Rozvoj přístupů k analýze způsobilosti procesů

Abstract

In the area of process capability analysis, failure to meet the data normality assumption is considered a non-standard situation. This disertation focuses on defining and understanding the causes of normality violations in the monitored characteristics, which is essential for quality planning and control, process improvement, and determining the appropriate procedure for process capability analysis. Based on the thorough review of research articles and a systematic investigation of the causes of violations of the assumption of normality in manufacturing processes, four main categories of non-normality have been identified that reflect different aspects of the occurrence of non-normality. The individual causes of normality violations in these categories are then analyzed in detail in terms of their identification, manifestation in the data, and impact on the results of the process capability analysis. Based on this analysis, a systematic procedure is proposed that allows the selection of an adequate capability analysis method depending on the nature of the violation of the normality assumption, thereby supporting the correct interpretation of results and qualified decision-making in the area of quality management. The identified causes of non-normality are devided into four categories – data characteristics, data structure, data collection methodology, and measurement systems. Process and technological aspects also play a significant role, as they can lead to multimodal distributions, asymmetry, or outliers. These causes manifest themselves in the data in different ways and have different impacts on the accuracy and reliability of statistical analyses. For selected causes of non-normality, procedures have been proposed and verified using both simulated data and real-world data from the field of mechanical engineering.