In order to solve the problem that the workpiece machining progress information mainly relied on manual statistics incurring frequent errors in case of mixed category machining, a kind of workpiece online identification method based on SVM algorithm and power information was proposed. At first, through analyzing characteristics of spindle power variation during machining process of workpiece, a group of power curve time domain statistical features, and machining power characteristic parameters were selected to constitute characteristic vector for identifying workpiece. Then characteristic vector samples of various workpieces waiting to be processed were acquired by trial machining, and these vector samples were trained to build up classifiers, which were related to a type of workpiece respectively. During identification stage, characteristic vector of workpiece under machining should be extracted in real time and then compared with each classifier. According to the value of decision-making function, the type of workpiece shall be identified. The test results show that the identification method is of good robustness and generalization and its performance is superior to BP neural network and a template matching algorithm.

job shop, machining, workpiece identification, motor power, support vector machine(SVM)

针对目前作业型车间中工件加工进度信息主要靠人工统计，且在混类加工时经常出现统计出错的问题，提出一种结合工件加工功率信息特征分析及支持向量机(support vector machine，SVM)分类的工件在线识别和统计方法。该方法通过分析工件加工过程功率变化特征，利用一组功率曲线时域统计参数及加工功率特征参数组成的特征向量来区分加工工件，对新工件进行试加工以获得多种工件的特征向量样本，对样本数据进行训练从而得到与工件类别一一对应的工件识别分类器。在线加工时，实时提取加工工件的特征向量并与各分类器进行对比，根据决策函数值即可识别该工件的类型。实验结果表明该方法具有较好的鲁棒性及较强的泛化能力，识别性能优于BP神经网络和一种基于模板匹配的算法。

作业车间, 机械加工, 工件识别, 机床功率, 支持向量机