基于KP CA和CIP SO￣P NN的煤与瓦斯突出强度辨识模型∗

摘要

In order to achieve accurate identification for coal and gas outburst,a new algorithm based on Kernel Principal Components Analysis( KPCA) and Improved Probabilistic Neural Network is proposed. Based on different conditions of coal seam and operation,the original index parameters that affect coal and gas outburst are determined and dealt with dimension reduction by KPCA method. The chief factors representing coal and gas outburst are ex￣tracted and put into PNN for identification of coal and gas outburst. Aiming at the incorrect classification defect caused by single smoothing factor,Chaos Immune Particle Swarm Optimization( CIPSO) is adopted to optimize pa￣rameters of PNN,which avoids the influence of artificial factors and improves the identification accuracy of the mod￣el. The numerical analysis results show that the proposed method has better performance compared with BP,PNN and PSO￣PNN approaches.%为了实现对煤与瓦斯突出强度等级的准确辨识，提出将核主成分分析( KPCA)和改进概率神经网络相结合，建立煤与瓦斯突出的强度辨识模型。根据煤层条件和生产条件，确定影响煤矿瓦斯突出的相关基础参数并对其进行测定，采用KPCA对该参数集进行降维处理，提取出可以表征煤与瓦斯突出的敏感参数作为辨识模型的输入值。利用混沌免疫粒子群算法( CIPSO)优化概率神经网络(PNN)的σ参数，以克服PNN中平滑参数σ单一而导致的分类错误，避免了人为因素的影响，提高辨识模型的精度。实例分析结果表明，相比BP、PNN、PSO￣PNN等方法，该方法对煤与瓦斯突出强度进行辨识，结果更为准确。