Correlation-based feature extraction from computer-aided design, case study on curtain airbags design

• Correlation-based feature extraction is applied to CAD model parameters. It shown that parameters as input, makes the problem high dimensional. The volume simulation of a curtain airbag used case study. Using area, medial axis, etc., concept ‘Sleeping Parameters’ introduced. Sleeping enables real time prediction tool in early design phases. Many high-level technical products are associated with changing requirements, drastic changes, lack information, and uncertainties input variables which their process iterative simulation-driven. Regression models have been proven be useful tools during design, altering resource-intensive finite element models. However, building regression from computer-aided (CAD) challenges such dealing too many low or coupled impact on studied outputs ultimately requires large training dataset. As solution, hidden features presented application airbags concerning geometric changes loops. After creating prototype covers all aspects airbag, its analyzed find out correlation between output. Next, using experiment latin hypercube sampling method, 100 samples generated corresponding for each sample was assessed. selected not highly correlated consequently lowers accuracy Various entities, extract several (referred sleeping ). new performance precision through two analyses studied. result shows choosing reduces dimensionality need use advanced algorithms, allowing designers more accurate predictions (in this approximately 95%) reasonable number samples. Furthermore, it concluded regression-based creates real-time ability development stage could contribute lower lead by eliminating iterations.