Depaola, MatteoMatteoDepaolaDe Gregoriis, DanielDanielDe GregoriisBornoff, RobinRobinBornoffVandevelde, BartBartVandeveldeMoens, DavidDavidMoens2026-11-172026-11-172026979-8-3315-6250-22833-8553https://imec-publications.be/handle/20.500.12860/60837Digital Twins (DT) enable Thermal Virtual Sensing in electronics, to support e.g. thermal management strategies. A physics-based DT may consist of an Augmented Kalman Filter (AKF) employing a thermal Reduced Order Model. The AKF estimation performance depends on the tuning of the process noise covariance matrix. Many strategies, such as the L-curve method, have been introduced to address the problem of tuning. However, these approaches typically are time consuming, scaling along with the model size, and rely on expertise to be correctly used and interpreted. In this work, a new method is presented to tackle the tuning more time-efficiently and without requiring a-priori expertise. The proposed method is tested on the thermal model of a printed circuit board assembly (PCBA) system with multiple unknown Quantities of Interest. Results indicate similar estimation performance when compared to alternative tuning methods, whilst achieving a speed-up with factors ranging from 59 to 5000 to perform the tuning itself. This approach enables a time-efficient and expertise-free tuning of AKF, paving the way for a broader and faster industrial adoption of DTs.engEfficient Augmented Kalman Filter Parameter Tuning for Thermal Virtual Sensing in Electronic SystemsProceedings paper10.1109/eurosime69483.2026.11511902WOS:001900864700021