Process Capability Analysis for Non-normal Processes with Lower Specification Limits

Acknowledgements I am heartily thankful to my advisor Hendry Raharjo, PhD for his invaluable guidance. His attitudes of motivation and his insightful comments helped me throughout my study. Although it was fatiguing, I must admit that I enjoyed what I did. I would also like to thank Prof. Bo Bergman. Without his brilliant idea, this study would not have been possible. Last but not least, my sincere gratitude and deepest love to my beloved mother and father for being supportive at every step I take; to my dear Başak for making me laugh when I am too tired to smile; to Doruk and Ebru for their logistics support; to Necla and Anders Bey for their help and for making me feel that I am not alone; to my Aunt Hülya whose strong will is a source of inspiration to me; and to Ahu for being always there regardless of the distance between us. Abstract Process capability analysis is an important element of any quality improvement initiative. However, estimating process capability is often problematic when it comes to non-normal distributions since the conventional methods sometimes give misleading results. In this Master's thesis, a new method for estimating process capability of non-normal distribution with only lower specification limit is proposed. The proposed method only considers the left tail of the distribution rather than taking all of the data points. For estimating the process capability, it employs least squares technique and normal approximation to the selected observations from the left tail. Furthermore, the proposed method is tested on log-normal, 3-parameter Weibull, and 3-parameter Gamma distributions. Simulations and real-world data analysis are used for verification and validation purpose, respectively. An easy and practical guideline is also developed. Following the proposed guideline, capability of a non-normal process can be estimated with a relatively high success rate.