IMECE 2025 revised Copy

This study uses hybrid intelligent algorithms to investigate the predictive modeling of oscillatory heat transfer coefficients (OHTC) in thermoacoustic refrigeration systems. Thermoacoustic refrigerators provide an environmentally friendly alternative to conventional cooling technologies; however, predicting OHTC is completely due to varying operational parameters such as pressure, frequency, and stack geometry. To address this, the research integrates Adaptive Neuro-Fuzzy Inference Systems (ANFIS) with Particle Swarm Optimization (PSO) and Genetic Algorithms (GA) to develop robust predictive models. A previously published experimental dataset comprising twenty-four samples was used to train and test various ANFIS configurations under different hyperparameter settings. Model performance was evaluated using R² as an indicator. The results reveal that ANFIS-PSO achieved the most consistent and reliable predictive performance across training and testing phases. The optimal ANFIS-PSO configuration (population size = 50, c1 = 1.5, c2 = 2.0) attained a training R² of 0.898 and a testing R² of 0.868, reflecting a well-balanced fit and minimal overfitting. While ANFIS-GA achieved a higher isolated testing R² in some instances, it exhibited greater variability, making it less suited for predictive applications. These findings demonstrate that ANFIS-PSO is better suited for practical predictive tasks in thermoacoustic refrigeration systems due to its synergism capability and model stability. The study contributes a validated framework for enhancing the performance of thermoacoustic cooling systems through intelligent, hyperparameter-optimized modeling.