Model calibration relevant (meta )data

exp. data (model error?)

• sensors (with spatial positions, correlations, sensor accuracy)
• setup (e.g. compression test on concrete, tensile test of steel)
  • material
  • geometry
• access information (publicly available?)
• store query, version of data
• store result of query ?
• preprocessing steps, pipeline (where to put this, workflow?)

model (forward)

• parameters (values or just types?, (deterministic, stochastich, inference?)
• simulation procedure
  • FEM, CFD,
    • constititutive model (plasticity, fatigue)
    • discretization in space and time
    • inital / boundary conditions
• software
  • url/git
  • version
  • maintainer, company

inference

• parameters (initial values, priors)
• likelihood
  • model error
  • noise model
• inference scheme (MCMC, Gibbs, Variational Bayes, deterministic)
  • parameters (num_samples, burn_in)

results

• (posterior) parameters (samples, distribution - common information)
• applicability of model / model parameters (e.g. temperature, strain range, ...) Difficult to describe?
• metrics describing the quality of model calibration
  • model evidence
  • average/root mean square error (different experiments/sensor types)
• similar results with other models (e.g. Bayes factor)

How was the complete workflow setup (computational workflow?)