DATA-DRIVEN COMPUTATIONAL INVESTIGATION OF FLUID FLOW AND HEAT-TRANSFER CHARACTERISTICS IN AN IN LINE TUBE BUNDLE UNDER PULSATING CROSS-FLOW

Abstract

Reciprocating pulsating flow offers a controllable approach for improving convective heat transfer in tube-bundle systems. This study combined secondary experimental data with computational modelling to evaluate 180 regimes covering 18 frequency-amplitude-duty-cycle combinations across ten Reynolds-Prandtl operating blocks. Factorial ANCOVA, HC3 robust inference, uncertainty-weighted regression, leave-one-block-out validation, dimensionless correlation development, bootstrap analysis, and multi-response optimization were applied. Pulsating flow improved heat transfer in 95.56% of the regimes, while 81.67% achieved at least 10% enhancement. Mean Nusselt number and heat-transfer enhancement ratio were 58.87 and 1.247, respectively. Frequency and amplitude ratio were the strongest controllable factors, while a duty cycle of 0.2 performed better than 0.5. Blocked validation yielded R2values of 0.9779 for Nusselt number and 0.7782 for enhancement, with corresponding mean absolute percentage errors of 2.56% and 4.76%. Strouhal-based correlations achieved blocked R2values of 0.9593 and 0.8431. The strongest tested condition, f = 0.450Hz, A/D = 15, and ψ = 0.2, produced Nu = 71.63and Nup/Nust = 1.520, approximately 52.7% above the low-pulsation reference. Its selection across all validation folds and 189 sensitivity scenarios confirmed robust thermal optimality within the measured domain.