Research on heat transfer model prediction of Tesla Valve heat sink based on training neural network method

Yan-Xiao Jia, Yan-Zuo Chang, Guo-Xing Yang, Ruo-Yu Yang, Yi-wei Zhang, Ting-Hao Zhang

Tesla valve heat sink, neural network, CFD numerical simulation.

With the increasing power density of electronic equipment, heat dissipation technology has become the key to ensure the stable operation of equipment. Because of its unique structural design, the Tesla valve heat sink shows great potential in the heat dissipation of high-power electronic devices. However, the traditional heat transfer model prediction method has the problems of complex calculation and low efficiency. The purpose of this study is to explore a method of heat transfer model prediction based on training neural network to improve the accuracy and efficiency of heat transfer efficiency prediction of Tesla valve heat sink. The heat transfer data of Tesla valve heat sink under different structures were collected by numerical simulation. The data were then used to train a feed forward neural network. Through a lot of training and verification, the neural network model shows good generalization ability and can accurately predict the heat transfer efficiency under unknown conditions. In this study, the effects of network structure, training algorithm and optimization strategy on model performance are discussed, and an improved network architecture is proposed to improve the accuracy of prediction. Finally, the advantages of the proposed method in computational efficiency and prediction accuracy are verified by comparison with traditional methods.

Received: 12 Aug 2024, Receive in revised form: 12 Sep 2024, Accepted: 19 Sep 2024, Available online: 23 Sep 2024

10.22161/ijaers.119.3