International Journal of Mechanical and Thermal Engineering : 2026 - 7(1)

Experimental analysis of an impinging circular jet on a flat plate

B Praneeth Kumar — Thu,1 Jan 2026

B Praneeth Kumar, B Venkata Sai Raghu Vamsi, D Vasubabu, J Rama Krishna, G Sahil

International Journal of Mechanical and Thermal Engineering 2026 7(1):166-171

Owing to its ability to provide high-intensity thermal exchange over concentrated surfaces, jet impingement is extensively employed in thermal management and industrial processing. This study experimentally investigates the thermodynamic behavior of a single circular air jet impinging orthogonally on a flat surface. The influence of two key parameters normalized nozzle-to-plate distance (H/D), adjusted between 2 and 8, and Reynolds number (Re), ranging from 41,440.8 to 66,107.5 on the convection coefficient and dimensionless temperature is evaluated. To ensure measurement accuracy, thermocouples were calibrated, and an error analysis was conducted initially. Results indicate that both the peak convection coefficient and the dimensionless temperature increase with higher Reynolds number, while both parameters decrease as H/D increases. Additionally, it was observed that at Re values above 57,227.4 and H/D ratios greater than 6, the variation in maximum dimensionless temperature significantly diminishes, suggesting that the plate cooling become independent of H/D.

Optimization and prediction of PETG properties in FDM using Taguchi method

T Venkata Naga Suresh — Thu,1 Jan 2026

T Venkata Naga Suresh, D Tarun, P Pujeth Sai Naga Vasu, M Venkata Basava Poornaiah and P Soma Shankar

International Journal of Mechanical and Thermal Engineering 2026 7(1):172-180

This research focuses on the optimization of Fused Deposition Modeling (FDM) process parameters to enhance the mechanical properties and surface quality of Polyethylene Terephthalate Glycol (PETG) components. Key parameters including layer height, infill density, infill pattern, and print speed were systematically analyzed using the Taguchi L27 orthogonal array to evaluate their impact on flexural strength and surface roughness. Experimental results revealed that the infill pattern is the most significant factor influencing both mechanical performance and surface finish. Specifically, a maximum flexural strength of 106.08 MPa was achieved under optimized conditions, while surface roughness (Rz) values ranged from 2107.7 to 6490.3. Statistical analysis through ANOVA and the development of linear regression models validated the experimental findings, providing a predictive framework for material behavior. This study demonstrates that the Taguchi method is an effective approach for improving the structural integrity and reliability of PETG parts for industrial applications.

Comparative Evaluation of Machine Learning Methods for Predicting Thermal Performance of Air-Cooled Heat Sinks: SVR, KRR, ANN, and Random Forest

Khemraj Beragi — Thu,1 Jan 2026

Khemraj Beragi

International Journal of Mechanical and Thermal Engineering 2026 7(1):181-184

Getting accurate predictions of how well a heat sink performs thermally is quite important in the design of modern electronic cooling systems. In this study, we looked at four commonly used machine learning (ML) approaches — Support Vector Regression (SVR), Kernel Ridge Regression (KRR), Artificial Neural Network (ANN), and Random Forest Regression (RFR) — and checked how well each of them can estimate heat transfer coefficients (HTCs) in air-cooled heat sinks with rectangular channels. We used three-dimensional CFD simulations to build our dataset, covering six different heat sink designs under laminar flow conditions up to Re = 2200. About 83% of the data went into training, and the remaining portion was kept aside for testing. Our results show that all four methods worked quite well. SVR turned out to be the most accurate, with a peak error of just ±1.4%. KRR came next at ±1.9%, followed by ANN at ±2.5% and RFR at ±3.1%. We also ran a detailed sensitivity study on the hyperparameters of each model and talked about how much computing power each method needs. The findings here suggest that ML-based surrogate models are a practical and fast option compared to running full experiments or detailed physics-based simulations every time.