High-Temperature Creep and Stress-Relaxation Modelling of 316H Stainless Steel

Project Summary

This project develops a computational framework for high-temperature creep and stress-relaxation modelling of Type 316H austenitic stainless steel. Real neutron-diffraction measurements and finite-element simulation results are combined to study stress redistribution, time-dependent deformation, and constitutive behaviour at elevated temperature.

The workflow focuses on processing experimental and simulation datasets, validating finite-element predictions against measured stress states, and calibrating simplified stress-relaxation models for structural-integrity applications.

Key Contributions

• Real-data analysis using neutron-diffraction measurements from 316H stainless steel
• Finite-element validation using interpolated stress tensors at measurement locations
• High-temperature stress-relaxation and creep behaviour assessment
• Constitutive modelling using Norton, Norton–Bailey and Garofalo-type creep laws
• Calibration of a reduced-order stress-relaxation model from experimental/FEA data
• Automated processing of MATLAB, CSV and finite-element output datasets
• Reproducible Python workflow for material modelling and validation

Research motivation

Type 316H austenitic stainless steel is used in high-temperature structural components where time-dependent creep deformation and stress relaxation can control long-term integrity. For safety-critical components, it is not enough to know the initial stress state; the stress tensor evolves during high-temperature exposure. This repository demonstrates how real neutron-diffraction measurements and interpolated finite-element results can be processed into a reproducible modelling workflow for stress relaxation at 550 °C.

Real data included

This repository includes a compact numerical subset from the University of Bristol dataset:

Harry Coules (2020), Multiaxial creep stress relaxation in 316H stainless steel, University of Bristol data repository, DOI: 10.5523/bris.rhg1bk2424a6262ecb9nouzp9.

The uploaded source archive contains:

• neutron-diffraction stress/strain tensor measurements,
• finite-element model results interpolated at neutron measurement locations,
• Double-Cantilever Beam specimen information,
• Abaqus input/data files and Fortran creep-law files.

For GitHub practicality, this repository includes the compact .mat files needed for the numerical workflow, not the full large Abaqus .dat and .cae files.

What this repository implements

1. Real-data ingestion

   • Reads Bristol 316H MATLAB .mat files.
   • Converts neutron-diffraction stress tensors and interpolated FEA tensors into tidy CSV files.
   • Joins measured and simulated stress states at the same measurement locations and exposure times.

2. Real-data validation

   • Compares neutron-diffraction von Mises stress with interpolated FEA predictions.
   • Produces location-wise residual maps and relaxation-evolution plots.
   • Stores reproducible validation metrics.

3. Stress-relaxation modelling

   • Fits a compact exponential stress-relaxation law to the real 316H FEA relaxation trend.
   • Provides a bridge toward more advanced Norton, Norton-Bailey, Garofalo, or Chaboche-type viscoplastic models.

4. Constitutive creep modules

   • Norton power-law creep.
   • Norton-Bailey time-hardening creep.
   • Garofalo hyperbolic-sine creep law.
   • Simple stress-relaxation integration under fixed total strain.

Key processed outputs

After running the real-data scripts, the repository creates:

data/processed/bristol_316h_neutron_stress_tensor.csv
data/processed/bristol_316h_fea_interpolated_stress_tensor.csv
data/processed/bristol_316h_joined_nd_fea.csv
data/processed/bristol_316h_relaxation_summary.csv

figures/bristol_316h_stress_relaxation_nd_vs_fea.png
figures/bristol_316h_fea_vs_neutron_svm.png
figures/bristol_316h_location_residuals_heatmap.png
figures/bristol_316h_fitted_relaxation_law.png

results/bristol_316h_realdata_metrics.json
results/bristol_316h_relaxation_law_fit.json

Current real-data results

Using the Bristol 316H compact numerical dataset, the workflow processed 72 measurement points: 6 exposure times and 12 neutron measurement locations.

For the interpolated FEA versus neutron-diffraction von Mises stress comparison, the current validation metrics are:

{
  "n_measurements": 72,
  "temperature_C": 550.0,
  "mae_sVM_MPa": 24.09,
  "rmse_sVM_MPa": 33.12,
  "r2_sVM": -0.72
}

The negative R² is not hidden. It is discussed as a useful scientific result: the sparse neutron measurements and interpolated FEA fields do not match perfectly point-by-point, which is realistic for multiaxial stress-relaxation validation. The mean FEA relaxation trend is smoother and can be fitted well by a compact relaxation law:

{
  "sigma_inf_MPa": 80.05,
  "sigma0_MPa": 93.91,
  "tau_h": 381.84,
  "r2": 0.94
}

Repository structure

src/
  ingest_bristol_mat.py             # real Bristol .mat ingestion
  analyze_bristol_realdata.py       # validation plots and metrics
  fit_relaxation_law_realdata.py    # simple real-data relaxation-law fit
  creep_models.py                   # Norton / Norton-Bailey / Garofalo laws
  calibration.py                    # fitting utilities
  stress_relaxation.py              # stress-relaxation ODE integration
  parse_abaqus_outputs.py           # Abaqus text parser utilities

data/
  raw_external/bristol_316h_subset/ # compact real-data .mat files
  processed/                        # generated tidy CSV datasets

figures/                            # generated validation and fit figures
results/                            # generated JSON metrics
reports/technical_note.md

Scientific honesty

This repository does not claim new laboratory experiments. It uses real open experimental/FE data from the Bristol 316H dataset and builds an original, reproducible computational workflow around it. The constitutive-law fitting is intentionally compact and educational; the next research step would be implementation of a full temperature-dependent creep/viscoplastic law and validation against complete FE fields or additional creep rupture data.