A Study on the Influence of Skewed Populations on Statistical and Neural Network Control Charts for Process Mean
Loading...
Date
2009-05
Authors
Yew Jin, Ooi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Shewhart control chart is an important statistical tool for detecting the presence of special cause
n the process. It is widely used in different fields of manufacturing processes and it is based on
he assumption that the means of samples which are drawn from the population in the process are
1ormally distributed. Some actions such as using Box-Cox power transformation and increasing
he sample size are used when the normality assumption is not valid in practice. In this
lissertation, the average run length (ARL) performance is compared among the neural network
based control chart, Shewhart control chart, EWMA control chart and CUSUM control chart by
Using the gamma distribution generated data in various skewnesses to measure the robustness of
he control charts. Dev C++ is used to develop and analyze the data in neural network based
:ontrol chart, while SAS is used to generate the gamma distributed data, develop and analyze the
lata in Shewhart, EWMA and CUSUM control charts. From the results obtained, the neural
Letwork based control chart is less robust compared to the statistical based control charts which
ncludes Shewhart X control chart, EWMA and CUSUM based control charts. Based on these
easons, we concluded that the EWMA based control chart is more robust to be applied when the
kewness is high in the data.
Description
Keywords
Skewed populations , Process mean , Neural network control