基于模糊支持向量机的岩石抗剪强度参数确定方法研究

摘要

针对岩石抗剪强度参数具有随机性与模糊性的特点，提出一种基于试验样本到超平面距离作为隶属度函数的模糊支持向量回归方法，通过更加合理地设计隶属度函数，提高支持向量机的稳健能力，得到最优的参数估计。在此基础上，对岩石抗剪强度参数进行建模仿真，并进行估计。结果表明，基于模糊支持向量回归的方法与最小二乘法、随机－模糊法相比，不仅能减少异常试验数据对岩石抗剪强度参数的影响，而且得到的回归方程更符合试验样本整体分布，可反映岩石力学参数固有属性。%For the randomness and fuzziness of determining shear strength parameters for rock,support vector regression theory was proposed based on membership function of the distance from test sample to hyper-plane. By designing membership functions rationally,improving robust ability of SVM,optimal parameter estimation could be obtained. A model of shear strength parameters for rock has been built and the parameters were estimated based on that. The results showed that,compared with the least squares and random method,fuzzy support vector regression not only could reduce the impact of abnormal shear test data on the parameters,but also could achieve regression equation,which was further conform to the overall distribution of the test sample and reflect intrinsic properties of rock mechanics parameters.