O&G Equipment Failure Prediction

Situation: In onshore US, shale oil & natural gas (O&G) production involves large-sized high pressure pumping equipment. They are subject to catastrophic failures leading to production disruption and expensive repair costs.
Approach: We adopted an innovative approach to identify informative features from telemetry and engineer those features to train the machine learning model under normal operating conditions. We also adopted a likelihood based approach to perform multivariate density estimation and outlier analysis.
Desirable levels of false positives and false negatives were achieved with advance (in terms of days) notification for any anomalous behavior that confidently predicted an impending failure. Model was successfully cross validated across numerous data sets.
Machine learning algorithm was successfully deployed in Google Cloud Platform (GCP)
Developed for publicly traded O&G company

Likes (0)