基于QGA-SVM的堆石料离散元细观参数标定模型

针对堆石料离散元三轴试验中存在的细观参数标定影响因素多、耗时严重等问题,在总结分析堆石料细观模型现状的基础上,建立基于量子遗传算法(QGA)和支持向量机(SVM)的细观参数标定模型。模型采用拉丁超立方抽样生成细观参数组,并使用离散元计算其应力-应变曲线;采用QGA对SVM进行训练,使其达到最佳学习效果,以模拟细观参数与应力-应变曲线间复杂的非线性关系;依据堆石料室内三轴试验成果,发挥SVM计算速度优势,采用QGA搜索堆石料细观参数,实现堆石料的离散元细观参数标定。堆石料细观参数实例标定结果表明,所建立的模型可快速、精确地标定离散元细观参数,具有工程应用价值。

Mesoscopic parameter calibration model of discrete elements in rockfill material based on QGA-SVM

The triaxial test of discrete elements in rockfill material has the problem of excessive influence factors and time-consuming for mesoscopic parameter calibration. On the basis of summarizing and analysing the current rockfill mesoscopic models, a mesoscopic parameter calibration model based on the quantum genetic algorithm(QGA)and support vector machine(SVM)was established. The Latin hypercube sampling was used to generate the mesoscopic parameter groups, and then the stress-strain curves were calculated by the discrete element method. In order to simulate the complex nonlinear relationship between the mesoscopic parameters and the stress-strain curves, the QGA was used to train SVM to achieve the best learning effect. According to the indoor triaxial test results and taking advantage of the speed of SVM, the mesoscopic parameters of rockfill were calibrated by the QGA searching process. The calibration example of rockfill shows that QGA-SVM can quickly and accurately calibrate the mesoscopic parameters of the discrete elements, indicating a good application value in practical engineering.