Baker Hughes leverages Matlab ML/NN toolboxes

Huge savings expected from failure predictions derived from terabyte frac fleet training data set.

GE unit Baker Hughes reports that it has used Matlab’s statistics and machine learning and neural network toolboxes to predict failure and optimize maintenance of its frac fleet. High pressure pumping equipment accounts for some $100,000 of the $1.5 million total cost of a truck. To monitor the pumps for potentially catastrophic wear and to predict failures, a terabyte data set from 10 operating trucks was analyzed with a neural network.

Baker Hughes’s Gulshan Singh said, ‘Matlab converted unreadable data into a usable format and allowed us to automate filtering, spectral analysis and transform steps for multiple trucks and regions.’

Singh sees many advantages in using Matlab, ‘The first is speed, development in C or another language would have taken longer. Matlab also helped automate the processing of large data sets. Finally, Matlab offers a variety of technologies working with data, including basic statistical analysis, spectral analysis, filtering, and predictive modeling using artificial neural networks.’

MathWorks helped Baker Hughes develop a script for parsing binary sensor data. Signal processing determined which signals in the data had the strongest influence on equipment wear and tear. Neural network analysis found that pressure and vibration sensor data was the best predictor of failure. The application is expected to bring savings of ‘more than $10 million per year’ and ‘reduce overall costs by 30–40%!’

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.