Optical Measurement And Roughness Evaluation Of Machined Surface Topography

With the rapid development in intelligent manufacturing industry upgrade recent years,the requirement for equipment manufacturing industrial standardization building and product quality control have become increasingly urgent.Product surface roughness plays a significant role in mechanical equipment and instrument performance and service life.The lack of efficient and rapid surface reconstruction and surface quality evaluation method hinders the product on-line quality monitoring and quality traceability in its entire service life.In order to meet the requirements of high-efficiency on-line in-situ surfaces roughness measurement and surface texture reconstruction,by using combined methods of theoretical research and scientific experiment,this research has carried out a study related to local machined surface texture noncontact measurement and roughness evaluation.By using residential networking(ResNet)method in machined surface images,the research acquires the prior knowledge to guide the following feature selection.This dissertation also proposed a curvature filtering enhanced shape from shading(SFS)surface reconstruction method with the engagement of differential geometry theory.The main research contents and achievements are as follows.1.The feature extraction capability of dual-tree complex wavelet transform(DTCWT)is investigated in both one-dimensional and two-dimensional signals.Based on numerical simulation,the feature extraction characteristic of real discrete wavelet transform(DWT)and DTCWT are described in detail as well as the multi-scale decomposition capability.Combined with the convolutional neural network(CNN),the research proposes an intelligent fault diagnosis method for one-dimensional signal and verified in simulation and experimental data.Using the multi-directional selectivity of DTCWT,an edge contour extraction method was also proposed and applied in indexable insert digital image.2.A two-tree complex wavelet enhanced SFS method is proposed.The theory of the SFS method and the solving process of the partial differential equation(PDE)are investigated.By introduce Wallis filtering algorithm in remote sensing technology,the uniform illumination image is acquired,and then the reconstruction is performed through the SFS method.Finally,the surface feature contour information is achieved by the DTCWT.The proposed method can effectively evaluate the machining quality of aviation aluminum alloy surface,and can be effectively applied to the online non-contact measurement and metal milling process state monitoring or processing quality evaluation.3.An intelligent surface roughness evaluation method was proposed.The effect of surface roughness measurement direction on the measurement result was researched.A digital image skew correction algorithm based on enhanced Sobel operator and Hough transform was proposed.The ResNet artificial intelligence algorithm is used to classify the digital image's surface roughness,and the recognition accuracy was 95.14%in validation set.At the same time,the performance of each layer filter in the network structure is also analyzed;the research found that the function of the established model is similar with the surface roughness comparison sample block which commonly used in industrial.4.The dissertation proposed a curvature enhanced SFS method.Combined with differential geometry theory,the research embeds the digital image in three-dimensional space and optimizes the curvature.The proposed curvature optimization was used to replace the smooth constraint term in the SFS surface reconstruction process.The proposed curvature enhanced SFS method was successfully applied in a simulation model and cast iron surface.