CorePy is a data analytics tool designed to integrate core-based geological data for machine learning characterization
- The primary focus of CorePy is to classify high resolution X-ray fluoresence data into chemofacies
- unsupervised and supervised clustering tools are applied
- Folder structures are developed to simplify working on multiple cores and formations
- Visualizations are used to validate clustering results
- CorePy consists of individual scripts and Pip install python package called corepytools
1) Additional notes and tips about GitHub and steps I use are here: https://github.com/Totilarson/MyCheatSheet
1) fork the CorePy repo to your github account
2) command line: navigate to the folder you want to install CorePy
3) make a local clone:
- command line: `git clone https://github.com/Totilarson/CorePy.git`
- if it is necessary to delete the local clone use: 'rm -rf .git*'
pip install -r requirements.txt
1) Inspect //CorePy/Data
- CoreData folder contains folders to store XRF, attribute, and beta data specific to each core.
- Open the Public_XRF file and undertand that each XRF data point is depth referenced using Core-Box-inch sticker references
- Open CoreAttributes folder. The folder 'Public' is the core name and in that folder are _TOC, _UCS, and _XRD attribute files
- All attribute files are linked with the XRF file using core-box-inch referencing
- Naming patterns for core box sticker location, wireline depths, and elemental concentrations are shown
2) Inpect /Coredata/Corebeta folder
- Public.json contains core-specific data that is accessed by CorePy scripts. Wireline logs are also written to this files
- several settings have been shifted to the Public.json file, and this will become the repository for formation tops
1) There are a series of codes in Corepy codes that have specific functions. The general path is to select the core in settings.py, then run Attribute_merge.py to build integrated datasets
2) Settings.py
- 'settings.py' contains variables for all the Python scripts
- "Lease_Name" , "Depth_model", "Formation", and "RockClassification" are the primary values that are changed by the user
- machine learning parameters are stored in settings.py
- 'chemocolor' is generated here. It makes formation-specific color schemes. If you add a new formation you have to add its colorscheme here
- Run_settings.json file is created when settings.py is executed. variables are stored here
- Elements: the complete list of elements analyzed by XRF. Elements_plotted: a specific order of elements for plotting. Model_elements: elements included in machine learning classification
- Run_settings['Elements_Depth'] was added and is formation-specific. Through time, there are key elements for each formation and I wanted an easy place to store them
- <corename>.json files are stored for each core in //CorePy/CoreData/CoreBeta
- files provide core-specific data that is referenced in each script
- Wireline scripts also write data to the .jsom file
- The .json files can be editted with core specific data.
- Merges attribute data from //CorePy/CoreData/CoreAttributes/<core name> with XRF input file
- Merges wireline data from //CorePy/CoreData/WirelineLogs/<core name> with XRF input file
- The files are merged based on Core-box-inch input from the XRF and attribute files
- Wireline log data is resampled based on core XRF data spacing
- output is a .csv file that merges XRF and attribute data
- if no attribute data is in folder it will skip over it
- Running Attribute_merge.py will build additional output folders
- Running PCAexample.py will run PCA-Kmeans.
- Output files are in output folder. CSV file includes additional columns of data with colun titles Chemofacies_PCA
- the number of K-means clusters is selected in the settings.py file
- Machine learning parameters for Neural model and XGBoost clustering have been added to settings
- these scripts build and apply results from supervised chemfoacies classifications
- An example training dataset is included in //CorePy/CoreData/CoreNeuralModel
- model parameters are output _XGB and __NN files in //CorePy/CoreData/CoreNeuralModel
- output .csv file has additional classification columns
- machine learning parameters are stored in settings.py
1) Elemental plots and depth chemostratigraphy plots
- elemental cross plots. Elements are selected from Run_settings['Elements_plotted']
- elements plotted with respect to depth: Depth model selected by: Run_settings['Depth_model']
- element box plots. majors and trace
- pie chart of chemofacies abundance
- Depth referenced chemostrat column output in a folder //CorePy/CoreOutput/CrossSection/
- This code does take trial and error to get the bounding parameters correct
- line 37 Corebeta['CoreBox_crop_points'] are stored for each core in the Corebeta .json file. They need to be adjusted depending on the core box photos
- Core depths are collected from the file name. The file name structure is: PublicCore_3978.jpg where 3978 refers to the top depth of the box.
- top and bottom depths for each core tube are calculated based on the box top depth, core tube length, and number of tubes in a box. This data is stored in the corebeta .json file
- Corebox photos are unique and it takes time to get this part correct
- the output are coretubes that are depth referenced
- Designed to overlay chemofacies results on corebox photographs
- Requires coreboxphotographs be converted to 'coretubes'
- Coretubes are created from Corebox_Crop.py
- Coretubes are depth registered and in folder //CorePy/CoreData/CoreTubes/
- 'Every core is unique'....typically Depth_calculated is used to identify the position of the XRF sticker, but there are issues specific to each core
- the primary problem is when a core box overlaps another corebox
- This is a plotting function and develops descriptive stats for each chemofacies based on attributes
- Core_attribute.py is run after Attribubte_merge.py
- output is a .csv file with descriptive statistics
- Box plots and depth plots show attribute results with respect to chemofacies
- It is necessasry to add "Attribute_plotted" to the core .json file
- download IrfanView https://www.irfanview.com/
- rotate images: Image -> Custom/Fine rotation. Adjust angle until left side of box is vertical.
'OK'. File -> Save (Original Folder) -> 'Save'
- cropping photos: Edit -> Create custom crop selection -> "Save and draw on image"...move box
crop to desired cropping. Edit -> Crop Selection (cut out) -> Save Original Folder
- cropping photos (2) -> manually click on each photo to crop. Faster this way and you can visualize
if teh core photo needs to be rotated.
- Crop photo notes -> the goal is to be consistent. Having some of teh cardboard box around the photo
is OK
Batch Rename: File -> Batch conversion/rename -> batch rename -> Batch rename settings (options) ->
Name pattern: R1_#####_#####, starting counter: 10500. increment: 12, where 12 is the length of the core box.
Add all files -> Sort files: "by name, ascending natural"
Start batch : I put them in a separate file so nothing gets deleted
CorePy is being developed by Toti Larson at the University of Texas at Austin, Bureau of Economic Geology, Mudrocks Systems Research Laboratory (MSRL) research consortium.
-
Toti E. Larson, Ph.D. - Research Associate at the University of Texas at Austin. PI MSRL research consortium
-
Esben Pedersen, M.S. - Graduate student (graduated 2020) at the University of Texas at Austin.
-
Priyanka Periwal, Ph.D. - Research Science Associate at the University of Texas at Austin.
-
J. Evan Sivil - Research Science Associate at the University of Texas at Austin.
-
Geoforce students - Ana Letícia Batista (2020) - Jackson State University
Install corepytools using pip install corepytools Follow over to the authors Github account to download example Python scripts that use corepytools
CorePy
|-LICENSE.txt **MIT**
|-README.md **edited in markdown**
|-gitignore
|CoreData
|CoreAttributes
|CoreBeta
|CoreBoxPhotos
|CoreNeuralModel
|CoreXRF
|CoreOutput
|CorePycodes