into local optimum and its convergence speed can be increased. Then, ten major factors are selected as the inputs and twenty practical cases from a characteristic rock burst coal mine are chosen to establish the PSO-BP evaluation model. The newly established model is further compared with the standard BP model. Results show that the evaluation accuracy is increased by 15 percent by the proposed model in this paper. Finally, the proposed PSO-BP evaluation model is finally applied to evaluate the rock burst risk degree of two coalmines, where the feasibility and applicability of the method are verified.

Key words: mining engineering, rock burst, risk evaluation, neural network, particle swarm optimization