SME manufacturers operating thermal process equipment face significant energy inefficiencies caused by gradual degradation of critical systems, including combustion drift, boiler heat-transfer fouling, plate heat exchanger scaling, and steam trap failure. These largely invisible issues can result in annual fuel losses of €14,000–€25,000 per facility. Existing predictive maintenance solutions are often unsuitable for SMEs, requiring costly sensor infrastructure, extensive historical failure datasets, and annual software costs exceeding €30,000.
This project develops and commercialises a physics-informed Bayesian Influence Diagram comprising a 38-node probabilistic graphical model that identifies the causes of excess energy consumption without requiring additional sensors or historical failure data. By combining first-principles thermodynamic models with readily available operational data—including fuel consumption, production volumes, and maintenance records—the system compares expected and actual energy performance, attributes inefficiencies to likely degradation mechanisms, detects emerging issues such as steam trap failure, and generates cost-weighted maintenance recommendations.
The approach has already been validated using 146,101 litres of annual fuel consumption data from Glenisk Co-operative and 352 maintenance records, where the model identified a 42% fuel reduction following six targeted maintenance interventions. Over 12 months, this project will advance the technology to TRL 6–7 through multi-site validation, commercial deployment preparation, and the establishment of a spin-out opportunity. The resulting platform will provide SMEs with an affordable, sensor-free route to reduce energy costs, improve operational efficiency, and accelerate industrial decarbonisation.
