Research On Non-contact Detection System Of High-speed Steel Roll Roughnes

As an excellent roll variety in modern metallurgical production field,high-speed steel roll is widely used in iron and steel manufacturing industry.In recent years,with the continuous improvement of industrial production technology,the surface quality inspection of high-speed steel roll is paid more and more attention.Surface roughness,as one of the important parameters to measure surface quality,has a great impact on the overall life and working performance of high-speed steel rolls.Currently,most of the roughness detection methods for high-speed steel rolls are contact detection.This method has slow measurement efficiency and certain wear on the roll surface,which does not meet the production requirements of modern industry.The fast and non-destructive non-contact testing technology is more suitable for measuring the roughness of high-speed steel roll.Therefore,this paper takes the non-contact detection system of high-speed steel roll roughness as the research object,and compares two optical methods,light scattering method and optical speckle method,which are most commonly used in the current non-contact detection technology,to make the following research:(1)Based on the principle of light scattering method,the relationship between light scattering and surface roughness is analyzed,and a set of light scattering image acquisition platform is designed to collect four groups of scattered images with different roughness levels.Image processing techniques such as image enhancement,Gaussian filtering and image rotation are used to obtain ideal scattered images,and feature extraction is carried out on the processed images.Fourteen groups of characteristic parameters,including scattering characteristic value and standard difference,were obtained and fitted,and the parameters with monotonicity with roughness were selected for further study.(2)Based on the analysis of the relationship between the principle of optical speckle method and the surface roughness of optical speckle,the speckle image acquisition platform was modified on the basis of the light scattering image acquisition platform.For speckle images with larger noise,the extracted image feature values had little correlation with the roughness after image clipping,gray enhancement and non-local mean filtering.In order to improve the unsatisfactory protection of speckle image information by traditional image processing,a non-subsampled contour wave(NSCT)transform based on stationary wavelet(SWT)was proposed in this paper to extract 8 groups of feature parameters of decomposed speckle image gray co-occurrence matrix and Tamura texture feature.By analyzing the correlation between the characteristic parameters and the roughness in the three scales after NSCT transformation,the decomposed image in the direction of 2 scales and 1 direction is finally selected as the research object.(3)Analyze the principle of support vector machine classification model optimized by genetic algorithm,conduct training tests on the extracted characteristic parameters,and compare the classification effects of four optimization algorithms,namely cross validation,sparrow search,particle swarm optimization and genetics,and verify that the classification accuracy of support vector machine optimized by genetic algorithm is higher.By comparing the classification results,Finally,speckle method with better classification effect is selected as the scheme to detect the roughness of high-speed steel roll.(4)Design GUI user graphical interface based on MATLAB,including the design of image acquisition module,image processing module,image display module and classification and recognition module.After the completion of interface development,experiment verification is carried out.Through the identification results,the feasibility of non-contact detection system for high-speed steel roll roughness based on optical speckle method is verified...