Acceleration of Geometric Computations in CAD and CAM by Using Polygon Rendering Hardware

为了适应顾客不断变化的喜好，迅速销售具有市场竞争力的产品，日本制造企业努力缩短设计完成后的生产准备时间。生产准备的大部分时间用于制造冲压、铸造和塑料成型零件的模具。模具通常使用具有各种形状刀具的3轴数控（NC）铣床制造。可靠刀位轨迹的计算一直是一个难题。铣削一个复杂的模具，通常需要几个小时来计算整个刀具路径。本项目的成员共同开发了一种新的技术，用于快速生成和验证数控铣削刀具路径。我们的技术使用图形处理单元（GPU），这是目前在大多数计算机上配备的，用于加速几何计算的工艺规划，数控路径生成和路径验证。作为项目的成果，实现了以下技术：（1）开发了一个实用的GPU加速CAM系统。该系统目前已作为日本某汽车制造商的核心系统运行。（2）提出了一种将前一次加工的剩余加工区域投影到下一次加工的刀位曲面的算法。（3）开发了曲面定义和几何质量评价的GPU加速算法;（4）开发了非制造领域几何计算的GPU加速算法。提出的算法实现和实验的CAD/CAM/CAPP程序使用这种技术进行了演示。

In order to follow frequently changing customer preferences and to quickly distribute products with high market competitiveness, Japanese manufacturing companies make efforts to shorten the production preparation time after the design completion. Most time period of the production preparation is consumed for making dies and molds for stamping, casting and plastic molding of parts. Dies and molds are usually fabricated using 3-axis numerically controlled (NC) milling machines with various shape cutters. Computation of the reliable cutter path is still a difficult problem. Several hours are often needed for computing whole cutter paths for milling a complex die. Members in this project jointly develop a new technology for fast generation and verification of the cutter path for NC milling. Our technology uses graphics processing unit (GPU) which is currently equipped in most computers, for accelerating the geometric computation in process planning, NC path generation and path verification. As the result of the project, the following technologies are realized ; (1) a practical GPU accelerated CAM system is developed. This system is currently running as a kernel system of one Japanese automobile manufacturer, (2) an algorithm for projecting the machining remained area in a prior machining to the cutter location surface for the next machining is proposed. This projection is useful for automating some process planning task of mold machining, (3) some GPU accelerated algorithms for defining the curved surface and evaluating its geometric quality are developed, (4) several algorithms for accelerating geometric computations in non-manufacturing fields are developed also. Proposed algorithms are implemented and an experimental CAD/CAM/CAPP programs using this technology are demonstrated.