Abstract: To improve the automation of anomaly detection in machining process,the original datasetsbased on mathematic description of control chart were constructed by Monte-Carlo method,a novel method based on neighborhood rough set was introduced to reduce the control chart time domain features,and an abnormal pattern recognition model of control chart based on support vector machine was proposed.Through the simulation experiment,the main identification model parameters were optimized with genetic algorithm,and the recognition accuracy of different kernel functions and classification models were analyzed and compared.The effectiveness and availability of the proposed model were verified with the use of actual production data.

Key words: control chart pattern, support vector machines, time domain features, neighborhood rough set, genetic algorithms

摘要： 为提高加工过程异常模式检测的自动化程度,在建立控制图数学描述的基础上,利用蒙特卡洛法构建了控制图数据集,研究了基于邻域粗糙集的控制图时域特征约简方法,提出了基于支持向量机的控制图异常模式识别模型。通过仿真实验,使用遗传算法优化了异常识别模型的主要参数,并对不同核函数、不同分类模型的识别精度进行了分析与对比。通过实际生产数据测试验证了所构建模型的有效性与可用性。

关键词: 控制图模式, 支持向量机, 时域特征, 邻域粗糙集, 遗传算法