Research On Algorithm And System Of Grinding Allowance Detection And Surface Reconstruction For Engine Blade

With the development of aviation industry,the market demand of aviation engine is growing.As the key component of aero-engine,blade grinding becomes the key technology that determine blade quality,automatic grinding and on-line detection of blade become the focus of research.In this paper,the residual detec tion algorithm in blade grinding,the blade point cloud filtering algorithm and the surface reconstruction algorithm of blade three-dimensional model are studied.The main contents are as follows:First of all,according to the actual needs of the project,establish three dimensional scanning and model processing system of blade,puts forward the design indexes of the algorithm.Establish the relationship between the coordinate system,derive matrix formula leaves point cloud model acquisition and obtiane the 3D point cloud model leaves.Based on the error detection and reconstruction process,analyzes the sources of system error,given the blade precise 3D model can not be obtained,using 0 gauge to measuring error.Get the 3D model and the actual block gauge by comparing the scanning measurement,system accuracy can meet the requirements of coarse grinding blade removal allowance is designed in this paper.Secondly,the detection algorithm of blade grinding allowance is studied.According to the characteristics of blade and abrasive belt,the method,principle and grinding path of residual detection is determined.According to the principle of grinding and characteristics of data of accurate balde profile,contour fitting NURBS equation and compute the normal position.Analysis of the rough contour of the leaf in different noise form,remove the large noise by mean filter algorithm,for holes that caused by mean filter and small noise,proposed the adaptive moving least squares support domain radius(MLS)fitting algorithm that remove the small noise and to fill the holes.Finally,the grinding allowance at different positions is calculated,and the engineering requirements are satisfied by engineering practice.Thirdly,aiming at the problem of noise in the leaves of point cloud model,the filtering algorithm is studied.The types of noises occurring in the process of blade scanning are analyzed,and the outlier points in the cloud points are removed by statistical analysis.Comparison of different algorithms o f point cloud surface feature estimation,using PCA principal component analysis method to calculate the normal and curvature of point cloud model.According to the deviation of normal calculation,by setting the threshold to the normal angle of sampling p oints in a normal neighborhood limit angle range,a correction factor is introduced to modify the normals.Finally,the three edge filter operator is used to smooth the small noise.Through the contrast experiment mean filtering,bilateral filtering algori thm,the algorithm is time-consuming,but the noise points are removed while retaining the details of the surface details of the blade.Finally,the surface reconstruction algorithm of blade point cloud is studied.In order to solve the problem of large number of cloud points and slow reconstruction time of the filtered blades,a down sampling algorithm is used to simplify the cloud data.Because of the contradiction between the number of point cloud and the reconstruction effect and the reconstruction time,the optimal ratio of the number of points cloud to the reconstruction effect and time is compared by experiment.On the basis of the region growing algorithm,the blade point cloud in the neighborhood of sampling points is projected to the plane according to the method,and a variety of nearest neighbor screening methods are used to find the best nearest neighbor point and complete the fast reconstruction of the blade point cloud.Finally,the experimental results show that the proposed algorithm improves both the reconstruction effect and the reconstruction time.