Functional data modeling and monitoring applied to particleboard manufacturing

Density of particleboards strongly affects its functional properties. In this paper we show that appropriate statistical methods aimed at monitoring the entire VDP shape, instead of some specific features, can improve the overall quality of the produced boards. The proposed method is based on recent literature on profile monitoring, which suggests to firstly model the profile via functional data and then design appropriate multivariate control charts for monitoring coefficients of the functional data model. The model combines BSpline functions as regressors and a small-scale model for dealing with the autocorrelated errors. Alternative strategies for placing the Spline knots are made. A performance comparison study based on real data is carried out to show that the proposed approach outperforms traditional methods considered in the industrial practice.