Optimal design of container heat dissipation safety based on five-block method and radial basis function

Abstract

The ability of the ocean to provide a good thermal environment and not take up land space resources has led to subsea data centres due to their energy efficiency, low failure rates and green sustainability. As a result, there is a growing need for subsea data centres to be structurally sound. This paper investigates the optimal design of data centre capacity and cooling structures to improve their thermal safety performance to a certain extent.Firstly, the temperature quantification of the heat dissipation and heat generation processes is carried out using Newton’s cooling theorem and Fourier’s law of thermal conductivity to provide maximum and minimum constraints for optimization. Then, the " five-block method" is used to optimize the arrangement of internal servers with temperature constraints, considering the problem of packing same-sized items in closed spaces. Based on the quantification of the solid-liquid and solid-solid heat transfer processes and the local temperature analysis as well as the optimization of the boxing, the model yields a maximum of 384 servers for the data center.Secondly, the fin structure is then further optimized for heat dissipation. An approximate model of the fin structure and thermal performance is developed and solved by a radial basis function (RBF) neural network algorithm. Then, based on 100 sets of parameters for the serrated fin structure, the RBF approximation model is used to optimize the range of values for height, pitch and thickness. Based on the three ranges of parameters, Monte Carlo simulations are used to generate different combinations of parameters for different shapes, and to obtain the exact parameter that will increase the surface area by 13.8%. In the case of the first task, a solution is obtained to store 396 servers.