Abstract
Performance of the heat exchanger with micro dimensions is influenced by variety of factors that have been evaluated in different experimental studies. Similar to other thermal devices, it is possible to propose model for performance characteristics of micro heat exchangers by means of intelligent methods. The aim of the present study is development of models based on intelligent techniques for estimation of Nusselt number in a micro heat exchanger using hybrid nanofluid. Furthermore, sensitivity analysis is implemented to assess the importance of considered inputs in the models. In this study, experimental data from another research work on the nanofluidic micro heat exchanger is used to propose model based on multilayer perceptron (MLP) and Group Method of Data Handling (GMDH) neural networks. Comparison between the outputs of the model and data provided in the experimental study reveals great exactness of the proposed models. It is found that the exactness of the model and errors in the prediction are dependent on the applied configuration and algorithm. R2 of the models by use of GMDH and MLP are 0.9948 and 0.9972, respectively. Average Absolute Relative Deviation (AARD) of the models based on the noted approaches are 2.43% and 1.56%, respectively. In addition to modeling, sensitivity analysis is carried out and it is observed that concentration of nanomaterial, TiO2-ZnO, in the working fluid has higher importance than Reynolds number and height of plate in the value of Nusselt number.