Thermal performance investigation of the miniature revolving heat pipes using artificial neural networks and genetic algorithms

Abstract

The miniature revolving heat pipes (MRVHPs) are promising candidates for the cooling structure design of rotating machines, however, the relation between the thermal performance of MRVHPs and their thermophysical properties is still absent. Therefore, MRVHPs are tested in various operational conditions firstly. And then, a power-law empirical correlation is developed based on experimental data, Ku, Ja, Pr, Bo and Fr are determined as main effective dimensionless parameters, while the structural parameters of MRVHPs Di/R, Le/Lc, Le/Leff and filling ratio φ are considered as well. However, the prediction accuracy is not satisfactory enough. Considering ANNs have been widely used in varied applications and demonstrated to be particularly credible in system modeling and identification, therefore, a BPNN model which parameterized by GA is developed for thermal performance prediction of the MRVHPs. The experimental data are divided into training data set and testing data set. Ja, Pr, Bo, Fr and φ are regarded as inputs, while Ku is output. The results show that the established GA-BPNN model could predict thermal performance of MRVHPs with a very good accuracy. Comparing with the predicted results of semi-empirical correlation, the square of correlation coefficient (R2) is increased by 13.265%. Meanwhile, the evaluation method for the optimal filling ratio of the MRVHPs under specified working conditions is developed and an acceptable accuracy is obtained.