numgeo-ACT¶

Automatic calibration of Karlsruhe fine sand with the Hypo-ISA model. Black lines are the numgeo simulations, coloured dashed lines the experiments; the objective function (top-left) decreases as the optimizer converges across oedometer, drained/undrained triaxial and several undrained cyclic triaxial tests.

What numgeo-ACT does¶

Calibrating an advanced constitutive model by hand is tedious: the models have many interacting parameters, and a parameter set that fits one test often degrades the fit of another. numgeo-ACT automates this process. You describe which model to calibrate, which parameters are free (and within which bounds), and which laboratory tests to honour. The optimizer then searches the parameter space, repeatedly simulating every test and scoring how closely the simulations match your data, until it converges on a single parameter set that represents the best compromise across all tests.

How it works¶

flowchart LR
    A[Excel database<br/>lab tests] --> B[excel.collect]
    B --> C[setup<br/>model + free parameters + tests]
    C --> D{{DEEM optimizer}}
    D -->|candidate parameters| E[numgeo<br/>element-test simulations]
    E -->|simulated curves| F[similarity measure<br/>vs. experiments]
    F -->|objective value| D
    D -->|converged| G[Optimized parameters<br/>+ PDF report + plots]

In code, a calibration is only a handful of steps — read the data, declare the model and its free parameters, set the bounds, and call the optimizer:

from ACT.utilities.excel import excel
from ACT import globals, DEEM
from ACT.models import hypoplasticity_isa

# 1) read the experimental database (mutates the reader instance)
database = excel()
database.collect("database-kfs-monotonic-cyclic.xlsx")

# 2) choose the model (default parameters and search bounds are pre-set)
model = hypoplasticity_isa()

# 3) describe the calibration problem
globals.setup(
    Model           = model,
    Free_parameter  = ["hs", "n", "ed", "ec", "alpha", "beta"],
    oedometer       = database.oedometer,
    triaxCD         = database.triax_CD,
    triaxCU         = database.triax_CU,
    path            = "./calibration/",
    Experimental_database = database,
)

# 4) run the optimization
DEEM.optimize(maxiter=200, n_cpu=8)

A complete, runnable walk-through is given in the Quickstart.

Related software¶

Citing numgeo-ACT¶

If numgeo-ACT contributes to your research, please cite the foundational publication:

A full list of works in which numgeo-ACT has been applied by our group is given on the References page.