부산대학교 도서관

The design of soft sensors for industrial processes requires knowledge of time-delays, caused either by mass transport phenomena or by information acquisition. The authors previously introduced a method for estimating finite-time delays through multiple correlation analysis, utilizing data obtained from historical databases. To validate its applicability in real industrial scenarios, it is imperative to assess the method's robustness against noise. In a previous paper, data from a sulfur recovery unit. were used to assess the suitability of the procedure. Here, the same dataset was utilized, and the input variables were intentionally corrupted by additive noise with various distributions and levels. An investigation was conducted to evaluate the impact of additive noise on the finite-time delay estimation. The results demonstrate the robustness of the proposed method to additive noise, when considering Gaussian or uniform distributions.