高应变率下岩石的蚁群智能动态本构模型

针对蚁群算法在优化网络权值过程中存在搜索速度慢和精确度的问题,将梯度下降法作为优化算子嵌入蚁群神经网络,提出了一种新型混合蚁群神经网络,并根据花岗岩的高应变率动态实验,构建了一种动载高应变率作用下花岗岩本构关系分析的蚁群智能模型,其拟合结果与实际情况吻合,对于研究在高应变率下岩石动力学性能具有借鉴意义。

An Intelligent Ant Colony Neural Network Model for the Dynamic Constitutive Relationship of Rock under High Strain Rate

To solve the disadvantages of ant colony algorithm including slow searching speed and low precision in optimizing weights of neural network,a novel hybrid ant colony neural network is proposed,in which gradient descent method is set as a kind of optimization operator,and based on the results of compression test under high strain-rate,a intelligent ant colony model on the constitutive relationship of granite under high strain-rate induce by dynamic load is constructed.The simulation results of the intelligent model are very close to the test results, this shows that the constructed model is valuable for studying on the dynamic mechanic properties of rocks.