Multiple Correlation Analysis for Finite-Time Delay Estimation for Soft Sensor Design in the Presence of Noise

The design of soft sensors (SSs) for industrial processes requires knowledge of time delays, caused either by mass transport phenomena or by information acquisition. A method based on multiple correlation analysis to estimate the finite-time delay was previously proposed by the authors. The method is based on processing data obtained from historical databases. Characterization of the robustness of the method to noise is performed here as a mandatory step to validate its suitability before it can be applied to real case studies in industrial settings. Data from an industrial database were exploited to create a set of synthetic case studies characterized by different nonlinearities. The input and output signals were corrupted by additive noise with different distributions and different levels. A systematic study of the effects of additive noise on finitetime delay estimation was performed. Finally, the analysis wasextended to an industrial case study. The robustness of the proposed method to additive noises confirms its suitability for finite-time delay estimation in industrial applications.