formation-top-picker

Problem

Interpreting formation tops is a task undertaken by a geoscientist to determine significant boundaries in the subsurface to identify correlation between the wells or to develop a geological concept of the area. However, there are many wells need to be interpreted during the development phase which will be time consuming to go through all of them.

Solution

This repo is a ML based framework and app to train a machine learning model and perform prediction for picking formation tops or lithology zoning using the well logs provided by user. It also allow user to visualize the prediction through multiple colours of lithology zoning for easy interpretation by geologist.

Quick Start

1. Install using pip

   pip install git+https://github.com/haizadtarik/formation-top-picker
   

Run WebApp

1. Run the following command to launch the web apps:

   streamlit run app.py
   

Run in your code

1. Load and process the data

   from toppicker import WellData
   
   data = WellData(<PATH_TO_LAS_FILES_DIRECTORY>,[<FEATURE_NAME>])
   train_df = data.process_data()
   

2. Train a new model and save

   from toppicker import Trainer
   
   trainer = Trainer()
   trainer.train(<TRAIN_DATAFRAME>)
   trainer.save(<FILENAME>)
   

3. Load trained model and perform prediction

   from toppicker import Trainer
   
   trainer = Trainer()
   trainer.load(<FILENAME>)
   yhat = trainer.predict(<TEST_DATAFRAME>)
   

References

1. McDonald, A., 2021, Python and Petrophysics Notebook Series. https://github.com/andymcdgeo/Petrophysics-Python-Series