Mike Pennell, Jeff Hsiung and Venkat Putcha of OspreyData’s Data Science team presented at the SPE Western Regional Meeting, one of the Oil & Gas industry’s leading events, on April 25, 2018. The team presented a poster on OspreyData’s novel Human Augmented machine learning methodology and fielded questions by interested engineers & experts in the Oil & Gas industry. The audience showed a high interest in OspreyData’s applied neural networks, random forests and other machine learning models for detection of failures and suboptimal states in rod pumps. Jeff Hsiung’s presentation about using Convolutional Neural Networks for Dynacard description, an implementation very similar to image recognition used by smartphones and ID scanners, sparked more interest in conversations.
Their accompanying paper “Detecting Failures and Optimizing Performance in Artificial Lift Using Machine Learning Models” dives into the work they’ve done in diagnosing and predicting Artificial Lift problems using machine learning. This scope covers common tubing and pump failures along with sub-optimal performance of sucker rod pumps and and gas injection units.
These achievements are just the beginning of the valuable work in machine learning based failure prediction they are contributing to the field.
Read the full paper here: OnePetro.org
Venkat Putcha and Jeff Hsuing answer questions with attendees
Mike Pennell chats with audience members about OspreyData’s Human Augmented machine learning methodology
Venkat Putcha chats with conference attendees
