(33b) On-the-Fly Root Cause Analysis in Multi-Unit Processes: A Machine Learning Approach with Granger Causality and Fault Trees (Application: Heat Recovery Network)

Energy-intensive industries (EIIs) are heavily dependent on the efficient functioning of interconnected, multi-unit processes. Pinpointing the root cause of performance issues in these intricate systems is vital for reducing energy usage and enhancing production. Despite the growing popularity of machine learning techniques for fault diagnosis, their broad implementation is hindered by challenges such as data expertise requirements and adaptability to real-world scenarios. A novel, multi-faceted framework for root cause diagnosis in EIIs is introduced. This framework harnesses the power of explainable machine learning, Granger causality analysis, and engineering insights via fault tree analysis. This synergistic approach allows operators to swiftly identify the root cause of anomalies without the need for extensive retraining of machine learning models. The efficacy of the proposed method in diagnosing root causes in complex industrial processes is demonstrated through a case study.