Characteristic extraction of machined surface using wavelet transformation

In the development and design of machinery products, product performance and tolerance stack-up analysis are generally calculated using computer software. In these calculations, the product shape is modeled as a set of primitives, which are simple geometrical shapes, such as plane, cylinder and sphere. However, the shape of an actual product is different from these shapes because machining error cannot be avoided. Actual shapes usually exhibit deviations in scales, such as geometric deviations, waviness and surface roughness. Therefore, the error between the result of computer calculation and the actual behavior of the product is caused by the shape simplification. A method of for generation of a pseudo shape is required to reduce the calculation error. In this study, a new method for random generation of skin model shapes, which maintain the surface characteristics in multiscale levels, is proposed. In the method, first, an optimal mother wavelet should be decided because they depend on extracting the surface characteristics. In this paper, eight representative mother wavelets are used in wavelet transformation for comparison. The results are discussed based on the results obtained through a case study, and latent problems are also observed.