Global Modeling of a Textile Process

The latest enhancements in off-line quality control deal with the possible functional relationship between the variability of the process response and the regressors ( exper imental variables controlled by the experimenter and measured with negligible error). In the past, the analysis of experimental data generally focused on means or location parameters, and variability was only taken into account to apply generalized least squares. Currently, however, the dispersion parameters are estimated using a least squares analysis of the logarithm of the sum of the squared deviations from the within- replications mean. The adopted dispersion model is used thereafter to reduce the location model to homoscedasticity to obtain a global model, which will help to as certain the conditions that make the best of a process, i.e. minimum variance, closeness to target, and robustness to noise.