Experimental Investigation On The Optical Measurement Of Supersmooth Surface Characteristics

Scattering characteristics and surface roughness on the surface of the materials make a very important influence on the performance of the product. The multi-wavelength fiber sensor for measuring surface roughness and surface scattering characteristics based on laser scattering was investigated. The probe of fiber optic sensor uses the special geometric design, specimens with different surface roughness(aR =0.10mm, 0.20mm, 0.40mm, 0.80mm) were analyzed by using 650 nm, 1310 nm and1550nm laser as the light source, respectively. The working distance of 2mm was chosen as the optimum measurement distance. The experimental results indicate that:(1) Under the same wavelength, the reflection intensity measured from the reflective surface of grinding samples decrease with the increase of surface roughness. Under the same roughness, the incident wavelength is longer, the reflection intensity is bigger.(2) The multi-wavelength fiber sensor can accurately measure surface roughness,and can effectively reduce the systematic error. The range of relative error of fiber optic sensor by analyzing system error is about 3.56%~7.43%.(3) The light scattering intensity ratio has a good linear relationship with the surface roughness.(4) The minimum relative error of the surface roughness is 2.92%, the maximum relative error is 13.4%, and the average relative error is about 7.48%. The accuracy for measuring surface roughness by multi-wavelength fiber sensor is about twice as large as that by single-wavelength fiber sensor.The measurement error is bigger by multi-wavelength fiber sensor for measuring surface roughness. In order to better measure the roughness, support vector machine is introduced to predict the surface roughness by using support vector regression. In this paper, the specimens with different surface roughness(aR = 0.012mm, 0.025mm, 0.05mm,0.10mm) were analyzed by using support vector machine when the incident wavelengths were 650 nm, 1310 nm, respectively. The working distance of 2.5mm~3.5mm was chosen as the optimum measurement distance. The target value is achieved by the regression algorithm of the measured data using LIBSVM software. The experimental results indicate that the mean squared error of regression prediction is 6.40444×10-7, and squared correlation coefficient is 0.999705 at 650 nm. The mean squared error of regression prediction is 6.72513×10-7, and squared correlation coefficient is 0.999838 at1310 nm. The average relative error is about 2.669% at 650 nm, while it is about 2.431%at 1310 nm. The average prediction error is reduced with the increase of incident wavelength. The prediction error of the roughness is less than 3% by using SVR.The speckle patterns are formed in the scattering space when the incident light irradiate on the rough surface, and a large amount of surface structure information are carried by the speckle field, so the research on speckle field formed by the rough surface is of great significance. The speckle images of rough surface are collected by beam quality analysis system. The speckle images are processed using Matlab software, and speckle contrast values are calculated by spatial-average method. The relationship between speckle contrast and surface roughness is fitted using second-order polynomial.The experimental results show that the roughness is larger, the relative error is smaller,and the maximum relative error is 12.5%. The surface roughness measurement method based on the average contrast laser speckle is feasible. The method has the advantages of non-contact, high efficiency, simple device.In conclusion, the optical fiber sensing method, support vector machine(SVM) and differential scattering method for measuring the surface roughness is feasible. Great help is provided for the precision measurement of optical components in optical engineering.