Novel hybrid neuro-fuzzy model to anticipate the heat transfer in a heat exchanger equipped with a new type of self-rotating tube insert

In this investigation, a combination of the wingsuit flying search (WFS) and teaching-learning-based optimization (TLBO) algorithms is developed as a new combinatorial optimization algorithm. The proposed combinatorial algorithm is tested over some well-known benchmark functions and then integrated with the artificial neural network (ANN) to construct a novel hybrid model. After that, the obtained hybrid model is employed to anticipate the experimentally obtained values of the average Nusselt number (Nu), average friction coefficient (f) as well as thermal-hydraulic performance ratio (η), in a heat exchanger equipped with a new type of self-rotating tube insert, against governing parameters. The insert is placed in the tube side of the water heater to heat natural gas. The proposed insert consists of various numbers of self-rotating modules. Indeed, the rotating insert is introduced to create effective secondary sweeping flow on the inner side of the tube. Since this type of tube insert simultaneously provides heat transfer enhancement and undesired pressure drop, a thermal-hydraulic performance ratio is defined to consider both of them. The governing parameters are the number of inserts (0 ≤ N ≤ 30), reservoir's temperature (40 °C ≤ T_R ≤ 50 °C) as well as Reynolds number (6 × 10³ ≤ Re ≤ 18 × 10³). It was found that the WFS-TLBO enhances the effectiveness of the main ANN in anticipating the Nusselt number (Nu), average friction coefficient (f) as well as performance ratio (η). Moreover, introducing the WFS-TLBO algorithm into the neural network provides an enhancement in the effectiveness of the hybrid models based on the single WFS and TLBO algorithms in anticipating the same parameters.