Inverse conjugate mixed convection in a vertical substrate with protruding heat sources: a combined experimental and numerical study

Abstract

heat inputs, CFD simulations were again run to compare the temperature distribution at the back of the PCB board with measurements. List of symbolsa Absorption coefficient (m −1 ) A Area of heat source (m 2 ) B Distance from the edge of the plate to the nearest chip edge (m) C p Specific heat (J/kg k) D Spacing between the heat sources (m) d Height of the heat source (m) Gr Grashof number, Gr = gβ∆TL 3 υ 2 g Acceleration due to gravity (9.81 m/s 2 ) H Height of the duct/channel (m) h Convective heat transfer coefficient (W/m 2 K) I Total radiation intensity (W/m 3 ) k s Thermal conductivity of solid (W/m K) L Characteristic length (height of the heat source in this case) (m) N Length of substrate in Z direction (m) n Refractive index p Pressure (N/m 2 ) S Sum of the residues, as the case may be (Eq. 12) s Direction vector Q Power input (W) Q ANN Heat input value measured by ANN (W) Q Expt Heat input value measured by experiments (W) Re Reynolds number, U α L υ Ri Richardson number, Ri = Gr Re 2 S Spacing between the two walls of the channel (m) t Heat source thickness (m) T Temperature (K) U ∝ Inlet air velocity (m/s) V Velocity vector (m/s) Abstract This paper reports the results of a combined numerical and experimental study to estimate the heat inputs of three protruding heat sources of the same size placed on a vertically placed PCB board of height 150 mm, depth 250 mm, and thickness 5 mm. First, limited measurements of temperatures were recorded at eight locations along the height of the back of the PCB board for different (and known) values of heat inputs of the protruding heat sources and different velocities. These were followed by three-dimensional calculations of fluid flow and conjugate heat transfer for various heat transfer coefficients on the backside of the PCB board. The difference between the CFD predicted and experimentally measured temperature distributions on the back of the PCB board was minimized using least squares and the best value of heat transfer coefficient was obtained. Using this 'data assimilated' CFD model, detailed CFD simulations were done for various values of heat input values and Reynolds numbers (each of these can be different from one another) of the flow. The temperatures at the same eight locations at the back of the PCB board were noted. An artificial neural network was then developed with ten inputs (eight temperatures together with the input velocity and the ambient temperature) to estimate the three outputs (three heat inputs) after carrying out extensive studies on the architecture of the network. This inverse solution was then tested with experiments for validating the ANN approach to solve the inverse conjugate heat transfer problem. Finally, with the ANN estimated

Comparison study of CFD and artificial neural networks in predicting temperature fields induced by natural convention in a square enclosure

“…Their study showed that the ANN was able to accurately predict temperature trend of air-flow and plate. Ahamad and Balaji [21] developed an ANN to solve the inverse conjugate heat transfer problem, i. e. the heat source was predicted with trained temperature distribution by ANN. They validated the efficiency and accuracy of the ANN by comparing ANN predictions with experimental data.…”