In periods of peak demand, Baker Hughes crews work around the clock to tap oil and natural gas reservoirs. At a single well site, as many as 20 trucks may operate simultaneously, with positive displacement pumps injecting a mixture of water and sand at high pressures deep into drilled wells. These pumps and their internal parts, including valves, valve seats, seals, and plungers, are costly, accounting for about $100,000 of the $1.5 million total cost of the truck.
To monitor the pumps for potentially catastrophic wear and predict failures before they occur, Baker Hughes analyzes pump sensor data with MATLAB® and applies MATLAB machine learning algorithms.
“We saw three advantages in using MATLAB to develop our pump health monitoring system,” says Gulshan Singh, reliability principal and team lead for drilling services at Baker Hughes. “The first is speed; development in C or any other language would have taken longer. The second is automation; MATLAB enabled us to automate the processing of large data sets. The third is the wide variety of technologies that MATLAB provides for working with data, including basic statistical analysis, spectral analysis, filtering, and predictive modeling using artificial neural networks.”