Abstract
Ruled surfaces provide a good capability of complex product modeling. Such geometry has been widely used in the aero-space, automobile, and mold industries. For example, turbine blade is a key component in energy and air-conditioning applications. The shape design involves advanced fluid mechanics. The manufacturing task is generally performed by companies specialized in five-axis CNC machining. It requires computer-aided manufacturing (CAM) technology and cutting experiences. The development of turbine blade is considered an indicator of industrial competence. In current practice, the cutter goes along the rulings of a ruled surface, thus finishing the machining job. However, this machining method does not produce optimal machining quality. Thus, several studies had attempted to improve it. Most of the adopt local optimization approaches or heuristic algorithms, which do not guarantee globally best result. Their computational efficiency is poor. As a result, the tool path planning of five-axis flank milling is highly limited. The machining quality cannot be precisely controlled. To overcome this deficiency, this research proposes a novel computational scheme that estimates the machining error using general-purpose graphic processing unit (GPGPU) in five-axis flank milling. GPGPU provides excellent functionality of parallel processing. The processing speed of floating number is faster than CPU. This computing technology has been applied to many engineering problems in addition to computer graphics. This research is one pioneering work that applies GPGPU in five-axis tool path planning. It cooperates physical limitations into the path planning such as feed, rotation angles, and cutting length. Quadrilateral patches are used for generation of optimal tool path in terms of minimal machining error. We convert a geometric problem into a math programming task. First, a network is constructed for modeling feasible solution of tool motions. The machining errors induced by individual tool motions become the weights between the nodes of the network. It thus transforms into a shortest path problem and can be easily solved by the Dijkstra’s algorithm. For complex machining conditions, we employ GA (Genetic Algorithm) for searching optimal tool path empowered by the parallel processing provided by GPGPU. The resultant GA algorithm can perform cross-over and mutation operations rapidly. It also considers linear interpolation between tool motions. Efficient encoding/decoding schemes are developed to fulfill the actual requirements of 5-axis machining. Simulation result produced by commercial software is obtained for verification purpose. Finally, a real cutting experiment is conducted to validate the effectiveness of the proposed scheme. CMM measure of the machining part shows that it significantly improves the machining quality of 5-axis flank milling. This work provides advantages on computational efficiency, machined quality, and planning flexibility for 5-axis flank milling of complex ruled geometry.