Passive Heat Transfer Augmentation in Low-Temperature Organic Rankine Cycle Systems Using Twisted Tape Inserts: Impact on Working Fluid Flow Behavior and Power Output

INTRODUCTION: The Organic Rankine Cycle (ORC) systems are considered to be the best for low-grade heat recovery. However, limited studies have integrated passive heat-transfer enhancement, deep learning prediction, and ORC optimisation within a unified framework under varying operating conditions. OBJECTIVES: Here, an integrated framework based on Deep Neurone Networks (DNNs) is proposed which includes prediction and optimization to model the non-linear relationships with great precision and thereby improve the entire ORC system performance. METHODS: The DNN was trained on the simulation data generated from different combinations of twist ratios, heat inputs, flow rates, and thermodynamic conditions. RESULTS: The DNNs predicted key thermal-hydraulic and thermodynamic performance indicators with high accuracy under varying ORC operating conditions. This model exhibited excellent predictive capability as indicated by R² values surpassing 0.98 for all the outputs, and very low errors like MAE = 0.0027 and RMSE = 0.0036 for the thermal performance factor prediction. CONCLUSION: Further optimization revealed major improvements, resulting in Nusselt number of 769.108, thermal performance factor of 1.6452, and net power output of 4.672 kW under the best conditions. Additionally, thermal and exergy efficiencies grew to 6.961% and 25.160% respectively indicating that heat transfer and energy utilization have become more effective. In summary, the suggested framework is a reliable and precise tool for assessing and optimizing ORC systems, thereby providing better insights into performance for thermal system design and operational decision-making.