Model-driven Design Framework And Methods For Automobile Passive Safety Systems

The capability improvement of digital design in engineering is critical for the transformation of national industrial structure,while model-based design technologies are the basis of digital design.In the past decades,experience-based design is the main method to achieve product development,which is especially true for vehicle passive safety system.Recently,experience-based design gradually gives its priority to computational model based design.Computer Aided Engineering(CAE)based design has get its popularity in nearly all the industrial areas.By using model to serve as the surrogate of physical prototype for designing,testing,and evaluating system performances,the experimental cost for product development has been considerably reduced.However,form the perspective of the digitalization of design optimization,directly employing CAE models has the defections regarding design validity,process flexibility,and information utilization.CAE model is merely the carrier of knowledge and information in specific area.The mining,propagation,integration and utilization of knowledge carried by model is urgently demanded by digital design vision.This thesis focuses on the development of a model driven digital design optimization framework and corresponding methods for complex engineering systems based on the design logic of “Model-Data-Knowledge”,which includes CAE model evaluation,CAE model decision making,metamodel updating,and design domain identification.The topics of this thesis include the following ones:(1)“Model-Data-Knowledge” mode based design framework is established through the analysis on the requirements and key issues of passive safety systems design problem.Two sub-modes,say model validation and design optimization,are included in the proposed framework,while the methods for CAE mdoel evaluation,CAE model decision-making,metamodel updating,and design domain recognition are to be studied.Reliable model validation is the basis of accurate design,while accurate design is the feedback of model validation process.These two modes supplement each other to form the cornerstone of the proposed methodology.(2)Time-frequency transformation based model validation metric is developed for CAE model evaluation through considering the responses characteristics and ideal validation metric.Comprehensive evaluation for model error/fidelity can be achieved based on the description,non-parametric reconstruction,and features decoupling for dynamic responses.(3)Statistcal decision making method is proposed for multivariate nonlinear systems based on dimension reduction and hypothesis testing techniques,considering the characteristic of nonlinearly correlated multivariate responses.Hierarchical dimensionality recution is achieved based on Kernel-Princial Component Analysis(KPCA)and Bayesian classification,exploring the mechanism of qualitative and quantitative model decision making.(4)Metamodel updating strategy based on model selection and bias correction techniques is proposed for the improvement in predictive accuracy of metamodels involved in engineering optimization.Reliability and optmimality under uncertainty can be achieved based on similarity theory and Bayesian extrapolation technique,which are included in the model updating strategy.(5)Design domain identification strategy is proposed based on root cause analysis for the purpose of optimization efficiency improvement.The methods and mechanisms of feasible/robust domain classification and identification are explored based on decision tree theory and Probabilistic Principal Component Analysis(PPCA).Optimization efficiency and effectiveness can be improved through applying the proposed hierarchical strategy in large-scale engineering problems.Oriented by digital design vision in engineering and driven by computational engineering models,the model-driven design framework and corresponding methods proposed in this thesis are applied to address the outstanding issues in vehicle passive safety system.By taking advantages of the procedures such as CAE models ultalization,deep data mining based knowledge and information acquisison and numerical model construction,the proposed methods are able to provide insights for the development of model-based digital design architechture in industry.