ML-Driven optimization of two-phase microfluidic cooling using acoustofluidic bubble actuation and nanoarray-coated...

Area of Science:

• Microfluidics
• Thermal Management
• Nanotechnology

Background:

• Conventional passive cooling methods face limitations in heat dissipation for high-power electronics.
• Effective two-phase microfluidic cooling requires precise control over boiling dynamics and surface interactions.

Purpose of the Study:

• To develop and optimize a novel two-phase microfluidic cooling strategy.
• To investigate the impact of acoustofluidic bubble activation and nanoarray-coated micropins on thermal performance.
• To utilize machine learning for performance prediction and optimization.

Main Methods:

• Integration of acoustofluidic bubble activation with nanoarray-coated micropin structures.
• Application of deep neural networks (DNN), Long Short-Term Memory (LSTM), and statistical correlation (Spearman, Kendall) for data analysis.
• Utilizing SHAP (DeepSHAP) and partial dependence plots (PDP) for model interpretability.

Main Results:

• The LSTM model demonstrated superior prediction accuracy (lower MAE, SMAPE, RMSE) compared to the DNN.
• Initial temperature and chipset material were identified as key factors influencing the heat transfer coefficient (HTC).
• Acoustofluidic excitation was the primary positive contributor to thermal performance.

Conclusions:

• The proposed microfluidic cooling strategy significantly enhances heat transfer.
• Machine learning, particularly LSTM, provides accurate performance prediction and aids in understanding influential parameters.
• This integrated approach offers a promising solution for advanced thermal management challenges.