Haibarakiku

Generates an Abaqus FEA training dataset for surrogate/ML models using Latin Hypercube or sparse-pattern sampling over a parameterized design vector. It handles batch job submission, crash recovery, and writes a unified dataset index, ideal for sweeping design parameters or building training sets.

Converts Abaqus FEA outputs into ML-ready (X, Y) wide-table CSVs by pivoting irregular mesh node displacements onto a regular N×N grid, using the final frame as the deformation target, and aggregating data from multiple cases into X_amplitude.csv (design vectors) and Y_grid_uz.csv (flattened grid displacement).

Performs closed-loop inverse-design validation by solving an inverse problem on a trained surrogate model. It then verifies with Abaqus FEA, comparing predicted versus true displacement fields and reporting error metrics to quantify the surrogate-FEA gap.