Application Research Of CUSUM Control Charts Using Normality Transformation

Statistical process control (SPC) are widely applied in industry for monitoring stabil-ity of process. Whereas, Conventional statistical process control charts (such as Shewhart, CUSUM, EWMA, etc.) are often based on the normality assumption. The process re-sponse distribution should be normally distributed. While the normality assumption is not always satisfied in practice. Applying conventional control charts in monitoring the non-normal process will lead to serious error. We try to define a function of random vari-able X which is itself a normally distributed, so use this function to transform non-normal data to normal data and create control charts.In this paper, we describe nine kinds of normality tests, i.e. We compare the power among these tests and also make some corresponding conclusions. Then, we define three kinds of normality transformations. We develop a quick and simple method to estimate the parameters of the function and also do some comparison among these transformations in the capability of transforming normality and process capability analysis. Finally, we propose three transformed CUSUM control charts and compare them with traditional CUSUM charts. Some advices are provided for practitioners on to create proper control charts with non-normal process.