适用于表面污染检测的机械臂轨迹生成研究

发布时间：2017-12-27 17:09

本文关键词：适用于表面污染检测的机械臂轨迹生成研究　出处：《西南科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：核退役是核工业机器人重点应用场景之一,其中的智能化核退役工程是重中之重。在核退役工程中涉及到繁杂的核工业退役工艺流程,其中很重要的一项是表面污染检测,检测结果将影响被退役核设施的处置方案实施。本课题以当前国内外核退役技术、机器人技术以及表面污染检测技术的研究现状为背景,主要围绕机器人进行表面污染检测时,如何使机械臂末端的检测仪器,在满足表面污染相关检测特性的前提下,可以顺利完成表面污染检测任务展开研究。课题研究目的是设计一套可行的方案。在本方案中使用线结构光扫描系统完成待测物表面点云数据的采集,经过法向估计和适于表面污染检测特性的点云法向偏置处理,得到检测轨迹所在曲面的点云。通过点云切片等方法规划出机械臂任务空间的检测轨迹,机器人根据生成的轨迹自主完成检测目标的表面污染检测任务。本课题针对表面污染检测,主要研究了α、β这两种核退役表面污染检测中重点关注的放射性污染源辐射特性及常用的检测方法,结合对中核821厂的待退役核反应堆及其周边辅助设施的实地考察,对待检测对象和表面污染检测工程的具体条件和需求进行了深入地分析和总结。课题使用D-H方法对MOTOMAN-MH6S机械臂建立连杆模型,求解其正运动学方程,针对使用传统方法求解机械臂逆运动学存在计算量比较大、且传统BP(Back Propagation)神经网络在被用来求解机器人逆运动学时会出现输出误差偏大的问题,这主要因其易陷入局部极值导致,论文对一种基于PSO优化的BP神经网络在求解机器人逆运动学中的应用进行较深入地研究。在满足表面污染检测任务的相关需求下,为了能够获取任务空间的检测轨迹,在研究过程中,对待检测物进行表面深度信息扫描,获取其点云数据,经过对获取到点云数据进行预处理,得到更加有利于生成任务空间检测轨迹的点云数据。再通过点云数据表面法线估计以及法向偏置算法,得到分别适用于α、β这两种放射源检测距离下检测轨迹所在曲面的点云数据,对其进行点云切片等算法处理,最终完成任务空间的轨迹生成,并对生成的轨迹进行了实验验证。本课题完成了表面污染检测特性及核辐射探测方法研究、机械臂运动学研究、任务空间中相关任务需求下的轨迹生成研究,并基于C++语言结合PCL开源库在VS2010平台下对各算法进行了设计和实现,同时对本文研究结果进行实验验证。实验结果表明:本文方案生成的检测轨迹在进行模拟表面污染检测任务过程中,针对直径60cm、70cm、80cm、90cm、100cm管道上切割下来的待检测曲面,进行α和β表面污染模拟检测时,检测距离平均有效率分别达70.25%和81.27%、71.83%和82.67%、72.41%和83.24%、73.76%和84.57%、75.03%和85.88%,从而验证了该方案的可行性和有效性。
[Abstract]:Nuclear decommissioning is one of the key application scenes of the nuclear industry robot, and the intelligent nuclear decommissioning project is the most important. In the nuclear decommissioning project, there is a complicated process of decommissioning of nuclear industry. A very important aspect is surface contamination detection. The test results will affect the disposition of decommissioned nuclear facilities. The project is based on the current status of nuclear decommissioning technology, robot technology and surface contamination detection technology, mainly around the robot surface contamination detection, how to make the instrument manipulator, to meet the surface pollution related detection characteristics of the premise, can successfully complete the task of surface contamination detection. The purpose of the research is to design a feasible scheme. In this scheme, the line structured light scanning system is used to complete the collection of the point cloud data on the surface of the object to be tested. After the normal direction estimation and the point cloud normal bias processing for the surface pollution detection feature, we get the point cloud for detecting the track surface. Through the point cloud slicing and other methods, we can plan the detection track of the manipulator's task space, and the robot completes the detection task's surface pollution detection task independently according to the generated trajectory. This thesis mainly studies the surface contamination detection, radioactive pollution source radiation characteristics focus on the detection of alpha and beta of the two nuclear decommissioning of surface contamination and detection methods commonly used, combined with the field survey of 821 nuclear plants for decommissioning of nuclear reactor and its surrounding ancillary facilities, to detect objects and surface contamination detection project conditions and needs to conduct in-depth analysis and summary. Use the D-H method to establish linkage model research on MOTOMAN-MH6S manipulator, solving the kinematics equation, to use traditional method to solve the kinematics calculation is relatively large, and the traditional BP (Back Propagation) neural network was used to solve the inverse kinematics of robot hours will output the big error problem, which is mainly due to its easy the lead into the local extremum, a BP neural network based on PSO optimization in solving inverse kinematics of robot application is deeply researched. To meet the needs of related surface contamination detection task, in order to detect the track to get the task space, in the course of the study, to test the surface depth information scanning, obtain the point cloud data, the point cloud data obtained after the pretreatment, get more point cloud data to detect task space trajectory generation. Then the algorithm to bias by point cloud data and surface normal estimation method, point cloud data were applied to these two kinds of alpha and beta radiation source detection distance under the curved surface, the point cloud slicing algorithm, trajectory generation and complete the task space, and the trajectory of the generation of experiments verify. This paper completed the research of trajectory generation tasks related to nuclear radiation detection method and detection characteristics of surface pollution research, kinematics research, task space requirements, and based on C++ language combined with PCL Library in the VS2010 platform for the design and implementation of the algorithm, and experiments are carried out to validate the results of this study. The experimental results show that this scheme generated in the detection of trajectory simulation surface contamination detection tasks, for detecting surface cut diameter of 60cm, 70cm, 80cm, 90cm, 100cm pipe, alpha and beta surface contamination simulation test, the average efficiency of 70.25% and 81.27%, 71.83% and 82.67%, 72.41% and 83.24% 73.76%, and 84.57%, 75.03% and 85.88% respectively with the detection distance, which verifies the validity and feasibility of the scheme.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TL943;TP242

【相似文献】