摘要
在分析比较目前常用的质量辅助诊断方法局限性的基础上,提出了一种基于多分类支持向量机(SVM)的质量控制图智能诊断新方法。该方法以SVM技术为智能核心,较好地解决小样本学习问题,避免了人工神经网络等智能方法在对小批量生产过程质量诊断时所表示出的过学习、泛化能力弱等缺点。另一方面,通过结合投票法和决策树的基本思想,所提方法拓展出对控制图混合型异常模式的识别能力,从而提高了对质量过程诊断的全面性和准确性。与其它几种常见人工智能方法质量诊断的效果进行对比,实验表明,所提方法容易实现、诊断精度高,为实现小批量加工过程的在线质量诊断与控制提供可行的思路。
After analysis and comparison of the limitation of assistant quality diagnosis methods used in practice,a novel intelligent diagnosis method for quality control chart was proposed based on multi-class support vector machine（MSVM）.This method takes SVM technology as the intelligent core,which solves the small sample learning problems well.And it can overcome the shortcoming of over-fitting,longtime training and frail generalization power of artificial neural network（ANN） in the small-batch production process quality diagnosis.Moreover,combining the basic thought of voting method and decision tree,the presented method has the capability to recognize mixed abnormal pattern exists on the control chart and improves the accuracy and universality of the process quality diagnosis.Simulation experimental results were given to demonstrate that,in comparison with other intelligent methods,the presented one is convenient in operation and higher in accuracy.So,it provides a better way for the small-batch production process quality online diagnosis and control.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第6期1689-1692,1696,共5页
Journal of System Simulation
基金
江西省科技项目(2007328)
国家自然科学基金(50705039)
关键词
小批量
质量诊断
多分类支持向量机
决策树
small-batch
quality diagnosis
multi-class support vector machine（MSVM）
decision tree
作者简介
吴德会（1975-），男，安徽嘉山县人，博士后，兼任九江学院电子工程系副主任，副教授，研究方向为智能质量控制，智能测试技术。
