Maintenance spending

Maximising asset management and predictive maintenance

Natural Resources, Oil & Gas

How can companies acquire and transport the necessary parts maintenance with enough lead time to keep remote operations running.

Download PDF

Goal

Offshore oil drilling creates ever-present logistical challenges for oil companies. Supplies, crew, and equipment must be scheduled well in advance in order to avoid costly disruptions. That’s the challenge a major oil and gas company faced as it sought to optimise its maintenance strategy for gas compressors across several offshore platforms. The company’s primary objective was to build a model that would accurately estimate the remaining life of its machines and provide sufficient warning time to coordinate the necessary maintenance resources. Fortunately, data from sensors offered vital insight into equipment performance.

Insight and Action

QuantumBlack aggregated information from a number of different sources, complementing data from sensor readings with maintenance logs and deferment values, among others. Our team analysed more than 130 gigabytes of data to focus on a low-pressure compressor that had been identified as extremely critical. We developed a regularised predictive model using advanced feature engineering to extract information from highly dimensional data. Through this work, we identified 45 sensors that were most predictive of impending equipment failure.

Results

  • 80Percentaccuracy of predictions made more than five days prior to machine failure
  • 45Sensorsidentified as being most predictive of impending failure
  • 11Percentreduction in yearly maintenance cost