Multi-objective optimization and artificial neural network models for enhancing the overall performance of a microchannel heat sink with fins inspired Tesla valve profile

Microchannel heat sinks (MCHSs) are advanced cooling systems used in various applications such as electronics cooling, thermal management of high-power microprocessors, and microelectronic devices. These heat sinks are designed to efficiently dissipate heat generated by electronic components to maintain optimal operating temperatures and ensure device reliability. Employing appropriate fins within MCHSs is a crucial strategy to enhance heat removal efficiency. Since few investigations have been performed on the simultaneous thermo-hydraulic features in microchannels with Tesla valves, this research addresses this gap by suggesting that the microchannels of a heat sink be reinforced with novel Tesla valve-shaped fins. Tesla valves are designed to offer different levels of resistance depending on the flow direction. This characteristic allows for selective control of pressure drop and heat transfer efficiency by simply switching the flow direction. Utilizing Tesla valve-shaped fins is an almost new approach that differentiates the current research from conventional designs. The advantage of this structure is that under different circumstances, depending on whether more heat transfer is required or less pressure drop (ΔP), the flow direction can be changed, and the maximum efficiency of the MCHS can be accessed. The other innovative aspect of this study lies in the integration of artificial neural network (ANN) models to optimize the efficiency of MCHSs equipped with Tesla valve-shaped fins. The simulations were conducted to characterize the flow and thermal features of the MCHS. Then, these data were used as the benchmark for training the ANN models. Four ANN models were introduced to anticipate the Nusselt number (Nu) and ΔP in each forward and reverse direction. The divider's acute angle (α), the divider's obtuse angle (β), and the divider's radius (R) were selected as the input data. Multi-objective optimization and genetic algorithms were used to achieve the optimal conditions. In the optimal thermal design (Setting 2: MCHS with fin variables of α = 60 β = 180, and R = 80 μm in reverse flow direction), the overall performance had a significant enhancement of 57.1 % compared to the device lacking fins. Furthermore, in Setting 2, the overall performance of the MCHS and its diodicity experienced notable improvements of approximately 12.1 % and 38.5 %, respectively. Implementing the parameters of Setting 5 (MCHS with fin variables of α = 60, β = 152.32, and R = 73.10 μm in forward direction) during fin production resulted in a notable enhancement of about 41.1 % in the device's overall performance.