位移反分析的PSA-ANFIS方法研究

地下工程中，通常采用位移反分析方法来确定围岩的力学参数，而现有的位移反分析方法均存在诸多不足之处，为此，采用模式搜索算法、自适应神经模糊推理系统以及正交试验设计方法，建立了位移反分析的PSA-ANFIS方法；同时还对一标准弹塑性问题进行了解析计算，得到了相应的训练数据对、检测数据对和预测数据对；最后采用此方法对该弹塑性问题的力学参数进行了反演。反演结果表明，所建立的位移反分析的PSA-ANFIS方法除了能够充分发挥ANFIS的各种优点外，还能有效地减少模型建立和参数调节过程中的人为干预，因此是一种优异的位移反分析方法。

Study on the PSA-ANFIS method for back analysis of displacement

The methods for back analysis of displacements are usually used to determine the mechanical parameters of surrounding rock in underground engineering. According to the shortcomings of current methods for back analysis of displacements, the pattern search algorithm, adaptive neuro-fuzzy inference system and orthogonal test design method were used to establish a new method for back analysis of displacements. Meanwhile, the data pairs for training, checking and predicting were gained by analyzing a prescribed elasto-plastic problem. Furthermore, the method was used to inverse the mechanical parameters of the prescribed elasto-plastic problem. The results showed that this approach has some advantages of ANFIS and can reduce manual intervention during the process of model establishing and parameters adjusting. It is an excellent approach for back analysis of displacements.