Thermal virtual sensing for electronic components and systems: exploring the robustness of Kalman filtering

Digital Twins (DTs) of Electronic Components and Systems (ECS) can be used for thermal analysis and design optimization. However, one of the main difficulties is identifying the physical parameters of the numerical model underlying the DT. Indeed, these parameters can vary across devices and even during their lifecycle. A methodology that can accommodate for uncertainties is therefore desired. This work presents a Kalman Filter (KF)-based approach employing a Reduced Order Model (ROM) of a considered ECS. The robustness of the KF with respect to uncertain parameter values is investigated by evaluating its capability to correctly estimate the ECS thermal state while relying on a model with perturbed parameter values or boundary conditions. Evaluation of the KF performance regarding dealing with parameter uncertainties indicate the effectiveness of the proposed approach.