Model-based experimental analysis of enhanced boiling heat transfer by micro-nano porous surfaces

Abstract In recent years, numerous studies have reported that the heat transfer performance can be enhanced by micro-nano structures during boiling. However, their detailed mechanisms at all boiling states are still not fully understood. In this work, four samples of honeycomb micro-nano porous copper surfaces with successively increasing pore sizes are prepared at different electrodeposited current densities. Both pool boiling experiments and numerical simulations are used to study their heat transfer performances. The obtained boiling test results indicate that the four samples have superior heat transfer performance than that of the conventional plain copper surface. Particularly, Sample#M2 prepared at a deposition current density of 1 A/cm2 has shown the overall best heat dissipation capacity, with maximum critical heat flux and heat transfer coefficient reaching 197.6 W/cm2 and 25.3 W/cm2/K, respectively. Besides the experimental studies, realistic numerical simulations are conducted. The proposed mathematical model considers three-dimensional (3D) complex geometries of the four samples reconstructed by a computed tomography (CT) scanning technique and real effective thermal conductivities caused by different porosities in the samples. It is observed that the thickness of the four prepared samples gradually increases as the electrodeposition current density increases, which directly leads to higher heat dissipation through the side surfaces. By considering this effect, a parameterized mathematical model of heat fluxes through three-phase contact lines (TPCL) on the boiling surfaces as the Neumann boundary conditions is developed to mimic the physical reality as much as possible, and the appropriate model parameter C can be estimated by minimizing the difference of measured experimental wall superheats and model estimates. Considering all the above modeling aspects, the solutions of a series of well-posed 3D transient heat conduction problems at two representative boiling states are studied. Numerical simulation results show that the percentage of heat flowing out of the TPCL to the entire heated bottom surface at high heat-flux densities can increase to 90%, indicating that the strong heat dissipation capacity of the micro-nano porous copper surfaces can be achieved. Subject to the 3D transient heat-conduction forward model developed in this work, a multi-layer end-to-end convolutional neural network is for the first time constructed and trained to solve the corresponding ill-posed inverse problems of 3D transient heat conduction efficiently. The ill-posed inverse problems correspond to the estimation of unknown heat fluxes highly variable in time and space on the complex boiling surface structure from available temperature data on the heated bottom surface. Different from previous classical Tikhonov-regularization-based inversion approaches that in general require much computational effort, this new neural-network-based solution method provides a promising alternative way to develop a highly efficient temperature-to-heat-flux soft sensor technique for both boiling applications and many other similar engineering problems.