Application of Statistical Methods : Challenges Related to Continuous Industrial Processes

Abstract: For decades, many efficient statistical improvement methods have been available to improve the quality of processes and products. Statistical process control (SPC), process capability analysis (CA), and design of experiments (DoE) are among the most powerful process monitoring and problem-solving methods in the quality engineering toolbox. SPC and CA are methods that are more directed toward monitoring existing processes and assessing their capability related to customer requirements, while DoE typically is used to improve products and processes. It is increasingly difficult to understand and control industrial processes and products because of the increasing complexity of technical systems. Among the complications for statistical analysis of measurements in continuous industrial processes are the multitude of variables and the combination of high-frequency sampling of the measurement systems and process dynamics. Therefore, in industry today, process data are often multivariate as well as autocorrelated (i.e., dependent in time).The purpose of this research is to support the application of SPC, CA, and DoE. More specifically, the aims of this research are: [1] to analyze the use, and related barriers, of SPC, CA, and DoE in organizations; [2] to provide guidance in selection of appropriate decision methods for Cpk when data are autocorrelated; and [3] to adapt methods for analyzing designed experiments to manage dynamic process behavior and autocorrelation in continuous processes.The main contribution of this research is that it explicitly illustrates and describes special considerations and problems that can be encountered when planning, conducting, and analyzing real experiments in continuous industrial processes. Other contributions of this research are: the practical use and development of adapted analysis procedures for experiments in continuous processes; the presentation of comparative data that helps in the selection of decision methods for Cpk when data are autocorrelated; and the analysis of barriers that hinder the use of statistical methods in Swedish organizations.