NasgroPy Project Showcase

Here’s a breakdown of the NasgroPy project, focusing on the crack propagation models, design of experiments (DoE), and the machine learning techniques used.

Project Overview

NasgroPy replaces NASGRO’s batch process with a Python-based solution, integrating Design of Experiments (DoE) and machine learning to create surrogate models for predicting crack growth.

Key Crack Propagation Models

TC01 – Through Crack at the Center of a Plate
CC16 – Corner Crack at a Hole

These models were used to simulate crack growth under various stress conditions, and machine learning techniques were applied to predict outcomes based on the simulations.

Machine Learning Techniques

The project applied Gaussian Process, Neural Networks, and Random Forest models to build surrogate models that predict crack propagation behavior. By automating the process, we were able to scale up simulations and make accurate predictions.

Sample Code


# Example of Sobol sequence integration with NASGRO simulation
from sobol import sobol_sequence

def run_simulation(params):
    # Sobol sequence for parameter sampling
    samples = sobol_sequence.sample(params)
    # Run the NASGRO simulation with the sampled parameters
    results = nasgro_run(samples)
    return results