Explainable machine learning approach for unveiling the influence of conical strip design on heat exchanger performance under laminar flow

Abstract

In this study, decision tree classification and SHAP explainable machine learning techniques were applied for the first time to analyze and optimize the thermal hydraulic performance of heat exchangers equipped with conical strip inserts operating under laminar flow. A dataset containing 400 rows of digitized experimental measurements from six published studies was constructed to determine the impact of various design parameters, such as geometry, arrangement, and number of conical strips (SN), on key performance indicators, including the Nusselt number (Nu), friction factor (f), and performance evaluation criterion (PEC). Results showed that a high Reynolds number (Re), a low pitch ratio (PR), and a high SN enhance the Nu, while a high Re, PR, and twist angle increase the f. When the strip-to-width ratio (SW) is less than 0.93, Re is between 550 and 1000, and PR is less than or equal to 1.75, the non-staggered arrangement of conical strips yields a geometry angle of 52.5 degrees or less, allowing a high Nu to be achieved. Additionally, a low f can be achieved by maintaining Re and PR above 325 and 2.26, respectively. By analyzing existing literature data, this study identified optimal ranges and influential features to enhance thermo-hydraulic performance, offering valuable insights for future heat exchanger design. Ultimately, the research contributes to achieving the Sustainable Development Goals by improving energy conversion and efficiency.