Fusion- Materials Qualification Lecture 2 Flashcards
What does a credibility case require?
Verification, validation and uncertainty quantification (VVUQ)
What is verification and give an example?
The process of checking that the mathematical and/or computational implementation of a model is correct. Is the code bug free and mathematically correct?
Example is a mesh refinement study
What is validation and give an example?
The process of checking that the model agrees with the physical world with respect to the question being asked. In structural analysis the question is normally “will my component fail by mechanism X?”
Example is comparing measured experimental strain/temperature with model
Ideal and real validation dataset characteristics
Ideal: full 3D data of temperature and strain throughout a component, with uncertainty bounds.
Real: sparse data, combination of full-field (surface) and point sensors throughout a component, different uncertainties for different sensors.
What often causes differences between a simulation and experiment?
Difference in assumptions
What assumptions for a monoblock simulation might be problematic?
Geometry: as manufactured geometry not matching ideal CAD model.
Boundary conditions and load: heat flux distribution might not be uniform in experiment.
Material properties: neglecting temperature dependence or plasticity.
Physical modelling: assuming small deformation or one way coupling.
What is HIVE?
Heating by induction to validate extremes. Is a high heat flux test rig for in-vessel components (e.g divertors)
What is uncertainty quantification?
The science of quantifying, characterising, tracing and managing uncertainty in computational and physical systems
Why do we need uncertainty quantification and what are the two main types of uncertainty?
Allows for quantitative validation of simulations with experimental data.
Main types are systematic and random errors
What do systemic errors do and what are they related to?
Create an offset or bias in our measured results.
Related to the precision of our measurement and is often hidden
What do random errors do and what are they related to?
Crest a random shift in our measured values.
Related to the accuracy of our measurement.
Deterministic vs probabilistic simulations
Deterministic: single input and pass/fail, assume worst case scenario for input parameters (e.g minimum measured yield stress), gives very conservative design.
Probabilistic: the future, probabilistic input and failure probability, accounts for uncertainty in input parameters, distribution of output parameters and a probability of failure.
When is a model credible?
If we can rely on it in the decision making progress to answer an engineering question.
If the adverse outcome of a wrong decision is significant then increased evidence (e.g experiments+VVUQ) needed for model credibility and vice versa.
What will fusion component qualification be like?
Digital and probabilistic. Simulations will be used to answer engineering questions. Simulations will be supported by experimental data and an associated credibility case.
What is CHIMERA?
Combined heating and magnetic research apparatus. Components in vacuum up to size of the ITER test blanket module. Can use surface heating, internal heating, central magnetic field, pulse magnet (simulates plasma disruption), water cooling at high pressures.
