Modern digital twin applications in science and engineering commonly utilize synergies between physics-based and data-driven modelling to connect physical assets with corresponding digital models, while additionally integrating sensor data or other observations. In this context, methods and tools stemming from scientific machine learning and uncertainty quantification become instrumental for the realization of digital twins. On the one hand, scientific machine learning is key for the development of innovative hybrid models, data-driven yet physics-conforming, to meet the stringent requirements of digital twins in terms of predictive accuracy and estimation speed. On the other hand, uncertainty quantification allows to assess the impact of ever-present uncertainties in modelling assumptions, physical parameter values, and data quality, and complement digital twin estimates with critical robustness and reliability metrics.

This mini-symposium aims to bring together researchers and practitioners working on the utilization of scientific machine learning and uncertainty quantification methods – and the combination thereof – for the purpose of digital twinning. We welcome contributions concerned with novel methodological developments, case studies, practical applications, and emerging trends.