Prediction of thermo-hydraulic properties of flow in an innovative plate heat exchanger using machine learning algorithms

Reducing fuel consumption and toxic gas emissions is a major concern in modern energy research. This paper investigates the performance and heat transfer enhancement of an innovative plate heat exchanger (IPHE) using machine learning techniques. By optimizing the geometric parameters of the plate, we predict thermohydraulic characteristics-represented by the Nusselt number (Nu), coefficient of friction (f), and performance (P) within the Reynolds number range of 500-5000 based on numerical modeling data. This study addresses the need for improved efficiency in plate heat exchangers (PHEs) amid rising energy demands and environmental concerns. Traditional methods like numerical simulations or costly experiments have limitations, prompting interest in artificial intelligence (AI) and machine learning (ML) for thermal analysis and property prediction in PHEs. Various ML models, including Decision Trees, XGBoost, Gradient Boosting, and ensemble methods, are evaluated in predicting f, Nu, and overall performance (P). Our comprehensive experimentation and analysis identify top-performing models with robust predictive capabilities. For f, the highest R2 score was 0.98, indicating excellent prediction accuracy, with mean squared error (MSE) values consistently below 0.0016. Similarly, for Nu and P, top models achieved R2 scores of 0.979 and 0.9628, respectively, with MSE values below 0.0347 and 0.05. These results highlight the effectiveness of machine learning techniques in accurately predicting thermohydraulic properties and optimizing PHE performance.