摘要
现有各种位移反分析方法均存在着这种或那种不足之处基于最优化理论的位移反分析方法,解的稳定性较差,易陷入局部极小,反演参数较多时收敛速度较慢,且难以搜索到最优解;基于人工神经网络的位移反分析方法,当解空间稍大时便难以收敛到所需要的精度,且训练结果不具有唯一性,因而很难获得与实际岩体相吻合的反演结果;基于遗传进化的位移反分析方法,需对搜索过程进行大量经验性干预才能搜索到最优解;基于遗传进化和神经网络的位移反分析方法,亦只在较小的解空间内才有效。针对这些不足之处,应用自适应神经模糊推理系统的原理,建立了位移反分析的自适应神经模糊推理方法,并应用该方法对所设定的某一标准弹塑性问题的力学参数进行了反演,反演结果表明,在较大的解空间内,这种位移反分析方法收敛速度快、解的稳定性好、反演结果精度高,是一种优异的位移反分析方法。
Current approaches for back analysis of displacements have some shortcomings. For optimization theory-based approach,it is easy to get trapped at local optimalization,instable in identifying optimal solution,and slow in converging for many parameters to be inversed. For artificial neural network based-approach,it is difficult to converge with desired accuracy when the search space is relatively large,and,furthermore,to get the sole inversion results because of the indefiniteness of trained network. The genetic algorithm-based approach can get the optimal solution only if experience-based interference is implemented. The approach based on artificial neural networks and genetic algorithms is effective only if the search space is relatively small. As a result,new approach for back analysis of displacements has to be established to overcome the shortcomings. Adaptive neuro-fuzzy inference system is used to establish the new approach for back analysis of displacements. The approach has been used to inverse the mechanical parameters of a prescribed elasto-plastic problem. The inversed results show that this approach can rapidly get a stable and accurate solution within a relatively large solution space and the approach is superior to current approaches.
出处
《岩石力学与工程学报》
EI
CAS
CSCD
北大核心
2004年第18期3087-3092,共6页
Chinese Journal of Rock Mechanics and Engineering
基金
国家自然科学基金(50274043)
湖南省自然科学基金重点项目(01JJY1004)
湖南省教育厅科学基金(01A015)联合资助课题。
关键词
神经网络
模糊推理
位移反分析
弹塑性力学
numerical analysis,adaptivity,neural networks,fuzzy inference,back analysis of displacements