In Honor of Nobel Laureate Prof. Ferid Murad

Abstract Submission Open! About 500 abstracts submitted from about 60 countries

Featuring 9 Nobel Laureates and other Distinguished Guests

Abstract Submission

Leo Eskin

Snofox Sciences, Inc.

Using Physics-based Modeling (digital Twin) Methods And Machine Learning To Improve Energy Efficiency And Reduce Maintenance For The Global Cold Chain
Mauntz International Symposium (7th Intl. Symp. on Sustainable Energy Production: Fossil; Renewables; Nuclear; Waste handling , processing, & storage for all energy production technologies; Energy conservation)

Back to Plenary Lectures »


The global refrigeration industry (cold chain) encompasses a wide range of disciplines, including the food sector, where temperature-controlled warehouses, trucks and shipping containers maintain food safety, and the healthcare industry, where refrigeration preserves medicines and pharmaceuticals, including vaccines. It is estimated that the refrigeration sector consumes approximately seventeen percent of the global electricity production [1] and this is expected to grow in the coming years due to global warming.
Significant performance enhancements, reduction in energy consumption and greenhouse gas emission, and improved maintenance intervals can be achieved by using physics-based thermodynamic modeling methods [2-5] to develop a digital twin for a range of industrial refrigeration systems. Implementations have been demonstrated for stand-alone, single-loop commercial vapor compression refrigeration systems (refrigerators or commercial cooling units) and for multi-loop, multi-compressor industrial refrigeration systems used in temperature-controlled warehouses up to several hundred thousand square feet in size. Such digital twins enable real-time performance monitoring by computing mass- and energy-balances using measured data, and the calculated results can be trended and used by machine learning algorithms to identify common equipment failures and alert personnel to operational problems.
Examples are presented illustrating how the trended calculated results enable root-cause identification of a 40+% cooling capacity reduction, and a machine learning algorithm is presented demonstrating highly (98+%) accurate identification of the eight most common refrigeration system failure modes.