Abstract: In view of the issues of low efficiency and high cost in creating hexahedral mesh models from industrial CT images, a method to directly generate hexahedral meshes from industrial CT images is proposed based on grid-based method. Octree is utilized to divide initial meshes from the industrial CT images and extract the core mesh; A surface fitting method is proposed to generate the boundary elements based on the difference of the industrial CT images boundary gray scales; Laplacian smoothing is used to optimize the quality of boundary elements; Experiments are carried out using industrial CT images of gear, support and bracket. The results indicate that the Jacobi ratio of generated hexahedral meshes ranges from 1 to 4, 95% of the units have a Jacobi ratio between 1 and 2, and the method can meet the requirements of finite element analysis. This method can directly generate hexahedral mesh models from industrial CT images of workpieces, and is more efficient compared with traditional reverse engineering methods because there is no need to rebuild the CAD model.