Variable Time Delay Estimation for Industrial Processes: Robustness to Noise

Estimating finite-time delays, whether stemming from mass transport phenomena or data acquisition, is essential for soft sensor design. Previously, the authors introduced a method named time variant multiple correlation delay estimation (TV-MCDE), designed to estimate time variant delays in dynamic processes. To validate the applicability of TV-MCDE in real-world industrial settings, robustness against additive noise is assessed in this paper on a real-world case-study. Data from a debutanizer distillation column are considered, and the method's robustness was evaluated by intentionally corrupting input variables with additive noise of various distributions and intensities. The results provide insight into the method's resilience to additive noise under Gaussian or uniform distributions.