摘要
提出了控制图模式识别的基本框架,描述了控制图异常状态的三种形式,即基本模式、特殊模式和混合模式。针对特殊模式和混合模式,提出了将输入数据经小波分解后的近似系数与各层细节系数的能量成分组成的特征向量作为概率神经网络的输入进行控制图模式识别的方法。仿真实验结果表明,该方法结构简单、收敛速度快、识别精度高、Ⅰ型错判和Ⅱ型错判低,适合于控制图模式识别。
The general framework of control chart pattern recognition was presented.Three types of control chart's unusual patterns(general patterns,special patterns and mix patterns)were de- scribed.As for special patterns and mix patterns,the combined wavelet transform with probabilistic neural network(PNN)was proposed.Input data was decomposed by wavelet transform into several detail co- efficients and approximations.The approximation obtained and energy of every lever detail coefficients was for the inputs of PNN.The simulation results show that the performance of the proposed method has many ad- vantages,such as simple structure,quick convergence,high aggregate classification rate,low type I error and typeⅡerror,which can be used in control chart pattern recognition.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2006年第S2期135-139,共5页
China Mechanical Engineering
基金
福建省教育厅A类科研基金(JA03114S)
关键词
小波变换
概率神经网络
控制图
模式识别
wavelet transform
probabilistic neural network
control chart
pattern recognition
