Research On Credibility Evaluation Criterion Of Structural Numerical Model Based On Non-probability Measurement

For the actual engineering structure,the test of complex physical models usually has the difficulty of strict test conditions,high cost,long cycle,etc.And even in some particular cases,it is difficult to test.In recent decades,the rapid development of computer technology has promoted the fast improvement of computational efficiency.Therefore,numerical simulation technology based on computer technology has been widely used in various fields,especially in the field of machinery.The numerical model based on numerical simulation technology is only an approximate model of the physical model,which is inevitably different from the physical model,thus it leads to the fact that the experimental results based on the numerical model and the physical model can't be completely consistent.Therefore,the study on measurement for qualitative and quantitative evaluation of the credibility of numerical models has become a hot research topic.Aiming at the difficulty in the credibility evaluation of single-output non-probabilistic numerical model of mechanical structure,the dissertation based on the non-probability metric method,achieves the transition of non-probability model response from non-probability to pseudo-probability,and then uses the area criterion and u-pooling criterion to evaluate the credibility of the single-output non-probability model.The following studies are carried out in this dissertation:(1)Based on the pseudo-probability metric,the credibility evaluation criteria applicable to the single-parameter single-output non-probability model are studied.For the single-parameter single-output non-probability model,it is assumed that the model response has a monotonic function relationship with the model parameters and one-to-one correspondence.When the model response equation divides the uncertainty parameter into two parts,the ratio of the length of the parameter segmentation point to the lower bound of the parameter to the length of the parameter uncertainty range is defined as pseudo probability.At the single verification point,the local credibility evaluation of the single-output non-probabilistic model can be achieved by using the area criterion in the model credibility evaluation.At the multi-verification point,the global credibility evaluation of the single-output non-probabilistic model can be achieved by using the u-pooling criterion in the model credibility evaluation.(2)Based on the interval convex model to measure the uncertainty parameters,the credibility evaluation method for multi-parameter single-output non-probability model with interval uncertainty is studied.Firstly,the interval convex model is modeled for multiple uncertain parameters without correlation.For the monotonic model,the model response equation divides the uncertainty domain of the parameter into two regions.Then,using the space conversion technology to transform the original uncertainty domain into the standard space,and using the first-order Taylor expansion of the model response function to obtain the linear model response equation instead of the original function,the uncertainty domain of the parameter is cut by the linear model response equation.The area ratio or volume ratio of the segmentation region and the parameter uncertainty region is solved based on the triangle similarity principle.Finally,calculating the area ratio or volume ratio,solving the pseudo-probability of the non-probabilistic model response,and realizing the credibility evaluation of the interval multi-parameter non-probability model.(3)Based on the ellipsoidal convex model to measure the uncertainty parameters,the credibility evaluation method for multi-parameter single-output non-probability model with ellipsoid uncertainty is studied.Firstly,the ellipsoidal convex model is modeled for multiple uncertain parameters with correlation.For the monotonic model,the uncertainty domain of the parameters will be divided into two regions by the model response equation.Then the space transformation technique is used to transform the original uncertainty domain into the standard sphere space,and then the first-order Taylor extension of the model response function is used instead of the original function.The area or volume of the arcuate region and the standard sphere is calculated by linear model response equation cutting uncertainty domain of the parameter.Finally calculating the area ratio or volume ratio of the arcuate region to the standard sphere domain,solving the pseudo-probability of the non-probabilistic model response,and realizing the credibility evaluation of the ellipsoid multi-parameter non-probability model.