WILLOW – Synthetic ultrasound datasets for pitting corrosion modelling (CEIT - Task 2.4)

SYNTHETIC ULTRASOUND DATASETS
--------------------------------------------------------------------------------
Setup & Simulations:
- Generated by simulating 3D ultrasound pulse propagation in pitted steel specimens with varying shapes and sizes using the k-Wave MATLAB toolbox.
- Two specimen thickness variants (1,000 synthetic signal samples each in .mat format):
    * Folder: "dataset_results_3D_3mm_1000samples" (3 mm thick steel specimens)
    * Folder: "dataset_results_3D_5mm_1000samples" (5 mm thick steel specimens)

Files & Structure:
- .mat Files (1,000 samples per dataset):
  Each file contains the following structures and parameters:
    * pit: Structure containing the input pit definition for the simulation (dimensions, shape, axial offset applied to shape, and backwall roughness).
    * pit_eff: Structure containing the actual effective depth and width (Ground Truth) calculated after geometry definition and grid discretization.
    * trans_len: Transducer diameter (either 6.4 mm or 3.2 mm, replicating laboratory transducers).
    * alpha: Apodization parameter used to define excitation in the simulation.
    * result:  The simulated raw ultrasound signal recorded up to the third (for 3 mm thick specimens) or up to the second (for 5 mm thick specimens) roundtrip echo.

- Processed .xlsx Datasets:
  Generated by processing the 1,000 .mat files for each respective specimen thickness.
    * dataset_ml_3mm_1000samples.xlsx
    * dataset_ml_5mm_1000samples.xlsx
  Structured similarly to the Harshlab .xlsx files but includes additional Ground Truth (GT) outputs.
  Key parameters/columns:
    * Ground Truth outputs: Depth (eff), width (eff), volume, surface area, surface-to-volume ratio, etc.
    * "metrics_*" columns: Pitting presence and properties indicators corresponding to the same descriptor definitions used for the Harshlab datasets.