A transformer model integrated with fuzzy strategy for the design of a new four-Leaf-clover-inspired hybrid pin-fins microchannel heat sink

High local heat flux significantly reduces the reliability and lifespan of electronic chips. Most existing microchannel heat sinks (MCHSs) are single solid structures. Such a structure reduces heat transfer area and fails to generate sufficient fluid disturbance. Published surrogate models cannot effectively characterise the uncertainties in sample data. Therefore, four new hybrid pin-fin microchannel heat sinks (NHPFMCHSs) are proposed. The computational fluid dynamics (CFD) simulations show that all four NHPFMCHSs significantly reduce maximum and average temperatures compared with the non-hybrid straight microchannel heat sink (NHSMCHS). The four-leaf-clover-like Case 4 achieves the best comprehensive performance under different Reynolds numbers (Re). This model uses four-leaf-clover-inspired pin-fins in the central region to enhance heat dissipation and straight-fins at the inlet and outlet to reduce pressure drop (Δp). Then, this paper, for the first time, presents a coupling of an improved fuzzy Transformer model with an enhanced shuffled frog leaping algorithm for the optimisation of Case 4. A fuzzy strategy is adopted to identify the uncertainty characteristics in the samples. Compared with BP and ANN, the proposed improved fuzzy Transformer model reduces the RMSE for Tmax by 38.89% and 33.33%, and for Δp by 48.96% and 39.68%, respectively. Compared with NHSMCHS, further optimisation of Case 4 leads to scheme No.15, which reduces the average Tmax by 8.74 K and increases the average Performance Evaluation Criterion (PEC) by 1.41 for all Re values.