Research On Visual Inspection Technology For Turbine Blade Surface Defect

Aero engine turbine blades are an important part of aero engines,and various scratch defects on their surfaces during the production process will affect the working efficiency,reliability and safety of the engine.Therefore,in order to ensure flight safety,it is necessary to systematically detect the defects of turbine blades before they are put into use.In order to improve the efficiency and accuracy of turbine blade defect detection,get rid of inefficient manpower detection,and meet the actual needs of industrial production,a new visual inspection scheme for surface defects of aero-engine turbine blades is proposed.The principal studies and deliverables of the dissertation are as listed below:(1)General layout.Based on testing needs,the monocular four-indexed fixed-point curved guide light source mechanism vision inspection system is proposed as a general design scheme,and the hardware system selection and software system functional interface design are determined.(2)Image preprocessing.Combined with the uneven sparse dot matrix and two-color reflection characteristics of the turbine blade surface to the light source,the specular highlight removal method based on L0 norm regularization is used to remove the local ununiform high reflection in the turbine blade image,and the image grayscale linear enhancement method is used to expand the gap between the defects and the background of the turbine blade,highlight the defect characteristics,and propose a multi-view turbine blade alignment technique based on apparent homogeneity of outer profile edges to align the turbine blade registration in different viewing angles,and the algorithm performs well in terms of combined accuracy and speed.(3)Image fusion.To address the difficulty of globally presenting defects in a single view for image defects of turbine blades with free-form characteristics,which is not conducive to later repair,a dual-scale decomposition and significance image fusion algorithm based on guided filtering are proposed,and experiments show that the fused defect target and metal background are relatively complete,uniform,robust,and the overall performance is better than some other comparative algorithms,and its subjective and objective evaluation comprehensive indicators are in the forefront,and have high real-time performance.It can be applied to the real-time defect detection site of aviation turbine blade industrial production.(4)Three-dimensional texture reconstruction.Aiming at the problem that it is difficult to distinguish between soft and hard scratches of turbine blades,a photometric stereo-based 3D texture reconstruction method for defects is proposed.Using the calculated fixed-point light source mechanism data,the lookup table light source calibration method is designed,and then the turbine blade surface defect depth information reconstruction is realized by solving the turbine blade surface normal vector,and after depth information data analysis,hard scratches and soft scratches in the defects are distinguished.