A rapid reconstruction strategy of full-domain temperature field based on discrete measurement points for electronics thermal management

Dynamic thermal management of multi-heat-source systems increasingly relies on full-domain thermal analysis and evaluations in real-time. However, the conventional discrete measurements and reconstruction techniques both struggle to capture the global temperature field evolution due to the real-time uncertainty of operating conditions. In addition, temperature field inversion of heat-source systems in the small data regime is also a challenging problem to be solved in practical engineering systems. Therefore, a rapid reconstruction strategy based on a small dataset and sparse sensors is described, herein to monitor the full-domain thermal states online, irrespective of variable operating conditions. Specifically, by dimensionality reduction, a series of low-dimensional eigenvectors can be identified from a small high-fidelity dataset under diverse operating conditions, characterizing the most dominant spatial distribution and evolutionary patterns of the thermal field. Online reconstruction is driven by dynamically adjusting the eigenvector coefficients by minimizing the error between real-time measurements and predictions. The global state is further estimated via assembling the order-reduced eigenvectors in a specific formula. In addition, QR decomposition is integrated for robust reconstruction. The feasibility and potential of the reconstruction technique was proved by analytical nondimensionalized temperature models. Finally, an extensive evaluation was concluded by referring to the simulation cases of a distributed multi-heat-source system with anisotropic thermal conductivity and in variable environments. The reconstruction results demonstrate the effectiveness and fast dynamic response of this approach, which can facilitate the synchronized monitoring of the global thermal distribution and effectively assist in regulating internal heat transport.