The Research On Change-point Identification In Nonparametric Situation In Phase I Based On Density Analysis

Statistical process control is an important method monitoring of multiple variables and control chart is one of the main tools to implement of SPC and MSPC. Control charts can be divided into two phases: phaseⅠ, the baseline period and phaseⅡ. In phaseⅠ, the observations are expressed as a single mode of operation. However, there may exist in more than one mode of operation. The existence of multiple operating modes brings some difficulties to SPC implementation. So it is unreasonable to monitor all modes of operation in a single statistical model. Thus, in baseline period, it is necessary to identify the implicit mode of operation of historical data, and filter data on this basic to extract the historical data in different operating modes and establish the control chart. Then we can ensure that the monitoring of process variation is effective.Different modes of operation reflect the diversity of the data distribution. There may contain a variety of different forms of data distribution in the baseline data. In this operating mode we cannot give an assumption of in-control distribution. Therefore, we must identify the quantity of operating modes and times without any assumption of data distribution. That is a solution of non-parametric multivariate problem in baseline period.In this paper, we will study on the analysis method under several changes in the process of multi-mode operation using the K-S test and kernel density estimation, which are important tools of statistical methods, without any assumption of the distribution. The problem in this paper came from the actual needs of the scientific issues, research in the chemical, semiconductor, food and other industrial production system. So it has broad application prospects. This study can enrich the baseline period study in the non-parametric multivariate distribution and improve the SPC theoretical system and provide a new basis for the theoretical study.