Optimal Design And Reliability Analysis Of Vehicle Occupant Restraint System Based On Non-probabilistic Model

With the rapid development of my country’s automobile industry,the number of users of vehicles in my country is also increasing.In order to reduce the injury of the occupants and protect the safety of the occupants’ life and property,improving the safety of the occupants after a collision has become the focus of people’s increasing attention.In the event of a car crash,a good occupant restraint system can significantly reduce occupant injury and reduce mortality.When optimizing the design of the occupant restraint system,the sensitivity analysis of the occupant damage response is carried out,and the variables that have a greater impact on the damage response are selected as the optimization design variables,which is beneficial to improve the optimization efficiency.When optimizing the occupant restraint system through the real vehicle crash test,which disadvantages are long computation time,low optimization efficiency.Furthermore,no expressions were shown between occupant injury response and model parameters.Therefore,the method of combining surrogate model and optimization theory is generally used for research.Considering the uncertainty of the structure and material of the occupant restraint system,as well as the reliability of the optimal solution of the occupant restraint system,reliability analysis of the occupant restraint system are required.Therefore,the specific research contents of this paper for the optimal design and reliability analysis method of the vehicle occupant restraint system are as follows:(1)The sensitivity analysis of the occupant restraint system is carried out using the design of experiments method.First,the simulation model of the occupant restraint system is established by MADYMO software.The simulation model is mainly composed of the Hybrid Ⅲ 50th percentile dummy model,the vehicle body model and the three-point seat belt,and whether the occupant restraint system model is accurate and reliable.Then,the optimal Latin hypercube experimental design is used to obtain sample points,and the human injury response value is obtai ned through the simulation model.Finally,the sensitivity of each variable to the response to occupant damage is calculated by the method of experimental design,and the sensitivity percentage of each variable is calculated by applying the normalization method,and the variable that has a greater impact on the response to occupant damage is screened out by combining the linear weighting method.As an optimization design variable,it lays the foundation for the subsequent optimization design.(2)The occupant restraint system is optimized by using the surrogate model and optimization algorithm.First,the initial sample points are obtained based on the optimal Latin design method in the experimental design,and the simulation values corresponding to the initial sample points are obtained by MADYMO software.The approximate optimization problem of the occupant restraint system is constructed by using the surrogate model.Then,the local evaluation index is used to determine whether to resample the local area of the surrogate model,and the resampling interval is determined.The optimal Latin hypercube test is used to obtain local sample points;the genetic Latin hypercube is used for sampling to obtain global sample points.The local and global sample points are screened by the weighted Euclidean distance criterion,and the qualified sample points are added to the initial sample space to update the surrogate model.Finally,the potential optimal solution is searched through a generational mapping genetic algorithm.The research shows that this method can efficiently solve the optimal solution of the occupant restraint system and ensure the safety of the occupants.(3)Reliability analysis of occupant restraint system using convolutional neural network and evidence theory.First,using the optimal Latin hypercube experimental design to obtain sample points,and the corresponding simulation values are obtained by substituting the sample points into the MADYMO simulation software.Then,the extreme value analysis of the surrogate model of the occupant restraint system is performed to determine the category of each focal element,and the focal elements of the occupant restraint system are redefined to obtain the training set of the convolutional neural network.After the convolutional neural network is trained on the training set,it can efficiently and accurately classify the focal elements of the occupant restraint system,so that the confidence interval of the evidence theory can be quickly obtained.