| In the field of modern automotive industry,the demand of product design and development for simulation rise increasingly.At the same time,the simulation time and calculation amounts are rapidly increasing with the improved model complexity.Under the premise of higher efficiency,how to simplify the models and evaluate the error of model simplification efficiently have become a research hotspot in the field of automobile industry.Through simplification,unnecessary calculation will reduce effectively,simulation efficiency will improve and the occupation of computing resources will decrease.However,with the improvement of model simplification,the accuracy of model simulation will inevitably decrease gradually.Therefore,how to effectively and quantitatively predict the errors caused by model simplification is of great significance to guide product development and design.In this paper,the following four aspects are researched about the above contents:(1)Model simplification and error evaluation method based on feature suppression.Based on the empirical rules of finite element model simplification and the related algorithms of model validation and validation,the method of model simplification and error evaluation for feature suppression is studied,and the quantitative evaluation index of model simplification error is established.At the same time,according to the requirement of large amount of training data in the development of prediction model in Chapter IV,the parametric modeling method for feature size data is studied,which provides a methodological basis for the training data establishment of subsequent prediction model.(2)Simplified feature extraction and recognition method of finite element model based on point cloud.In this paper,a simplified feature extraction and recognition method based on point cloud model is proposed,and the realization process of this method is demonstrated by taking a certain type of sheet metal parts as an example.This method includes three main steps: feature extraction,feature judgment and feature geometric information digitalization.It mainly solves the incompatibility between geometric model features and input data of neural network.It digitalizes geometric information of simplified feature of model and achieves the purpose of effectively reading geometric feature information of prediction model.(3)Simplified error prediction model based on machine learning model features.Based on the prediction model algorithm of machine learning,a simplified model error prediction method for feature suppression is proposed in this paper.During the development of the prediction model,taking the collision simulation of the front bumper beam model as an example,the training data of the prediction model,the training process of the prediction model and the accuracy verification process of the prediction model are demonstrated.The results show that the method can predict and evaluate the error of model simplification in vehicle collision simulation.(4)Based on the simplified error prediction method proposed in Chapter 4,and combined with the above three parts,a prototype system of error prediction is developed,and the feasibility of the prototype system is verified by substituting it into a more complex simulation model in the field of vehicle safety research.Focusing on the prediction and evaluation of model simplification error in vehicle safety simulation,this paper presents a set of perfect model feature simplification error evaluation algorithm,and develops a software platform for model simplification error evaluation based on the proposed method.According to this method,engineers can efficiently judge and analyze what geometric features can be simplified and the errors caused by simplification or suppression in product development and design stage.The method proposed in this paper is an effective solution to the problem of model simplification error evaluation. |
