Representations and Algorithms for Manufacturing Coordinate Metrology Point Clouds

成功的设计和制造需要使用最少的材料，并具有可重复的材料性能和形状生产。内存和并行处理的增加导致了对新的计算机辅助设计和制造网格和基于点云的表示和算法的需求，这些表示和算法可以提供工厂车间生产率的分析。解决这一需求缺口是研究计划的重点。具体的短期重点概述如下。 1)在零件拓扑优化中，指定了“非优化体积”(如装配配合曲面)约束，但剩余的“优化体积”可能会根据预期载荷和挠度/应变要求而减少。申请人实验室提供的软件将用于设计添加制造(AM)部件，这些部件将进行3D打印和测量，以确定自由状态几何形状和施加载荷下的挠度/应变。其结果将是对拓扑优化/AM流程的验证和更大的信心，以及行业更快地采用这一具有竞争力的新兴技术。 2)在使用计量数字化仪时，一个基本的任务是将点云与标称零件表面进行配准。对于非常薄的零件，例如板材，点与之匹配的曲面可能不明确。使用申请人的定制安装，对于每个表面点，将确定数字化仪朝向数字化仪的方向向量。然后，可以将具有相反法线向量的曲面作为匹配项拒绝。 3)最近获得的一台实验室数字化仪以接近每秒500,000个样本的速度收集点云。最多连接4个数字化仪，可以实现每秒200万个样本的数据点采集率。对于汽车镜面尺寸的零件，必须在6秒内将1200万个点与标称几何图形配准。单独使用串行台式计算机算法需要5分钟或更长时间进行相关计算。448并行核心图形处理单元视频卡算法的开发已将计算时间减少到15秒，最近实验室购买的2880核心技术提供了进一步加速的潜力。这项拟议的研究将利用这项新技术，目标是开发并行算法，以工厂底线的速度处理点云。 拟议的研究计划包括高就业能力的学生培训，使用最先进的软件，坐标计量数字化仪，和并行计算硬件。它解决了重要的AM设计优化和测量问题，对于保持加拿大在医疗器械、汽车、航空航天和消费品的轻量化设计和制造方面的竞争力至关重要。

Successful design and manufacturing requires minimum material use with repeatable material properties and shape production. Increased memory and parallel processing has led to demand for new Computer Aided Design and Manufacturing mesh and point cloud based representations and algorithms that can provide analyses at factory floor production rates. Addressing this demand gap is the focus of the research program. Specific short term emphasis is outlined below. 1) In part topology optimization, “non-optimization volume” (such as assembly mating surfaces) constraints are specified, but the remaining “optimization volume” may, depending on expected loads and deflection/strain requirements, be reduced. Software available in the applicant’s laboratory will be used to design Additive Manufacturing (AM) parts that will be 3D printed and measured to determine both free state geometry and deflection/strain under applied loads. The outcome will be validation and greater confidence in the topology optimization / AM process, and more rapid industry adoption of this competitive emerging technology. 2) When using metrology digitizers, an essential task is registration of the point cloud to the nominal part surface. For very thin parts, such as sheet metal, the surface to which the point matches can be ambiguous. Using the applicant's custom installation, for each surface point, the digitizer direction vector towards the digitizer will be determined. Surfaces with opposing normal vectors can then be rejected as matches. 3) A recently acquired laboratory digitizer collects point clouds at rates approaching 500,000 samples / second. Networking up to 4 digitizers, a data point collection rate of 2 million samples / second is achievable. For an automobile mirror sized part, 12 million points must, within 6 seconds, be registered with the nominal geometry. Sole use of serial desktop computer algorithms requires 5 minutes or more for the related computations. Development of a 448 parallel core Graphical Processing Unit video card algorithm has reduced the calculation time to 15 seconds, and recent laboratory purchase of 2880 core technology offers potential for more speed-up. The proposed research will exploit this new technology with the goal of developing parallel algorithms that will process point clouds at factory floor rates. The proposed research program involves highly employable student training in the use of state-of-the-art software, coordinate metrology digitizers, and parallel computing hardware. It addresses important AM design optimization and measurement issues essential to maintain Canadian competitiveness in lightweight design and manufacture of medical appliance, automotive, aerospace, and consumer goods.