Inferential sensor – Based adaptive principal components analysis for mechanical properties prediction and evaluation

Abstract

This paper is concerned with a method for on-line quality prediction and evaluation of mechanical properties in metal testing. This method uses an Adaptive Principal Component Analysis (APCA) as a multi predictor of different sub-models defining the mechanical properties such as constraints limits and elongation. The PCA technique, characterized by its multivariate component, is strongly recommended to model a multi-input-multi-output system. The complex system is generally known as a non-linear and unsteady state process. The PCA method is a linear projection. To adapt it and to improve the prediction accuracy, a variant of this method is considered based on iteratively using a specific algorithm. This kind of approaches is applied for constructing an inferential model, which allows a reliable and accurate predictor. Simulation results, based on the measured and computed data using the above-cited method, show that the proposed approach is easily implementable and give an accurate prediction.