Heat exchanger fouling determination using thermography combined with machine learning methods

Abstract

Disclosed is a methodology for determination and prediction of heat exchanger fouling, such as polymer fouling in the circulation loop that forms part of the heat exchanger system. The buildup of a polymer or other undesired material deposit in the heat exchanger provides a distinctive temperature signature (thermal gradient) on the surface of the heat exchanger asset, which is visualized using a thermographic camera. Coupling images (thermograms) from the camera with a machine learning algorithm identifies fouling and, with knowledge of the historical data of the asset and operating and ambient conditions, enables prediction of future fouling. The thermal images provide several types, or orders, of temperature information that are indicative of locations vulnerable to fouling. In one case, the method uses machine learning applied to time-based temperature change/gradient information to detect hidden polymer fouling in areas that form part of the heat exchanger asset.