论文部分内容阅读
采用模式识别方法,从切削过程的动态切削力和振动信号中抽取特征,对刀具的磨损状态进行了判别。通过时间序列分析建立反映切削状态的数学模型,从动态数据中凝聚信息,构成用于判别的特征向量。在分类器的设计方面,采用了在近邻分类法基础上的三种改进算法:编辑技术、边界抽取和边界补缀。采用上述方法处理的浓缩样本集,其识别率接近大样本集的 1-NNR 的结果,分类速度约提高了6倍。可望用子对刀具磨损的在线监控。
The pattern recognition method is adopted to extract the features from the dynamic cutting force and vibration signals during the cutting process to judge the wear state of the tool. Through the time series analysis, a mathematical model reflecting the cutting state is established, and the information is condensed from the dynamic data to form the eigenvectors for discrimination. In the design of the classifier, three improved algorithms based on the nearest neighbor classification are adopted: editing technique, boundary extraction and boundary patch. Using the above method to process the concentrated sample set, the recognition rate is close to that of the large sample set 1-NNR, and the classification speed is increased by about 6 times. Expected to use the child on the tool wear online monitoring.