An optical method and neural network for surface roughness measurement

Abstract

The measurement of surface roughness using stylus equipment has several disadvantages. A non-contact optical method is needed for measuring the surface roughness of engineering metals with improved accuracy. One candidate for an optical method is the use of a laser source, where the laser light intensity reflected from the surface represents the surface roughness of the illuminated area. A relation can be developed between the reflected laser beam intensity and the surface roughness of the metal. The present study examines the measurement of the surface roughness of the stainless steel samples using a He-Ne laser beam. In the measurement a Gaussian curve parameter of a Gaussian function approximating the peak of the reflected intensity is measured with a fast response photodetector. In order to achieve this, an experimental setup is designed and built. In the experimental apparatus, fiber-optic cables are used to collect the reflected beam from the surface. The output of the fiber-optic system is fed to a back-propagation neural network to classify the resulting surface profile and predict the surface roughness value. The results obtained from the present study are then compared with the stylus measurement results. It is found that the resolution of the surface texture improves considerably in the case of optical method and the neural network developed for this purpose can classify the surface texture according to the control charts developed mathematically.