The Three-dimension Evaluation System Establishment And Detection Method Research Of Casting Surface Roughness

Casting surface roughness has significant effects on the casting appearancequality and wear resistance, lubrication, fatigue resistance and corrosion resistanceperformance. Therefore, reasonable evaluation of casting surface roughness isparticularly important. However the traditional casting surface roughness evaluationmethod mainly is casting standard sample pieces and touch probe measuring method.The former detection results are greatly influenced by artificial factors, the latter hasbeen wandering in the two-dimensional measurement phase for years, which is verydifficult to reflect the micro topography of casting surface as a whole. Therefore, it isvery necessary to carry out more accurate technology research about casting surfaceroughness evaluation.Based on the thorough analysis and research on the two-dimensional andthree-dimensional parameters of surface roughness, combined with the random andisotropic textures characteristics of the casting surface, a set of relativelycomprehensive3D evaluation system was built. Various measuring methods of thesurface roughness were systematically analyzed, finally combined with computervision image spectrum analysis and neural network technology, a new method ofcasting surface roughness detection was proposed. This paper took a large number ofsurface images of casting standard specimens with different surface roughness, andpre-processed these images, including Image filtering and image enhancementprocessing, which makes it beneficial to the feature extraction. Then power spectrumimages were obtained through two-dimensional discrete Fourier transform of surfaceimages, the analysis found that the power spectral radius and the energy distributionof different circle ratio show approximate monotonic relationship with surface roughness value. With these characteristic parameters as input, and three dimensionalevaluation parameters as output, the BP neural network model was established.Finally, the measure software system was designed using MATLAB software, andthen the3D evaluation parameters of some casting specimens and actual castingswere forecasted by using the measure system, the results were compared with thosethat were measured by3D surface profile measuring instrument, the average errorrate of every parameter was not beyond5%, which satisfied the requirement ofmeasurement and proved the feasibility and accuracy of the method. It laid thefoundation for casting surface roughness three-dimensional evaluation research in thefuture.