Less Measurement Method Based On Micro-Vision For Micro-Heterogeneous Surface In Deep Hole

Surface roughness is an important index to affect the surface quality of the parts. It has a direct effect on the service performance, appearance and lifetime of the parts, especially for those parts with some special functions (sealed, relative movement, etc). Thus it is necessary to accurately measure the surface roughness of the work surface of the parts. However there are some difficulties to accurately detect the surface roughness for those deep-hole parts with micro-structural characteristics through the conventional surface roughness measurement methods, due to the poor accessibility caused by the small measurement space. Therefore, an adequate roughness measurement method for the micro-heterogeneous surface in deep hole is significant, which can detect and assess the surface morphologies, and then provide a detection basic and technical support for improving the surface quality of those parts.In this paper, the R-surface, located within the deep-hole valve, was taken as the object to study the surface roughness measurement method for the micro-heterogeneous surface in deep hole. Comparing the advantages and disadvantages of various surface roughness methods and their application scope, and combining with the measuring difficulties of R-surface, we proposed a surface roughness method for micro-heterogeneous in deep hole based on micro-vision. The microscopic image acquisition method for both the sample and test valves and the surface roughness value access method for the sample valves were introduced in detail firstly. Analyzing the microscopic image of R-surface by Gray Level Co-occurrence Matrix (GLCM) method, one of the statistical analysis methods, and then6characteristics were extracted from the GLCM:Angle of Second-order Moment, the differential entropy, correlation, Inverse Gap, Variance and the Maximum Probability. All of these characteristics show an approximate linear relationship with the surface roughness value. So the statistics characteristic parameters of the microscopic image of the sample vales and the corresponding surface roughness values consisted of the database to establish the relationship model between the two variants. Based on the database, two surface roughness estimation models were established by the neural network method:the single BP neural network model and the hybrid GA-BP model integrated of the Genetic Algorithm (GA) and neural network together. Then the comprehensive accuracy of the entire system was tested with the test valves, and the results showed that the BP model and GA-BP model both had high estimation accuracy, thus both of the two models can meet the accuracy requirement of surface roughness measurement. However, we could also find that the hybrid GA-BP relationship model had even higher measurement accuracy (95.25%). At last, taking Visual Basic Language and MATLAB software as the platform, the surface roughness measurement software for R-surface of valves was designed.The proposed surface roughness measurement method based on micro-vision for micro-heterogeneous surface in deep hole has high measurement accuracy. It is an effective solution to the surface roughness measurement difficulties while measuring the deep-hole micro-heterogeneous surface, and provides a new thinking for the surface roughness measurement of such parts. The proposed method has been successfully applied in a factory for the surface roughness measurement of such valves, and realizes the engineering applications of surface roughness measurement for the deep-hole micro-heterogeneous surface.