柔性机构动态可靠性分析的SVM回归极值法

SVM regression extremum method of dynamic reliability analysis in flexible mechanism

为改善柔性机构动态可靠性分析的效率和精度,基于支持向量机(Support Vector Machine,SVM)回归理论,提出了一种柔性机构动态可靠性分析高效率高精度的SVM回归极值法(SVM Regression Extremum Method,SREM)。首先,介绍了柔性机构可靠性分析的基本理论;其次,结合蒙特卡罗法(Monte Carlo,MC)和SVM回归理论,建立了柔性机构动态响应极值的代理模型,并利用代理模型进行了柔性机构的可靠性分析;最后,以柔性夹紧机构的可靠性分析为例,利用SREM加以验证。结果表明:SREM的计算时间约为蒙特卡罗法的20%,远远少于蒙特卡罗法;SREM的计算精度几乎与蒙特卡罗法保持一致,当可靠度大于98%时,SVM回归极值法的计算精度与蒙特卡罗的计算精度完全一致。

In order to effectively improve the efficiency and accuracy of dynamic reliability analysis in the flexible mechanism, based on the Support Vector Machine(SVM) regression theory, an SVM Regression Extremum Method(SREM) with high-efficiency and high-precision of dynamic reliability analysis in flexible mechanism is proposed. Firstly, the basic reliability analysis theory in FM is introduced. Secondly, the combination of Monte Carlo(MC) method and SVM regression theory is applied in FM, and the surrogate model of dynamic response extremum in FM is established. By using the surrogate model, dynamic reliability analysis in FM can be effectively implemented. Finally, SREM is verified by the dynamic reliability analysis of clamping FM. Results obtained show that the calculation time of SREM is far less than that of MC, which is about 20% of MC. The computational accuracy of SREM is almost consistent with that of MC method. When reliability is greater than 98%, the computational accuracy of SREM is equal to that of MC.