A novel prediction model of heat transfer characteristics of helically coiled tube heat exchanger using Dual Optimized LSTM (DOL) regression

Abstract

The investigation and examination of heat transfer and fluid flow behavior in micro device configurations is a highly pertinent subject in contemporary research. This is due to the increasing applications of micro devices, which require small size and high efficiency. Surface effects play a major role in micro scale applications, especially in heat transfer. The superior heat transfer characteristics of helical coiled tubes have led to their widespread adoption in industry. Moreover, altering fluid thermal conductivity offers a viable passive technique for augmenting heat transfer rates. Studies have revealed that the addition of nanoparticles to traditional heat transfer fluids significantly boosts their thermal conductivity, resulting in improved heat transfer performance. This work introduces a Dual Optimized LSTM (DOL) regression algorithm to predict heat transfer properties and nusselt numbers based on particle mass flow rate and Dean number. The efficiency of the proposed DOL prediction method is evaluated in a simulation environment using water and Al₂O₃ nanofluid flowing through a circular tube with uniform heat flux. The simulation results are analyzed for temperature-dependent characteristics, including nusselt number, friction factor, pressure drop, convective heat transfer coefficient, and inner nusselt number. The performance of the DOL regression model is evaluated under both laminar and turbulent flow conditions of nanofluids, with key metrics derived for each regime. The model's accuracy is demonstrated by a low RMSE of 0.00221 between experimental and predicted data, achieving a prediction accuracy of 99.5 %.