Statistical monitoring of multivariable dynamic processes with state‐space models

Abstract

Industrial continuous processes may have a large number of process variables and

A generic statistical methodology to predict the maximum pit depth of a localized corrosion process

“…Indeed, it should be noticed that the estimations of GLD parameters are less influenced by the central values than the other distribution estimations. Because of these various advantages, the family of four-parameter GLD have been used in many fields where accurate data modelling is required such as insurance and inventory management [30], finance [31,32], meteorology [30,33], pipeline leakages [30], statistical process control [34,35], independent component analysis [36,37], simulation of queue systems [38] or for generating random number [39]. For several years, the authors of the present paper have also developed the use of this versatile family of distributions in materials science [28,30,40] and statistical control process [27,41].…”

On the Run Length of a State-Space Control Chart for Multivariate Autocorrelated Data

“…To estimate state-space models in statistical process control, a number of approaches exist (Negiz and Cinar, 1997;Dorsey and Lee, 2003;Lee and Dorsey, 2004;Pan and Jarrett, 2004;Triantafyllopoulos, 2006;Xie et al, 2006;Jarrett and Pan, 2007a;Zantek et al, 2007), with different algorithms that provide different statistical behaviour.…”

“…In such a case, analysis of the residuals will reveal the "filtered" behaviour of the process, i.e., the behaviour of the non-foreseeable or innovational part, without going into an analysis of the "inertia factor". On a simultaneous basis, the works of Negiz and Cinar (1997), Dorsey and Lee (2003), Lee and Dorsey (2004), Pan and Jarrett (2004), Alfaro (2005), Triantafyllopoulos (2006), Xie et al (2006) and Zantek et al (2007) introduced multiple time series state-space modelling in statistical process control. This modelling allows to eliminate the autocorrelation existing in the data, preserving a great amount of information in the residuals, that in this methodology are denoted as innovations, and also allowing to describe the existence of inertia in the process by building a control chart for the states variables, estimate with the state-space model.…”

HACCP with multivariate process monitoring and fault diagnosis techniques: application to a food pasteurization process

“…In addition, data may have strong autocorrelation. In recent works, SPM based on state variables of a dynamic stochastic model of the process were developed from process data collected under normal operating conditions to deal with high autocorrelation and cross-correlation in process data (Negiz & Cinar, 1997;Norvilas et al, 2000). SPM of processes generating highly autocorrelated and cross-correlated data can be implemented by using the T 2 chart of canonical variate state space variables and the squared prediction error (SPE) chart.…”