一种改进的Douglas-Peucker数控加工轨迹压缩方法
Improved Douglas-Peucker CNC Machining Trajectory Compression Method
王品 1王婧如 2张丽鹏 2王森 2荆东东2
作者信息
- 1. 中国科学院沈阳计算技术研究所,沈阳 110168;沈阳中科数控技术股份有限公司,沈阳 110168
- 2. 中国科学院沈阳计算技术研究所,沈阳 110168;中国科学院大学,北京 100049
- 折叠
摘要
数控加工程序通常由计算机辅助制造系统生成,以微小直线段的形式"以直代曲"来指导数控机床进行直线插补运动.随着工艺复杂度和精度要求的提高,数控加工程序的数据量急剧增加,不仅增加了数据存储和传输的难度,而且会引起机床执行过程中速度的频繁调整.针对以上问题,提出了一种融合深度学习的改进Douglas-Peucker三维数控加工轨迹压缩方法,该方法通过引入曲率和距离容差度的超参数考虑了加工轨迹中数据点序列的几何特性,并通过深度神经网络模型动态地优化算法中的超参数,从而实现更高的压缩效率.此外,算法中利用了 KD树结构优化误差计算,确保压缩后的数据能够在给定的公差范围内精确呈现原始数据的特性.实验表明,该算法可大幅减少数据量,并确保压缩后的数据准确呈现原始数据的特性.
Abstract
Numerical control machining programs are typically generated by computer-aided manufacturing systems,guiding numerical control machine tools in linear interpolation motion by approximating curves with tiny linear segments.With the rising complexity and precision demands of the process,the volume of data in numerical control machining programs has seen a drastic increase.This not on-ly complicates data storage and transmission but also leads to frequent speed adjustments during the operation of the machine tool.In response to these challenges,an improved Douglas-Peucker three-dimensional numerical control machining trajectory compression method that integrates deep learning has been proposed.This method takes into account the geometric characteristics of the data point sequence in the machining trajectory by introducing hyperparameters of curvature and distance tolerance.It dynamically optimizes the hyperparameters within the algorithm using a deep neural network model,thereby achieving a higher compression efficiency.Addition-ally,the algorithm employs a KD tree structure to refine error calculations,ensuring that the compressed data can precisely represent the characteristics of the original data within a given tolerance range.Experiments have demonstrated that this algorithm can signifi-cantly reduce the volume of data and guarantee that the compressed data accurately reflects the attributes of the original data.
关键词
Douglas-Peucker算法/轨迹压缩/轮廓误差/深度神经网络/参数优化Key words
Douglas-Peucker algorithm/trajectory compression/contour error/deep neural network/parameter optimization引用本文复制引用
出版年
2025
NETL
NSTL
万方数据