Comparative study of control charts for residuals

Authors

  • Danilo Cuzzuol Pedrini
  • Carla Schwengber ten Caten

DOI:

https://doi.org/10.15675/gepros.v4i4.508

Abstract

Control charts (CCs) are among the most popular statistical tools for process monitoring in industry. When implementing CCs in practice, it is necessary to validate two basic assumptions: (i) data samples collected from the process are independent and (ii) identically distributed. If the data sampled are autocorrelated, the first assumption is violated and the control chart will yield a large number of false alarms. When the variable is dependent on process parameters, and these parameters vary during process execution, the reference model for the process is not the same for all data samples emerging from the process. In the first case, the usual approach is to monitor the residuals from a time series model adjusted to process data. For the second case, a regression model relating the dependent variable to process parameter is adjusted, followed by a control chart for the residuals. The present paper thus aims to compare the performance of these approaches in a same data sample. A flowchart is introduced to orient the application of these control charts. Key words: Control Charts; Residuals; Regression; ARIMA.

Published

2008-12-01

How to Cite

Cuzzuol Pedrini, D., & ten Caten , C. S. (2008). Comparative study of control charts for residuals. Revista Gestão Da Produção Operações E Sistemas, 4(4), 123. https://doi.org/10.15675/gepros.v4i4.508

Issue

Section

Articles