摘要
提出了一种适用于模式识别的新型神经网络模型———局部有监督特征映射 (RegionalSupervisedFeatureMapping, RSFM)网络,将其应用到质量控制图的模式识别中,为基于统计过程控制(SPC)的智能工序质量诊断分析系统提供了技术支持。文中研究了网络的基本性能并对其参数进行优化,提出了采用欧氏距离判别法作为混合型多特征异常模式的识别方法。实验证明,所提出的模型对控制图的基本模式和混合型多特征异常模式都能够有效识别,网络收敛速度快、识别精度高,可进行大样本训练,适用于控制图的在线实时模式识别。
Pattern recognition of abnormal control charts can provide clues to reveal potential quality problems in manufacturing process. It has been a necessary technology to realize the automatic recognition of abnormal patterns with the need of automation and intelligence of statistical process control (SPC). In this paper, a new neural network model named regional supervised feature mapping (RSFM) network is proposed to recognize the control chart patterns in the intelligent diagnosis system for machining quality. The performance of network is studied, and its parameters are optimized. Euclid distance discriminance is developed to recognize mixed abnormal patterns. Numerical simulation results show that this model possesses many advantages, such as quick training and good recognition performance, which can be used in an on line real time mode.
出处
《机械科学与技术》
CSCD
北大核心
2005年第1期30-34,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
航空科学基金项目(04J18012)资助
关键词
人工神经网络
智能诊断
控制图
模式识别
Artificial neural network
Intelligent diagnosis
Control charts
Pattern recognition
