Descripció del projecte
Computational models are widely used in industry as a prediction tool for product development. They aim at simulating the actual physical behavior of the devices under the expected input actions and predict the critical response. They are classically used assuming that the input of the model (material properties, geometry, loads…) is deterministic, and the analysis is performed by carrying out a parametric study. In order to properly account for the uncertainty of the output, the input data is stochastic (the input parameters are random variables). Solving this type of problems requires using technologies like Stochastic Finite Element Methods (SFEM), where in addition to the error coming from the Finite Element discretization (e.g. the coarseness of the finite element mesh) we have to account with the error arising from the stochastic approximation (e.g. the reduced number of Monte Carlo realizations).
The aim of the project is introducing SFEM in the SEAT portfolio of methodologies in the framework of complex automotive crashworthiness computational models in VPS (software of ESI Group). This includes accounting for uncertainty in the data coming from different sources (material properties variation, metal thickness non uniformity due to the manufacturing process, tolerance existing regarding barriers points of impact…) and assessing the error introduced by the numerical approximations.
Thus, the two specific objectives of the project are:
(i) Developing and implementing SFEM methodologies with VPS, to be used in SEAT projects
Analyze, assess and control the dispersion of the numerical results and correlate them with the experimental observations (understanding also the dispersion obtained from identical models ran several times).
