Definition of restraint components for vehicle occupant safety : Using Design of Experiments

Abstract: The Master thesis has been carried out at Autoliv B.V. & Co. KG in Dachau, which is a global automotive safety corporation. The objective of the thesis was to define component characteristics, which would reduce the load on a 5th percentile female dummy in a full frontal crash to achieve a five star in the new Euro NCAP full frontal crash test. Euro NCAP (voluntary vehicle safety rating system) tests new vehicles launched on the European market with the goal to encourage manufacturers to aim for more than the legislative minimum requirements. The safety performance of the vehicles is evaluated and rated in a star rating system where a maximum of five stars can be awarded depending on the impact on the dummy. The method Design of Experiments was used in the thesis and determined as a successful method to find an optimal component setting. The system with the 5th percentile female dummy was decided to be investigated in total nine parameters with different settings and eleven measured responses. The D-optimal design was considered as the most suitable design for the experiments since it was handling input parameters with different levels on the factors. Moreover, it was significantly reducing the number of required experiments. The D-optimal design was considered as suitable for optimization of parameters, however it was found as preferable to first execute a simple screening of the parameters to ensure that an appropriate range of the parameters had been selected. In this case a screening with Plackett Burman would have required 80% less experiments to investigate the selected range of the parameters. The two software programs, Minitab and Datalysor, were used to analyze the Design of Experiments. In the analysis, the main parameters, interactions and higher terms were identified, which significantly influenced the responses. Furthermore, regressions were evaluated and the programs strengths and weaknesses were compared. It was shown that the regressions made in Datalysor were more accurate and a better model of the reality. The reason for the better regression was because Datalysor had the possibility to remove different outliers and predictors for every regression, which was not possible in Minitab. From the regression model an optimal setting was found which fulfilled the requirements from Euro NCAP and five stars could be achieved.