Comparing performances of clements, box-cox, Johnson methods with weibull distributions for assessing process capability

Ozlem Senvar, Bahar Sennaroglu

Abstract


Purpose: This study examines Clements’ Approach (CA), Box-Cox transformation (BCT), and Johnson transformation (JT) methods for process capability assessments through Weibull-distributed data with different parameters to figure out the effects of the tail behaviours on process capability and compares their estimation performances in terms of accuracy and precision.

Design/methodology/approach: Usage of process performance index (PPI) Ppu is handled for process capability analysis (PCA) because the comparison issues are performed through generating Weibull data without subgroups. Box plots, descriptive statistics, the root-mean-square deviation (RMSD), which is used as a measure of error, and a radar chart are utilized all together for evaluating the performances of the methods. In addition, the bias of the estimated values is important as the efficiency measured by the mean square error. In this regard, Relative Bias (RB) and the Relative Root Mean Square Error (RRMSE) are also considered.

Findings: The results reveal that the performance of a method is dependent on its capability to fit the tail behavior of the Weibull distribution and on targeted values of the PPIs. It is observed that the effect of tail behavior is more significant when the process is more capable.

Research limitations/implications: Some other methods such as Weighted Variance method, which also give good results, were also conducted. However, we later realized that it would be confusing in terms of comparison issues between the methods for consistent interpretations.

Practical implications: Weibull distribution covers a wide class of non-normal processes due to its capability to yield a variety of distinct curves based on its parameters. Weibull distributions are known to have significantly different tail behaviors, which greatly affects the process capability. In quality and reliability applications, they are widely used for the analyses of failure data in order to understand how items are failing or failures being occurred. Many academicians prefer the estimation of long term variation for process capability calculations although Process Capability Indices (PCIs) Cp and Cpk are widely used in literature. On the other hand, in industry, especially in automotive industry, the PPIs Pp and Ppk are used for the second type of estimations.

Originality/value: Performance comparisons are performed through generating Weibull data without subgroups and for this reason, process performance indices (PPIs) are executed for computing process capability rather than process capability indices (PCIs). Box plots, descriptive statistics, the root-mean-square deviation (RMSD), which is used as a measure of error, and a radar chart are utilized all together for evaluating the performances of the methods. In addition, the bias of the estimated values is important as the efficiency measured by the mean square error. In this regard, Relative Bias (RB) and the Relative Root Mean Square Error (RRMSE) are also considered. To the best of our knowledge, all these issues including of execution of PPIs are performed all together for the first time in the literature.


Keywords


Process performance indices (PPIs), process capability indices (PCIs), process capability analysis (PCA), non-normal processes

Full Text:

PDF


DOI: http://dx.doi.org/10.3926/jiem.1703


Licencia de Creative Commons 

This work is licensed under a Creative Commons Attribution 4.0 International License

Journal of Industrial Engineering and Management, 2008-2019

Online ISSN: 2013-0953; Print ISSN: 2013-8423; Online DL: B-28744-2008

Publisher: OmniaScience