The Surface Defect Detection Based On Sensor Signals And Surface Texture

Aiming at problems of batch of workpieces machining quality classifying andsurface defects locating, in this dissertation, the study regards the spindle power signalsand the bore section in the machining process as the thesis materials. New approachesare applied in the batch machining process. The AMT feature of power signals isassociated with the surface machining quality, according to the amount of featuresinformation to finish machining quality classification. Then in accordance with thedirection characteristics of surface texture to locate the surface defects. The researchsurrounding the problems that must be studied in extracting the power signals AMTfeatures and surface texture features. The main contents include: the design andconstruction of the monitor signal and image acquisition hardware platform, the powersignal pre-processing and AMT features extraction, the image pre-processing and texturelocal direction feature extraction. The main aspects of the study are as follows:1. The KT5A/P-type sensor is adopted to collect power monitor signal in themachining process, and the Microvision products (including the workpiece placedplatforms, industrial digital cameras, image data acquisition card and lighting device) areselected to build the hardware platform of the workpiece surface image acquisition. TheMVtec HALCON software is selected to develop the programs for image acquisition.2. The spindle power signals of batch drilling were preprocessed by methods ofz-score normalization and segmentation, and then scale features of the monitor signalswere extracted from local to global by the improved AMT algorithm. Adopting PrincipalComponent Analysis method to reduce feature dimensions, and the PCA featureinformation is associated with the surface machining quality. According to the amount ofPCA information, the research realizes the classifying of workpiece processing quality.At last, compared with the classifying result and artificial test result, the research verifiesthe reliability of the method.3. The limitations of classical PCA algorithm is examined in the texture defectsegmentation. Without increasing the complexity of the algorithm, the classical PCAalgorithm is improved to reduce the sensitivity of PCA algorithm to the local lightuneven phenomenon. And the defect partition effectiveness is improved which is basedon the threshold. All of these are prepared for the information exaction of texture feature.4. The research analyzes the of traditional surface texture features in indicating thesurface defects, the local direction characteristic is proposed to analysis the surfacedefect. Considering the noise jamming in the image, filters is designed to filter theimages, and texture image is enhanced based on the image processing techniques.5. On the base of image enhanced, model is constructed to extract the localdirection features of surface texture. At the same time, the correspondence relationship isestablished between the local direction characteristic of textures and the positions ofsurface defects. Then，based on the changing of local direction characteristics to identifythe position of surface defects.In conclusion, this dissertation realizes the classification of workpiece processingquality based on the AMT feature Principal Component informative of power signals.The location of surface defects are achieved by considering to the surface image texturedirection characteristic. The feasibility of the proposed methods are verified bycomparing with the classifying and artificial test results which are simulated by using theMATLAB program.