一种基于改进成组算法的特征提取方法  

Feature extraction methodology based on improved grouping algorithm

作　　者：闫纪红[1] 李万召[1] 

机构地区：[1]哈尔滨工业大学工业工程系/现代生产技术中心,黑龙江哈尔滨150001

出　　处：《计算机集成制造系统》2009年第10期2023-2028,共6页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(70701012);黑龙江省自然科学基金资助项目(E200634);航天支撑技术基金资助项目~~

摘　　要：针对传统成组算法不能保证核元素为最优的缺点,提出了一种改进的成组算法,获得了最优的核元素,从而使每个分组中的元素最多,分组数达到最少。较之模糊聚类法,本算法更简单,运算速度更快,最终的分组数通过预先设定的相似度得到,更有利于处理未知信息。最后,采用刀具磨损实验中采集的振动信号对该方法进行验证,得到了比较理想的结果。Traditional group algorithm couldn＇t guarantee core element as optimal. To deal with this problem, an improved group algorithm was proposed. By this algorithm, optimum core elements were obtained. As a result, each group had the most elements satisfying similarity threshold consequently the number of groups reached the fewest. Compared to fuzzy clustering algorithm, the proposed method had better convergence time and simple structure. Vibration signals acquired from a cutting tool degradation experiment was used to evaluate this method, and satisfactory results were obtained.

关 键 词：成组算法 特征提取 核元素 非平稳信号 广义夹角余弦 刀具磨损 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]