Robust calibration strategy for selecting constitutive model and parameters from stress–strain data

Develop a robust calibration strategy that reliably determines both (i) which salt rock constitutive model to use and (ii) a representative parameter set for that model directly from stress–strain data, so that stable and accurate predictions can be made across laboratory datasets for salt rock mechanics.

Background

The paper emphasizes the complexity of salt rock behavior and the abundance of constitutive models with many parameters, which makes reliable calibration challenging. The authors note that, despite extensive literature, determining a single representative parameter set across datasets and identifying the most appropriate constitutive model from stress–strain data has been difficult.

They propose a multi-step calibration strategy using Particle Swarm Optimization and demonstrate it on synthetic data, but the abstract explicitly frames the general task of robust model-and-parameter identification from stress–strain data as an unsolved challenge, highlighting its importance for safety assessments and predictive modeling in salt caverns.