This repository is the result of scientific research using unsupervised ML algorithms to prevent failure during the drilling of oil&gas wells. The main goal is to produce a system that isn't essentially required a pre-drill model because the system utilizes a self-learning and self-adjusting model and proactively identifying an anomaly in wellbore condition and mitigates a stuck pipe incident before it occurs.
To run and output data to a local file:
./run.py
To run and output data to a matplotlib graph:
./run.py --plot
You must have matplotlib properly installed for this option to work.
To run and get code execution info:
./run.py --track
To run and finally get information & statistics about the state of the HTM:
./run.py --info
Model parameters is located in the model_params directory.
The chart produced with the --plot option contains red highlights where anomaly loglikelihood is above 0.35. The 0.35 threshold is low enough that it may provide only significant deviation.