基于图形制导复杂曲面最佳适配的梯度-模拟退火算法

针对复杂曲面类毛坯与设计原形存在误差，提出了一种基于图形制导复杂曲面最佳适配的梯度一模拟退火算法。该算法首先利用参数曲面的几何不变性，使曲面有一个较为理想的初始位置。然后使用梯度一模拟退火算法对目标函数进行优化，从而实现了复杂曲面的最佳适配。最后运用实例验证了所提出的算法的可行性及有效性。

A Gradient-Simulated Annealing Algorithm of Graph-Guided Best Fitting of Complex Surfaces

Blanks of sculptured surface usually do not have positioning references, and poor positioning of a workpiece on a machine tool will lead to a great deal of allowance fluctuation and may even result in making the finished workpiece unacceptable. In order to improve the blank positioning, this paper presents a gradient-simulated annealing algorithm of graph-guided best fitting of complex surfaces. The algorithm will be explained in sections 1 and 2. Section 1 starts from the research results of Ref.4 and explains in detail our method of prelocation with the help of Fig.1. Our prelocation method ensures that a reasonable initial position of the blank is obtained. Section 2 explains in detail why the gradient-simulated annealing algorithm we propose can ensure global optimization, whereas already existing algorithms now and then give only local optimization. Section 2 also gives seven steps of the gradient-simulated annealing algorithm we propose. This method has been used for the positioning of hydroelectric turbine blades; it is proved that the algorithm is very efficient and the final positioning result is excellent.