Dong Zhichao, Cheng Haobo
Appl Opt. 2015 Oct 20;54(30):8884-90. doi: 10.1364/AO.54.008884.
This study intends to address several problems that are still obstructing the complete practicability of the linear equation dwell time model (LEDTM) used in deterministic subaperture polishing, including memory cost, time cost, arbitrary boundary, and arbitrary tool path. For a large-scale surface error matrix, the memory cost and time cost are two major problems of concern. Here, we present a method that uses the operation of a sparse matrix to build and store the coefficient matrix of the linear equation, which can reduce the memory cost and time cost tens to hundreds of times, thus making LEDTM readily deal with a large-scale surface error matrix on a common personal computer, with a time cost of ∼1-6 s for a surface error matrix with a scale of 300×300 to 600×600. The compatibility for an arbitrary surface error boundary and tool path is also addressed. Using the proposed method, we believe the LEDTM reaches a complete practicability in engineering.
本研究旨在解决几个仍在阻碍确定性子孔径抛光中使用的线性方程驻留时间模型（LEDTM）完全实用性的问题，包括内存成本、时间成本、任意边界和任意刀具路径。对于大规模表面误差矩阵，内存成本和时间成本是两个主要关注的问题。在此，我们提出一种方法，该方法利用稀疏矩阵运算来构建和存储线性方程的系数矩阵，这可以将内存成本和时间成本降低数十到数百倍，从而使LEDTM能够在普通个人计算机上轻松处理大规模表面误差矩阵，对于规模为300×300至600×600的表面误差矩阵，时间成本约为1 - 6秒。还解决了与任意表面误差边界和刀具路径的兼容性问题。使用所提出的方法，我们相信LEDTM在工程中达到了完全实用性。
