摘要
在柔性机构(Flexibility Mechanism,FM)动态可靠性分析中,为了提高其计算精度和计算效率,通过融合蒙特卡洛和支持向量机回归理论,提出了一种新的SVM回归极值法(SVM Regression Extremum Method,SREM)。该方法借助ADAMS软件抽取FM动态响应极值的小样本,基于支持向量机回归理论建立FM动态响应极值的代理模型,使用此代理模型进行FM动态响应可靠性分析。最后,利用蒙特卡洛法、SVM回归极值法和另外两种方法对柔性曲柄摇杆机构的摇杆最大摆角可靠度进行分析。结果显示:在小样本情况下,SVM回归极值法的计算精度与MC相当,精度明显高于另外两种方法;SVM回归极值法的计算效率比MC大幅度提高,与另外两种方法计算效率相当。验证了在小样本情况下,SREM在FM动态可靠性分析中高效率和高精度。
In order to improve computational accuracy and efficiency in Flexibility Mechanism(FM) dynamic reliability analysis, a novel SVM Regression Extremum Method(SREM) is proposed based on the combination of Monte Carlo method(MC) and Support Vector Machine(SVM) regression theory. By means of ADAMS software, a small amount of samples for FM dynamic response extremum were generated. Based on these samples,the surrogate model of FM dynamic response extremum was established by SVM regression theory. Reliability of FM dynamic response can be assessed by the surrogate model. Finally, the reliability of rocker maximum angle in crank-rocker FM was analyzed by MC, SREM and other two methods. The results show that, in the case of small samples, the computational precision of SREM is almost the same as that of MC, and is obviously better than that of other two methods; the computational efficiency of SREM is significantly higher than that of MC, and is slightly higher than that of other two methods. It is proved that SREM is of high-precision and high-efficiency in FM dynamic reliability analysis in the case of a small amount of samples.
出处
《工程力学》
EI
CSCD
北大核心
2014年第12期208-216,共9页
Engineering Mechanics
基金
北京市自然科学基金重点项目(3102019)
国家自然科学基金项目(51175017)
关键词
柔性机构
支持向量机回归
蒙特卡洛法
动态可靠性
SVM回归极值法
flexible mechanism
support vector machine regression
Monte Carlo method
dynamic reliability
SVM Regression Extremum Method(SREM)
