改进Alpha Shapes和快速凸壳算法的SVM故障诊断  

SVM Fault Diagnosis Based on Improved Alpha Shapes and Rapid Convex Hull Algorithm

作　　者：宋仁旺[1] 杨磊 余百千 石慧[1] 董增寿[1] SONG Renwang;YANG Lei;YU Baiqian;SHI Hui;DONG Zengshou(School of Electronic and Information Engineering,Taiyuan University of Science and Technology,Taiyuan Shanxi 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,山西太原030024

出　　处：《机床与液压》2023年第13期212-217,共6页Machine Tool & Hydraulics

基　　金：山西省自然科学基金(201901D111259);国家自然科学基金青年科学基金项目(61703297)。

摘　　要：现有的基于凸壳的支持向量机(SVM)算法处理机械装备产生的大规模原始数据时间太长。针对这一问题,通过结合轮廓提取算法(Alpha Shapes)和快速凸壳算法,提出一种结合改进快速凸壳算法的SVM用于故障诊断研究。该融合算法利用改进简化的Alpha Shapes算法提取点集的边界数据点,作为改进的快速凸壳算法的对象,减少凸壳算法递归的工作量。实验结果表明:该算法平均只提取了数据集0.26%的数据点,且计算的时间也相应降低。最后实验同样表明该算法的性能优于单一的SVM算法。The existing convex hull based support vector machine(SVM)algorithm has too long time in processing large-scale original data generated by mechanical equipment.To solve this problem,by combining Alpha Shapes algorithm and fast convex hull algorithm,an SVM combined with improved fast convex hull algorithm was proposed for fault diagnosis.In the fusion algorithm,the improved simplified Alpha Shapes algorithm was used to extract the boundary data points of the point set as the object of the improved fast convex hull algorithm to reduce the recursive workload of the convex hull algorithm.The experimental results show that the algorithm extracts only 0.26%of the data points of the data set on average,and the calculation time is reduced accordingly.Finally,the experimental results also show that the performance of the algorithm is better than a single SVM algorithm.

关 键 词：大规模数据 支持向量机 凸壳超平面 Alpha Shapes 故障诊断 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]