A data-driven approach of quantifying function couplings and identifying paths towards emerging hazards in complex systems

• DOI: 10.1016/j.psep.2021.04.037
• Publication Date: 2021-04-29T23:30:19Z
• Authors: Mengxi Yu; Madhav Erraguntla; Noor Quddus; Costas Kravaris
• Keywords: Lift (data mining); Complex system

Abstract Hazardous scenarios emerging from complex system where number of functions are large and corresponding function coupling are humongous, are very difficult if not impossible to identify humanly. Today’s complex systems generate a very large dataset every minute and dynamic nature of the generated data makes it difficult to track such couplings. The Functional Resonance Analysis Method (FRAM) got success in recent years to understand hazards emerging from function couplings in complex systems, however, challenges remain to estimate aggregated couplings appropriately without quantitative analysis. The current study developed a data-driven approach to quantify function couplings using lift confidence intervals of association rules. Later, association rules were merged to identify the paths leading to a potential hazardous scenario. The paths were presented graphically and equipped with quantified coupling information and capable of providing guidance to prevent the emerging hazard scenario. The approach has been demonstrated with a case study of a polymerization process in process industry for which function couplings are represented by a very large dataset.