Crash Simulation Methods For Vehicle Development At Nissan

Author : Pierre Phalippou
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (124 download)

DOWNLOAD NOW!

Download or read book Data-enabled, Reduced-order Simulation of Dynamical Systems with Application to Vehicle Crash Simulation written by Pierre Phalippou and published by . This book was released on 2020 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Vehicle manufacturers face numerous constraints regarding security, comfort, weight, and fuel consumption when designing new automotive vehicles. Numerical simulation, and more precisely, the finite element method, is extensively used in the process to achieve cost reduction while allowing to test the impact of multiple design parameters on the system behavior. Nevertheless, industrial finite element models of automotive crash are growing prohibitively expensive in computation time, slowing down the design process. Numerous model order reduction methods have been developed in order to speed-up simulations by making use of the humongous amount of collected data and the repetitiveness of computations with slight changes to the design variables. This thesis aims to apply model order reduction methods to finite element simulations of automotive crash, involving material and geometric nonlinearity as well as contact. This project, conducted in close collaboration between the car manufacturer PSA group, the software developer Altair engineering, and the computational mechanics group at the UTC-CNRS joint laboratory Roberval, targets intrusive methods necessitating modifications in the finite element solver source code. Amongst different promising methods reviewed in the state-of-the-art, our work focuses on projection-based reduced order modeling and hyper-reduction. A reduced-basis of global shape functions is built upon relevant data in the online training phase and used online to reduce the model size, enabling the hyper-reduction of internal forces and computational speed-up. The contributions of this thesis concern three aspects of the training phase. Developments of the incremental singular value decomposition allow for a more flexible and faster reduced-basis training. The proposed sparse criterion enables the formation of basis functions subsets with enhanced performances and better correlation between the online training phase approximation error and the actual error in the online reduction phase. An innovative formulation of the hyper-reduction optimization problem involving mixed-integer programming, added constraints on polynomial integration, and volume conservation is suggested for the benchmarking of future heuristic methods. All implementations are performed in the industrial explicit finite element solver Altair Radioss with complete access to the source code. To the author's best knowledge, the implementation of such a method in industrial explicit finite element solver represents a significant differentiator of the present contribution. Implementation specific difficulties relevant to this aspect of the thesis are discussed. Finally, recommendations, as well as perspectives and further developments required for the robust application of the method to industrial numerical simulations of vehicle crash, are given.