The growing demand for renewable energy has amplified the need for efficient and reliable wind turbine technologies, where understanding aerodynamic performance and aeroelastic behavior plays a critical role. In this study, a high-fidelity computational fluid dynamics (CFD) model was developed to analyze the aerodynamic loads and structural responses of a 2 kW horizontal-axis wind turbine, while an artificial neural network (ANN) was trained using CFD-generated data to predict power output and aeroelastic characteristics. The work combines ANN predictions and CFD simulations to determine the feasibility of machine learning as a surrogate model, which is much less expensive in terms of computational costs and time, with no negative effects on the accuracy. Findings indicate ANN predictions are closely comparable to CFD results with under 5–7% deviation at optimal blade pitch angles, which was shown to be very reliable in capturing nonlinear aerodynamic trends at different wind speeds and blade pitch angles. In addition, the obtained result emphasizes the example of the trade-off between aerodynamic efficiency and structural safety, where the largest power coefficient (0.42) was achieved at $\left(0^{°}\right)$ pitch and the tip deflections were reduced by almost 60% as the pitch was raised to $\left(5^{°}\right)$. Such results substantiate the usefulness of ANN-based methods in the rapid aerodynamic and aeroelastic simulation of wind turbines and provide a prospective direction for effectively designed wind power generation and optimization.

* Grammatical corrections: There are some grammatical errors in the article. The file with the name " 8UR337_Grammatical corrections " that is attached shows the location of the errors and provides linguistic corrections. Use “Foxit Phantom PDF” to...

The present study explores combined free-forced convective flow and entropy generation in a constant-volume double lid-driven trapezoidal cavity. All configurations of the isosceles trapezoidal cavity were meticulously designed to possess identical leg lengths and constant volume, ensuring that the same amount of heat is transferred from the cavity’s legs. The cavity has left and right lid-driven walls capable of oscillating upward and downward, while all other domain boundaries remain stationary. The left wall is sustained at a consistently high temperature, whereas the right wall is kept at a stable low temperature, and the upper and lower horizontal walls are thermally insulated. The modelling of this problem was carried out based on the finite volume technique. The obtained results were carefully validated against existing literature related to similar problems. The influence of relevant parameters such as Richardson number (0.01 ≤ Ri ≤ 100), aspect ratio (0.4 ≤ AR ≤ 1) and three distinct moving arrangements (Case-A, Case-B and Case-C) were examined. The findings revealed that heat transport was restricted at high Ri for all the presented aspect ratios, especially for Case-C. For all the presented aspect ratios and cases, entropy generation decreases as Ri increases, with the lowest values observed for Case-A. Trapezoidal cavities with AR = 0.4, 0.6, and 0.8 generate lower entropy than the square cavity at high Ri, but higher entropy at low Ri.