Research And Application Of Numerical Analysis Grid For Vehicle Crashworthiness

Compared with traditional crash tests, numerical analysis methods become the principal means for the vehicle crashworthiness research because of its accurate calculations result and efficient implementation. However, with the trend of globalization, automobile industry gradually transplanted to the network environment. Transnational and trans-regional automotive manufacturers and suppliers collaborate to develop products which is helpful to improve product quality and enhance competitiveness of their products. So, the numerical analysis based crashworthiness researches which carry out at local site must adapt to the network manufacturing environment.This paper researches and builds the grid platform for vehicle crashworthiness numerical analysis based on distributed computing and numerical analysis theory, and also study the crashworthiness collaborative simulation and distributed data-mining method based on the crashworthiness theory. Based on the research above, this paper builds two application systems including the application system for vehicle crashworthiness collaborative simulation and the application system for vehicle crashworthiness data-mining, and two typical applications which are business practice related were implemented on these two systems separately.According to the need of the vehicle crashworthiness numerical analysis, this paper study the grid system architecture, security policy, human resource related workflow, as well as calling and notification method of the human services. Based on the research above, this paper constructs a grid platform comply with OGSA. This paper presents a concept and technique about grid application middleware for vehicle crashworthiness numerical analysis which enables the decoupling and separation between the application services and the basic grid services, and also enhances the flexibility and scalability of the system development. This proposed grid platform can be applied not only to vehicle crashworthiness numerical analysis, but also applies to other areas of similar works.Vehicle crashworthiness collaborative simulation is an important part of the vehicle crashworthiness numerical analysis. This paper researches the crashworthiness algorithm which is based on the explicit nonlinear finite element method and contact algorithm, and collaborative design method for finite element analysis based on information safety. Presents the combine method for finite element models based on information safety, and the separation method of the simulation results, based on which, developed the application services for crashworthiness collaborative simulation. Based on the research mentioned above, this paper build the application system for vehicle crashworthiness collaborative simulation by deploying the services that we developed to the application grid middleware, and developing the Web pages for the end-user in the application layer of the plat form based on the technique of Jsp, Servlet, workflow, et al. as well as business logic. This system enables car makers and part suppliers to share resources and to protect their own secret information about vehicle during the collaboration. In order to assess the performance of the system, a test bed has been established in Networks (Shanghai Automotive Engineering Research Institute, Yanfeng Johnson, the Shanghai Supercomputer Center and the Shanghai Jiaotong University), and a real business case was study based on this test bed.Data-mining for crashworthiness data is another important part of vehicle crashworthiness numerical analysis which aims to bring technique support for the vehicle crashworthiness structural design, especially the optimal design. This paper studied the meta-data computing technology, combining of the decision tree classifier based on data mining and vehicle crashworthiness theory, and proposed the distributed data-mining algorithm for crashworthiness data based on Stacking algorithm, Meta-learning algorithm and reverse strategies. This paper also proposed a distributed data-mining service which consisted of front-end services in charged of the whole project management and back-end services in charged of the calculation of meta-data and the local data-mining of the data stored in local host. Based on the research mentioned above, this paper built the application system for vehicle crashworthiness data-mining by deploying the services that we developed to the application grid middleware, and developing the Web pages for the end-user in the application layer of the plat form based on the technique of Jsp, Servlet, workflow, et al. as well as business logic. In order to assess the performance of the system, a test bed has been established in LAN, and a case about the data-mining to 183GB data stored in four nodes was studied to realize the vehicle optimal design based on this test bed.