In order to reduce automobile body weight and improve the crashworthiness of passenger cars, the use of high-strength steels is greatly increasing in years past. However, the high-strength steels generally become less ductile. As the results, the fracture of steel material and spot weld failure may occur due to local deformation at corner of automobile frame. Therefore, the accuracy improvement of local deformation to predict the fracture is highly demanded by automotive industries. However, current shell and solid elements of conventional FEM generally use bilinear shape functions. It was found that bilinear shape function was difficult to predict a complicated local deformation with high accuracy. To resolve this issue, Isogeometric analysis was considered as a better alternative. Isogeometric directly deals with 3D CAD constructed by NURBS and calculates the NURBS model using high order shape function elements. In this study, mainly 16 nodes shell and 64 nodes solid element with third order shape function were applied.

First of all, the local deformations of typical crash verification models were investigated, which were S-frame model with hat-section and 3-point bending model constructed by high-strength steels. In case of conventional FEM, local deformation of S-frame and 3-point bending model showed the triangle shapes at corner region of hat-section, while local deformation of Isogeometric results showed the elliptical or hyperbola shapes. And as to fracture location of 3-point bending model, it was found that Isogeometric results correlated well with the experimental results.　By contrast, fracture location of conventional FEM was not at corner region of hat-section.

Next, IGA preprocessor was developed to create the NURBS data. It has two types of NURBS data creation functions, which were named as inner trimming function and outer trimming function. A Body in White model was constructed using two types of functions. With regard to the creation time, the new functions were able to reduce the creation time by about 50% compared to FEM model.