Rod–airfoil configurations are commonly used to study turbulence–structure interaction noise, which is a major contributor to aerodynamic noise in various engineering applications. In this work, noise mitigation by regular and irregular leading-edge serrations is investigated for a rod–NACA0012–airfoil configuration using a direct hybrid CFD/CAA method. Large eddy simulations (LESs) are performed for the turbulent flow, and the acoustic perturbation equations (APEs) are solved to determine the acoustic field. The numerical results are validated by experimental data. The NACA0012 airfoil has different serrated leading edges, i.e., serrations with constant (regular) and varying (irregular) spanwise distributions of the amplitudes. Irregular spanwise amplitude distributions are generated by increasing or decreasing the amplitude of every second serration. The findings show an overall noise reduction for all configurations with regular and irregular serrations. The highest noise reductions are achieved by the irregular leading-edge serration configurations A3W2 and A6W2, with reductions of up to 11 dB in the Strouhal number range of $S_t$ = 0.25–0.75. Regular serration designs with greater amplitudes (A1W1 and A1W2) outperform other regular serration designs, reducing noise levels by up to 10 dB. For irregular configurations, the level of mitigation is found to be correlated with the base amplitude of their regular serrated counterparts.

This study investigates the influence of mesh resolution and turbulence model selection on the accuracy of numerical simulations for transonic flow, with particular emphasis on shock-boundary layer interaction phenomena. Accurate prediction of such flows is notoriously difficult due to the sensitivity to near-wall resolution, global mesh density, and turbulence model assumptions, and this problem motivates the present work. Two solvers were employed, rhoCentralFoam (unsteady) and TSLAeroFoam (steady-state), both are compressible and density-based and implemented within the OpenFOAM framework. The investigation focuses on three different non-dimensional wall distance (y+) values of 1, 2.5 and 5, each implemented with both moderate and fine mesh resolutions. Three turbulence models—Spalart-Allmaras (SA), k-ω Shear Stress Transport (SST), and k-ε‌ Realizable—were evaluated at M = 0.74, Re = 2.7 × 106, and α = 3.19°. Results showed that while both solvers achieved good overall agreement with experimental data, particularly in terms of pressure distribution, lift coefficient, and shock location, noticeable differences still emerged. The k-ω SST model consistently delivered the most robust performance across all cases, capturing the shock position on meshes with deviations below 0.02 compared to the experiment, and maintaining accuracy even at y+ ≈ 5. The k-ε‌ Realizable model was highly sensitive to near-wall resolution, displacing shocks downstream at higher y+ values, whereas Spalart-Allmaras remained broadly comparable to the k-ω SST model in predictive performance. The rhoCentralFoam solver achieved consistently better lift predictions, staying within about 2% of the experimental value on average, whereas TSLAeroFoam overpredicted it by around 4%. For transonic Reynolds-Averaged Navier-Stokes (RANS) simulations, unsteady k-ω SST with y+ ≈ 1 is recommended for maximum fidelity, whereas steady k-ω SST or SA simulations offer a practical option for quick and reasonably accurate aerodynamic predictions.

Air source heat pumps (ASHPs) have become a promising alternative to conventional heating and cooling systems, making accurate performance prediction increasingly important. This study presents a comparative analysis of Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) models for evaluating the ASHP performance under varying ambient conditions, examining the symmetry or asymmetry of prediction behavior across cold and hot regimes. Two experimental campaigns were carried out in a controlled climate room: the first primarily covering moderate to high temperatures ($-3 °\mathrm{C}$ to $36 °\mathrm{C}$), and the second mainly covering negative and low ambient temperatures ($-16 °\mathrm{C}$ to $18 °\mathrm{C}$). Performance data were collected to capture system behavior under diverse thermal conditions, making predictions more challenging. Both models were optimized, ANFIS through grid partitioning and ANN via architecture selection. Results demonstrate that ANN models achieved a superior overall accuracy, with mean absolute errors of 0.061 to 0.064 for cold and hot ambient conditions, respectively, showing a particularly strong performance under cold conditions. ANFIS demonstrated remarkable robustness in low-temperature predictions, maintaining less than 3% deviation across variations in water inlet temperature. Both approaches revealed temperature-dependent characteristics: cold-condition modeling required more complex architectures but yielded higher precision, whereas warm-condition modeling performed reliably with simpler configurations but showed slightly reduced accuracy.