tools

Tools — Data Regeneration

This folder contains the Python script used to generate TestData_FORGE_VDL.txt from the original Schlumberger DLIS source.

You do not need any of this to run the demo. The bundled test data file is checked in and the C# project loads it directly. This folder exists so that:

• the data lineage is auditable (anyone can verify that the demo data came from the cited DLIS file without alteration), and
• the conversion is reproducible for other channels, depth ranges, or wells.

---

Prerequisites

• Python 3.9 or later

• Two pip packages:

  pip install dlisio numpy

  Add lasio if you also want to process LAS files:

  pip install lasio

The script has no other runtime dependencies.

---

Source Data

The bundled TestData_FORGE_VDL.txt was generated from the Cement Bond Log DLIS file in the Utah FORGE 16A(78)-32 dataset:

• GDR submission: https://gdr.openei.org/submissions/1330
• DOI: 10.15121/1814488
• Licence: Creative Commons Attribution 4.0 International (CC-BY 4.0)
• Citation: McLennan 2021

The relevant file inside the submission is the CBL DLIS (~95 MB on disk). Download it from the GDR page above; the script reads it directly without unpacking.

---

Listing Channels

Before converting, run --list to see what's inside any DLIS or LAS file:

python convert_log_data.py path/to/CBL.dlis --list

The DLIS lister enumerates every logical file, frame, and channel, showing each channel's dimensions and units. Look for an array-shaped channel — VDL data is stored as [n_depths, n_time_samples], typically [n, 256] for Schlumberger CBL/QSLT tools.

LAS files use a different convention — array waveforms are stored as parallel scalar curves named VDL[1], VDL[2], ... VDL[256]. The LAS lister groups these automatically so you see one entry per array rather than 256 individual columns.

---

Regenerating the Bundled Demo File

The exact command that produced TestData_FORGE_VDL.txt:

python convert_log_data.py path/to/CBL.dlis VDL \
    --depth-range 3000 3500 \
    --depth-stride 1 \
    --out TestData_FORGE_VDL.txt

This crops to the cemented casing interval (3000–3500 ft, sealed cement section), keeps every depth row at the source 0.125 ft resolution, and normalises amplitude to a 0–100 range with the 99.5%-ile saturation default. The result is 4001 depths × 256 time samples ≈ 50 MB on disk.

For a smaller demo file suitable for a public repo without the LFS-size overhead:

python convert_log_data.py path/to/CBL.dlis VDL \
    --depth-range 3000 3500 \
    --depth-stride 4 \
    --out TestData_FORGE_VDL.txt

--depth-stride 4 keeps every fourth depth row, dropping the file from ~50 MB to ~12 MB. The contour-injection technique still works identically on the smaller grid.

---

Useful Flags

Flag	Effect
--list	Enumerate channels and exit (no output written).
--frame NAME	DLIS only: read the channel from a specific frame.
--out PATH	Output filename (default: TestData.txt).
--depth-range LO HI	Crop to depth interval, in source units.
--depth-stride N	Keep every Nth depth row. Use 4 or 16 for smaller demo files.
--time-stride N	Keep every Nth time sample.
--sample-us VALUE	Override the time sample interval. The script tries to read it from DLIS metadata (WSR / WAVE_SAMPLE_RATE / SR parameters); falls back to 6 µs if absent.
--raw	Skip the 0–100 normalisation and write source amplitudes verbatim.

---

Output Format

The C# loader (LoadVdlTestData in Mainwindow.xaml.cs) expects the following exact format. The script writes it byte-for-byte:

• Encoding: UTF-16 LE with BOM (FF FE)
• Line endings: CRLF
• Field separator: tab
• Three columns: time(µs)   depth(ft)   amplitude
• Empty time field when time == 0 (start of waveform)
• Empty amplitude field for null cells
• Ordering: outer loop = depth, inner loop = time

This format predates the project — it matches the original test data file shipped by Gigasoft for ProEssentials VDL examples. Keeping the exact format means the demo can drop in any conforming file without C# changes.

---

Adapting to Other Wells

The script is not FORGE-specific. Any DLIS file with a 2D array channel indexed by depth and a sibling depth channel (TDEP / DEPT / DEPTH / MD) will load. Schlumberger CBL, QSLT, and CBT tools all produce compatible arrays under various channel names — VDL, WF1, WAVS, depending on the tool generation. Run --list first to find the channel name in your file.

For LAS input, point the script at the array's base name and it auto-stacks the columns:

python convert_log_data.py path/to/well.las VDL --out TestData.txt

---

Licence

This script is MIT licensed, same as the rest of the example code. The data it processes is the user's responsibility — the FORGE dataset above is CC-BY 4.0 (attribution required); proprietary client DLIS files retain whatever licence terms govern the original.