Abstract: To improve the accuracy and timeliness of product quality control in intelligent manufacturing processes,a control chart pattern recognition method based on parameter-free clustering and improved support vector machine multi feature fusion was proposed.The control chart simulation dataset was generated using Monte Carlo method,and the multiple quality feature mean levels were divided.Using parameter-free clustering to extract historical data information features,integrating them with statistical and shape features,and obtaining the optimal feature combination through cross experiments.The support vector machine classifier was improved with Beluga whale optimization algorithm for accurately and efficiently identifying abnormal patterns in control charts.Through simulation experiments,the recognition accuracy and efficiency of different classifiers on different complex datasets were compared.The results showed that the classifier was less affected on the dataset complexity,and the recognition accuracy even on complex datasets could maintain over 98.63%.At the same time,it had the advantages of fast training speed and low computational complexity.

Key words: control chart, pattern recognition, feature fusion, parameter-free clustering

摘要： 为提升智能制造中产品质量管控的准确性和及时性,提出一种基于无参数聚类和改进支持向量机多特征融合的控制图模式识别方法。采用蒙特卡洛法生成模拟数据集,考虑了质量特征均值微动的情况。将无参数聚类提取的历史数据信息特征,与统计特征以及形状特征进行融合,通过交叉实验获取最优特征组合。借助白鲸算法改进支持向量机分类器,实现对控制图异常模式的准确高效识别。通过仿真实验比较了不同分类器在不同数据集复杂程度下的识别准确性和效率,结果显示,所提出的分类模型对数据集复杂程度的影响较小,即使在复杂数据集上也能保持98.63%以上的识别精度,并具备训练速度快、计算复杂度低的优点。

关键词: 控制图, 模式识别, 特征融合, 无参数聚类