基于支持向量数据描述的线性轮廓监控研究

摘要

In manufacturing processes,there exist a kind of processes in which the output of the process has a linear relationship between one or more independent variables.The output of this kind of processes is known as linear profile.In this paper,a support vector data description ( SVDD) based control chart is pro-posed for monitoring the linear profiles.Furthermore,the effects of the parameters on the SVDD classifier are analyzed and the method of control chart limit determination is presented.Simulation results show that the proposed SVDD-based control has better performance than the T2 control chart for detecting shifts in inter-cept and residuals.Howerver,the T2 chart is better than the proposed control chart when there are shifts in the slope of the linear profile only.%在制造过程中,存在一类过程输出与一个或多个独立变量之间有线性函数关系的情况,称为线性轮廓( Linear profile).针对线性轮廓控制的问题,提出了基于支持向量数据描述( Support Vector Data Description,SVDD)的线性轮廓控制图,并分析了SVDD参数对分类器性能的影响及控制图参数的确定方法.仿真结果表明,基于SVDD的线性轮廓控制图在监控截距和残差变异时比T2 控制图性能更好,而监控斜率的变异时,T2 控制图性能更好.