A study of neural network for surface characteristics in-process optical measurement

Abstract

The opaque property of the coolants would produce an inaccessible problem for in-process optical measurement of workpiece surface quality in machining process. A transparent window method can be used to produce an optical clean zone on the surface to solve this problem. The surface scattered images of specimens in the clean zone are captured and some parameters are extracted from the images in this study. Artificial neural network is developed to determine surface roughness by selecting the back-propagation algorithm. Flow rate of the transparent fluid, the thickness of the fluid layer, scattering feature along the main direction of the scattering stripe, standard deviation perpendicular to the main direction of the scattering stripe and gray feature of the scattered image are chosen as input values, and surface roughness is selected as output value of the neural network. Experimental results show that the proposed method is efficient and effective for in-process optical measurement of surface characteristics in terms of training and test accuracy of surface roughness.