Geometric Properties of Partial Least Squares Regression for Application to Process Monitoring

GODOY, J.L.; VEGA, J.R.; MARCHETTI, J.L.

Congreso; III Congreso de Matemática Aplicada, Computacional e Industrial; 2011

A decomposition of the input and output variable spaces of an arbitrary stochastic process is obtained by partial least squares regression (PLSR) and their main geometric properties are derived. The proposed decomposition can be used for detecting and identifying faults or anomalies in complex processes due to the existing relationships between statistics on each determined measurement subspace and the fault type.