基于APSO-BP耦合算法的岩体力学参数反馈研究  

Back-Analysis of Rock Mechanic Parameters Based on Coupling Algorithm of Adaptive Particle-Awarm-Optimization and Back-Propagation Neural Network

作　　者：杜小凯[1] 任青文[2] 郑治[3] 张国华[4] 许传华[5] 

机构地区：[1]河海大学水利水电工程学院,江苏南京210098 [2]河海大学土木工程学院,江苏南京210098 [3]中国水电工程顾问集团贵阳勘测设计研究院,贵州贵阳550002 [4]中国灌溉排水发展中心,北京100054 [5]中钢集团马鞍山矿山研究院,安徽马鞍山243004

出　　处：《中国矿业大学学报》2008年第6期756-762,共7页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(50539030);国家重点基础研究发展计划(973)项目(2007CB714104)

摘　　要：提出了基于自适应粒子群优化(APSO)与误差反向传播(BP)神经网络耦合反馈分析模型(APSO-BP).模型实现对网络结构、权重、阈值的同时优化,借助自适应粒子群算法全局优化能力强、收敛速度快的特点,提高了模型运算效率.采用Schaffer基准函数对该模型和传统遗传算法、BP神经网络、粒子群与BP神经网络组合算法进行测试对比.结果表明该模型更为优越.应用该模型对索风营水电站地下岩体力学参数进行反馈分析,计算位移值与实测值吻合较好,平均误差0.22 mm.The coupling feedback analysis model （APSO-BP） was proposed based on the adaptive particle swarm optimization algorithm and the Back-Propagation neural network. The model succeeds in optimizing the construction, the weights and the thresholds of the neural network at the same time, which gives the full stochastic global optimization ability of APSO, greatly improving the operation efficiency of the model. The Schaffer＇s function was applied to test the model, the genetic algorithm, the BP neural network and the PSO-BP algorithm. The results show that the APSO-BP model takes smaller time and has more precision level. The model is used to optimize the rock mass mechanical parameters of underground grotto of suofengying hydraulic power plant. The calculated result was close to the measured values; and the average error was 0.22 mm.

关 键 词：自适应粒子群算法 BP神经网络 反馈分析 岩体力学参数 索风营水电站 

分 类 号：TU45[建筑科学—岩土工程]