论文部分内容阅读
将人工神经网络与遗传算法相结合， 提出了一种用于位移反分析的进化神经网络方法。这种方法基于正交试验获得的样本进行学习， 用遗传算法搜索最优的神经网络结构， 并用最佳推广预测学习算法训练此网络， 以此训练好的网络描述岩体（土）的力学参数与岩体位移之间的非线性关系， 再应用遗传算法从全局空间上搜索， 进行岩体力学参数的最优辩识。作为例子， 文中给出了弹性问题的反分析， 结果是令人满意的。
Combining artificial neural network and genetic algorithm, an evolutionary neural network method for displacement back analysis is proposed. This method is based on the samples obtained from the orthogonal experiment, searches for the optimal neural network structure with genetic algorithm, and trains the network with the best generalized predictive learning algorithm. The trained network describes the mechanical parameters of rock mass (soil) And the nonlinear relationship between the displacement of rock mass, and then use the genetic algorithm to search from the global space, the rock mass mechanical parameters of the optimal identification. As an example, the paper gives a back analysis of the elasticity problem, and the result is satisfactory.