Research And Application Of Structural Reliability And Sensitivity Analysis Method Based On Active Learning Kriging Model

Structural reliability analysis is used to measure the risk level of the structure and obtain the failure situation of the structure.Sensitivity analysis is used to obtain the influence of uncertain input variables or their distribution parameters on the uncertainty of output response.In order to improve the accuracy and efficiency of structural reliability and sensitivity analysis,this paper studies structural reliability analysis,local sensitivity analysis and global sensitivity analysis.The main research work is as follows:(1)Some strategies for updating Kriging model iteratively with learning function are pointed out,and a new learning function EU is constructed.Combined with LHS,MCS and other methods,a structural reliability analysis method based on active learning Kriging model is proposed.Some numerical and practical engineering application examples are selected and compared with various methods.The results show that the proposed method improves the accuracy and efficiency of structural reliability analysis in various examples.(2)The truncated importance sampling method is introduced in the local sensitivity analysis.By combining this method with Kriging model and learning function EU,a local sensitivity analysis method based on active learning Kriging model is proposed.This method takes into account the number of calls of both functional functions and proxy models.The importance ranking of distribution parameters of uncertain input variables is obtained,which further improves the analysis efficiency while satisfying the analysis accuracy.The superiority of this method is verified by numerical and engineering application examples.(3)For the two common global sensitivity indexes,the Kriging model updated by learning function EU is used to simulate the sample point response values of the singlelayer MCS method,so as to avoid the huge computation of the single-layer MCS method,and then obtain the global sensitivity indexes based on variance.The Kriging model updated by learning function EU is used to calculate the unconditional and conditional failure probabilities,so as to improve the solving efficiency of the global sensitivity index based on failure probabilities.Finally,the importance ranking of variables corresponding to the two indicators is obtained.The accuracy requirement and efficiency improvement of the two indexes are verified by examples.(4)In order to further verify the analysis method and put the method into practical application,a six degrees of freedom robot is used as an example to analyze the reliability and sensitivity of the positioning accuracy when the input kinematic parameters are uncertain.The reliability degree of the positioning accuracy of the robot and the uncertain kinematic parameters which have a serious influence on the positioning accuracy are obtained.Through comparative analysis,the proposed method not only meets the analysis accuracy,but also greatly improves the analysis efficiency,which still has certain advantages in robot analysis.