Applications

We are working to quantify uncertainty in algorithms for highly coupled, multi-physics, and multi-scale exascale simulations of tokamak plasma, which is vital for advancing fusion knowledge and designing future tokamaks.

Through collaborations with the UCL Met Office Academic Partnership, SEAVEA's UQ methods are applied across various modelling domains including ocean, sea-ice, paleoclimate, climate change, and atmospheric turbulence. Our work also extends to Data Assimilation and Machine Learning for nowcasting, including deployments for the Air Quality Unified Model (AQUM).

Working with the DDWG “Turbulence at the Exascale,” SEAVEA demonstrates how VVUQ methods can quantify uncertainty in turbulence modelling, particularly in scenarios where existing models perform poorly and in studying complex transitions.

SEAVEA collaborates with the DDWG “Systems Engineering” to investigate the reliability of the scalable SCEMa multiscale application, which links molecular interactions to engineering-scale properties like strength and toughness. We apply semi-intrusive VVUQ to manage how uncertainties at one scale influence behaviour at another.

Essential for regulatory approvals, VVUQ procedures are applied using UKCOMES’ HemeLB code to create personalised human-scale models for high-resolution 3D blood flow simulation. The SEAVEA toolkit quantifies the impact of changes in physiological and numerical modelling parameters on blood flow throughout the body.

The IMPECCABLE approach, which combines molecular dynamics (MD) based ligand-protein binding free energy with machine learning (ML), is being enhanced with built-in uncertainty quantification. This ensures actionable rankings of compounds and includes research into uncertainty propagation between different models.

Brunel University of London (BUL) uses agent-based models like Flee and Homecoming to forecast the arrival and return of forcibly displaced people. BUL has used SEAVEAtk to reduce the cost of sensitivity analysis (SA) for this application, which resulted in collaborations with NGOs and a UNHCR-led migration forecasting policy brief.

BUL's Flu And Coronavirus Simulator (FACS) code, which mimics the spread of the virus, is used to perform SA. This helps quantify forecast uncertainties and understand how geographical region choice affects input parameter sensitivities. FACS was one of the foreasting  codes used during the COVID-19 pandemic to inform strategic decision making for several NHS hospitals in London.

Researchers at Universita della Svizzera Italiana (USI) have used SEAVEAtk to perform sensitivity analysis for electricity network models that have inherent uncertainties in supply, demand, and costs. The work conducted on this application led to the development of sensitvity analysis support for correlated parameters in EasyVVUQ.