A novel robust Gaussian filtering method for the characterization of surface generation in ultra-precision machining
A novel robust Gaussian filtering method for the characterization of surface generation in ultra-precision machining
复制标题
DOI：
10.1016/j.precisioneng.2006.01.005
复制
发表时间：
2006-10-01
影响因子：
3.6
通讯作者：
To, S.
中科院分区：
文献类型：
--
作者：
Li, Huifen;Cheung, C. F.;To, S.
在经验理解加工条件与表面特征关系的基础上，为了预测表面形貌并提供最优的加工参数，对表面生成机理进行了大量的研究。虽然新型几何产品规格（GPS）和验证框架体系的形成推动了相关研究工作向新的表征方法和国际标准的起草方向发展，但对超精密加工表面质量评价至关重要的表面表征技术的应用研究相对较少。本文提出了一种新的鲁棒高斯滤波方法，并将其用于表征超精密加工表面的表面形貌。采用三次b样条和m估计，使方法具有较好的鲁棒性和可靠性。在比较经典加权函数性质的基础上，定义了一种自研鲁棒加权函数（ADRF），提高了RGF的鲁棒性。为了验证所提方法表征的可行性，采用计算机仿真的方法对真实的超精密加工表面进行了分析。实验结果表明，RGF方法不仅可以有效地分离整个测量区域的表面成分，而且可以消除异常值的影响。(c) 2006爱思唯尔公司版权所有。
A lot of research work has been focused on the study of the surface generation mechanisms in order to predict the surface topography and provide the optimal machined parameters based on the experiential understanding of relationship of machined conditions and surface features. Although the formation of novel geometrical product specification (GPS) and verification framework system promotes the relevant research work to new characterization methods and draft of international standards, relative little research work was conducted on the application of surface characterization techniques to ultra-precision machining which is very important to evaluate the surface quality. In this paper, a novel robust Gaussian filtering method (RGF) is proposed and used to characterize the surface topography of ultra-precision machined surfaces. Cubic B-spline and M-estimation are used to make the method reliable and robust. Based on the property comparisons of classical weighting functions, a novel auto-developed robust weighting function (ADRF) is defined to improve the robustness of RGF. To verify the characterization feasibility of the proposed method, computer simulation is used and then the real ultra-precision machined surfaces are analyzed. The experimental results indicate that the RGF method cannot only separate the surface components effectively on the whole measured area and but also eliminates the influence of freak outliers. (c) 2006 Elsevier Inc. All rights reserved.