Verification and validation of models used in computer simulations of roadside barrier crashes
Submission Date: 2013
The Finite Element Method (FEM) is now regularly used by engineers to analyse the crashworthiness performance of roadside safety barriers. In particular, the improvements in non-linear Finite Element (FE) codes and the available access to supercomputing facilities have now allowed engineers to simulate in detail crash tests between vehicles and roadside safety barriers. Computer FEM simulations allow investigating the performance of new designs or retrofitted modifications to existing systems. However, it is essential that the numerical model is accurately verified and validated to provide reliable results. In particular, quantitative methods should be used to pursue an objective assessment of the level of Verification and Validation (V&V). The quantification of the V&V process is particularly important in the certification process for roadside hardware by regulatory authorities. This paper provides an overview of the guidelines for V&V of numerical models used for simulating roadside hardware crashes that were recently proposed under the National Cooperative Highway Research Program (NCHRP) 22-24 in the United States. After an initial description of the general concepts of the V&V process, the quantitative methods that objectively measure the level of validation of numerical models used to simulate the crash performance of roadside safety hardware, such as guardrails or concrete barriers are discussed. In particular, it is shown how the acceptance criteria were assessed for those selected validation metrics, based on an analysis of the typical scatter of results from a repeated series of identical or very similar vehicular full-scale crash tests. An example application for the proposed V&V procedures is also provided. Designers, policy decision makers and regulators will benefit from the use of the described V&V procedures, which provide a quantitative process and measurable level of the numerical model’s reliability.