Research On Reliability Method Under Uncertainty And Its Application In Automotive Structure Design

With the development of industrial technology and the increasingly complicated structure,the reliability and safety issues of automobile structures have become more and more important in the design of engineering structures.Due to the extremely harsh environment carried by the actual engineering structure,various uncertainties generally exist in the design parameter,which makes the structure frequently fail and causes catastrophic accidents.Good reliability design can avoid these accidents and guarantee the performance characteristics of the structural design.Structural reliability analysis and optimization design is a method to solve the optimal design variables with structural reliability as the objective function or constraint condition based on uncertain variables.Compared with the traditional structural design method,this method can obtain better structural characteristics and economic benefits,which has important research value and engineering significance.The high simulation cost makes efficient and accurate reliability analysis and design optimization an urgent problem to be solved.Especially in the case of multi-dimensional variables and non-linearity,there are still deficiencies in efficiency and accuracy when conducting reliability research.This paper systematically studies the structural reliability analysis and its optimization design methods,as well as the topology optimization,and proposes a data-driven automotive reliability design process framework,the probabilistic reliability analysis model of the structural system based on the improved downhill simplex algorithm interpolation response surface method(RSM),the reliability optimization design method of the approximate model based on local validity constraints,the non-probabilistic reliability topology optimization model based on the Chebyshev polynomial,the application of reliability-optimized design model based on probabilistic-non-probabilistic mixed variables in automobile structure.The main research contents and conclusions are as follows:1)For the lack of reliability design of the passive safety system for automobile structure,a data-driven reliability optimization design process of automobile structure parameter uncertainty is proposed.Based on the current reliability theoretical knowledge of structural system,the basic theories of uncertain probability models and non-probabilistic models are analyzed,and the basic rules of random and interval operations are sorted out.To combine the actual engineering applications of automobiles,a reliability analysis method,a reliability optimization design method,and a proxy model reliability optimization design method combined with finite element analysis are given.The deficiencies of the current reliability analysis and optimization design are summarized by description.An analysis method and an optimized design process for considering the uncertainty of the automobile structure are introduced,and the proposed design process is discussed and analyzed in depth to provide a reliable design process framework for the actual project.2)The hybrid method combining adaptive moving experimental point strategy and RSM is proposed,and a new RSM is described using the downhill simplex algorithm to efficiently evaluate structural reliability.Due to the high computational efficiency,the RSM has been widely used in structural reliability analysis.However,the RSF consisting of the initial experimental points is rarely fully suitable for the limit state function,which leads to incorrect design points.And for the highly nonlinear limit state function,the approximate accuracy of the reliability failure probability mainly depends on the design points,which leads to errors in the traditional RSM reliability evaluation.Based on the balance between efficiency and accuracy,this paper proposes improvement strategies.The principle is mainly to reconstruct the RSM approximate model by changing the search strategy of the basic downhill simplex algorithm,and combining the direct search advantage of the improved downhill simplex algorithm with the response surface reliability analysis mechanism.The analysis of examples shows that the modified RSM global optimization algorithm has good convergence ability and high calculation accuracy.3)Aiming at the accuracy problems of the approximate model in structural reliability optimization design,the local effective constraint reliability optimization design method based on proxy model is proposed.Due to the low-cost calculation process,the agent model method has been widely applied in structural reliability optimization design considering uncertain conditions.But the main limitation of this method is that it is difficult to quantify the errors caused by the approximation of the meta-model,which leads to inaccurate optimization results of reliability evaluation.This paper analyzes the root cause of the uncertainty in the output response,aims to improve the uncertainty performance in the project,identifies the important areas and feasibility areas and introduces the idea of local efficiency to solve the problem of reliability optimization design.To improve the efficiency of the algorithm,the constrained effective set is introduced in the search iteration process to avoid repeated calls in the design optimization process when the probability constraints appear invalid.Finally,the car example is given to demonstrate that the proposed method can effectively solve engineering application problems.4)Aiming at the uncertainties of structural topology optimization design variables,the topology optimization design method for continuum structures with uncertain but bounded parameters is proposed based on the non-probabilistic reliability quantified structural parameter changes of the multi-ellipsoid convex models.Considering that the material properties and loading forces are uncertain,the interval Chebyshev zero polynomial is used to approximate the true LSF of the normalized random variable.The single-loop reliability algorithm is used to calculate the optimal design point value under the corresponding target reliability index,so that the non-probabilistic reliability optimization problem can be transformed into a deterministic optimization problem.The results of calculation examples display that compared with the deterministic structural topology optimization design,the reliability topology optimization considering random variables can obtain a more reliable topology.5)Aiming at the problems of the lightweight design of automobile structure collision safety system,the optimization design method for automobile collision safety-lightweight reliability under the condition of random-non-probabilistic mixed variables is proposed.Uncertain parameters generally exist in the design of automobile structures.The collision safety and reliability of automobile structures are one of the important links in automobile safety design.Therefore,the structural design considering uncertainty is an inevitability of automobile performance design research.When the variables have some known distribution characteristics and some incomplete information,it is necessary to consider the uncertainty model of the mixed variable mode.Through Chebyshev sampling distribution,uniform sampling is introduced to optimize the sample points.Based on the simulation model analysis of highly non-linear automobile collisions,the important parts of the vehicle body is taken as the research object for the lightweight design to obtain an optimized design scheme with good crash safety performance under the constraints of reliability.The final results show that the total quality of the optimized key components is reduced by 13.44%,and the reliability constraints also meet the given reliability index,achieving a balanced design of collision safety reliability and lightweight structure.