Search Results

Finite Element Analysis (FEA) models are routinely being adopted as a means of up-front design for automotive body structure design. FEA models play two important functions: first as a means of assessing design versus an absolute target; secondly they are used to assess the performance of design alternatives required to meet targets. Means of assessing model capability versus task is required to feed appropriate information into the design process. Being able to document model capability improves the credibility of the FEA model information. A prior paper addressed assessing the absolute performance of model technology using a metric based on a statistical hypotheses test that determines membership in a reference set. This paper extends the use of quality technology to determining the capability of the FEA model to span the design space using Designed Experiments.

The problem of NVH CAE model correlation in light of test and product variation has been addressed. An objective metric based on statistical hypothesis testing has been proposed and evaluated. This technique has been shown to work for frequency response functions. The hypothesis test answers the question ‘Are the involved frequency response functions statistically different than those in a reference set?’ This paper demonstrates that vehicles are uniquely identifiable by their frequency response functions. Under certain restrictive assumptions, the average gross error normalized by the ensemble variance is chi-squared distributed. Using a chi-squared test, the probability that a NVH CAE prediction is a member of a reference (test) set can be estimated. Within the context of a reference (test) set, this metric represents the limit to predictability. The metric was applied to examples including two midsize car NVH CAE models.