| A large volume-ratio workpiece is a workpiece consisting of two or more parts with large difference in volume.As a typical representative,solid rocket motor is the most commonly used weapon and spacecraft propulsion mechanism.The success or failure of assembly is directly related to the quality of the product.The main components of a solid rocket motor are large-volume combustion chamber and small-volume nozzle.The assembly of these two parts is fixed by bolts between the end assembling faces.In addition,one or more step faces are often processed on the assembling surface(convex step and recessed step)to achieve sealing.The geomatric measurement project of the large volume ratio of the workpiece's assembling surface is very complicated,and the conventional measurement method cannot perform the detection efficiently or enen effectively.In this paper,we take the assembling surfaces of solid rocket motor which is the representative of large volume-ratio workpiece as the research object,aim at their geometric measurement,develop a large volume-ratio workpiece assembling surface geomatric measurement system and processe data of the assembling surfaces which is gained by the measurement system,the main research work is as follows:(1)Development of geometric measurement machine for solid rocket motor assembly surfaceCombined with the shape characteristics of the assembly surface,and based on the advantages of the coordinate measuring machine and the articulated arm coordinate measuring machine(AACMM),an articulated arm coordinate measuring machine based on the gantry,rotative arm and two-dimensional line laser sensor was developed.The two-dimensional line laser sensor can realize high-efficiency and high-precision full-scale detection of the assembling surface for combustion chamber and the nozzle.The introduction of the gantry structure ensures that the combustion chamber and the nozzle of large volume ratio can be detected in the same coordinate system.The use of a rotative arm provides the measuring machine with a flexible measurement space to cope with the challenges of complex measurement projects,while eliminating manual guidance,thereby reducing the introduction of human factors.(2)Laser signal denoising algorithmThe original laser signal is affected by ambient illumination,the shape of the surface to be tested and the physical limitations of the measuring machine during the acquisition process.Inevitably there will be noise.This paper proposes an adaptive mathematical morphology filtering algorithm for laser signal denoising.The algorithm fits the laser signal based on RANSAC algorithm,and then uses the segmentation line as the morphological filter template to adaptively control the template length,and achieves the protection of the sharp corner points while filtering out the noise.(3)Point cloud compression algorithmFor the massive 3D point cloud data of the assembling surface acquired by the measurement machine,a point cloud compression algorithm based on the law of universal gravitation law is proposed.The algorithm makes full use of the organization characteristics of point cloud data acquired by the measurement machine.By projecting the point cloud to the two-dimensional plane,the "combination force" parameter of each point is obtained,and based on this parameter,the point cloud is divided into "non-feature point set" and "feature point set".The "feature point set" is protected and the "non-feature point set" is sampled in large proportion,and the point cloud compression with feature retention is realized.(4)Point cloud sharp feature differential information estimation algorithmFor the sharp features such as steps,the traditional K nearest neighborhood(KNN)or the optimal K nearest neighbor algorithm optimized by the feature entropy of the structure tensor covariance matrix cannot effectively solve the problem that there are always multiple plane points in the neighborhood of the sampling point at the sharp feature.Based on the optimal K-neighborhood,this paper proposes a neighborhood selection method based on 2D eigen-value maximization.This method selects the point on the plane with the highest point density in the best neighborhood of the sampling point at the sharp feature.This neighborhood avoids the normal vector distortion caused by the existence of multiple planes in the neighborhood.(5)Assembling surface geometric value estimation algorithmIn this paper,3D feature information is composed into feature vector,and the "feature point set" segmented in point cloud compression process is refined based on this vector,and divided into "non-feature plane point set","thread hole point set" and "step cylinder point set".By slicing "non-feature point set" and plane fitting with RANSAC algorithm,the plane with the largest number of points is used as reference plane,and each point set in the point cloud is projected onto this plane separately and then the corresponding raster bitmap is obtained.The edge of the bitmap is detected by mathematical morphology gradient,and then the boundary points of point cloud are obtained.The RANSAC algorithm is used to fit the boundary point set to obtain the geometric information of circle features on the assembling surface.The "step cylinder point set" is again estimated by RANSAC algorithm to detect its rotative axe,and then the point set is projected onto the rotative axis,and the projected maximum and minimum values are calculated to gain the step height or depth.Flange workpiece is used to simulate the assembling surface of the solid rocket motor.The experimental results show that the maximum error of the measurement system for the inner and outer diameter of the threaded hole is 0.05 mm,the maximum root mean square error is 0.018 mm,and the maximum error for the step diameter detection is 0.04 mm,the root mean square error is 0.02mm;the maximum error of the outer diameter detection of the workpiece is 0.06 mm,the root mean square error is 0.023mm;the maximum error of the step height(depth)detection is 0.03 mm,and the root mean square error is 0.016 mm. |
PDF Full Text Request