Nonintrusive Stochastic Finite Elements for Crashworthiness with VPS/Pamcrash

Abstract

Crashworthiness analysis remains an important concern for the design of safety structures. In this context, uncertainties play an essential role in the response of a crash problem with non linear behavior. With this statement at hand, in this work it is presented a review of uncertainty quantification (UQ) techniques, with intrusive and non-intrusive approaches in stochastic finite element methods for crashworthiness. The well-known deterministic finite element solver VPS/Pamcrash is used to illustrate the currently available methods, developing a comparative analysis of these techniques in crashworthiness UQ. Finally, relevant non-intrusive methods are applied to analyze the behavior of a specific quantity of interest in a dynamic crash model.

Adaptive surrogates of crashworthiness models for multi-purpose engineering analyses accounting for uncertainty

“…In the literature different papers regarding kriging metamodeling [9,11]. Other surrogate modelling technique can be implemented as Polynomial Chaos or Separated Response Surface [10,11]. However, kriging shows a better performance for crash modelling [11].…”

“…In crashworthiness, having an accurate knowledge of the model allows to reduce the number of prototypes and, in consequence, their costs. Stochastic modeling for UQ has been used in the crash industry for decision making support [10,17]. The process to propagate the variability from the input to the output requires thousands of model evaluations.…”

“…In order to demonstrate the performance of the proposed approach, a benchmark crash problem is studied. The literature contains different works in crashworthiness UQ field [11,10,17], where different UQ approaches are shown implementing techniques as Monte Carlo, Polynomial Chaos, Quasi Monte Carlo, dimensionality reduction and surrogate modelling.…”

Nonintrusive Uncertainty Quantification for automotive crash problems with VPS/Pamcrash

“…The number of samples n s affordable in a real problem is generally not sufficient to produce a proper Monte Carlo assessment of the statistical properties of the output of the system. A review of nonintrusive UQ methodologies for crashworthiness, see [21], demonstrates that the standard Monte Carlo sampling is extremely demanding and, in practice, beyond the possibilities of standard industrial practitioners.…”