Predictive Maintenance for Oil and Gas Supermajors

Predictive maintenance in the energy industry aims to solve a billion dollar problem — Asset Integrity. The energy industry is seeing rapidly aging infrastructure. More than half of the pipelines in the U.S. are older than 50 years and in the past 5 years alone, there have been 3,300 leaks reported. This includes the largest in U.S. history with eighty deaths and close to 400 injuries. The challenge for asset integrity teams within oil and gas companies is that they are limited by legacy technology, computing resources, and antiquated methodologies including manual observation and recurring scheduled maintenance which inefficiently over allocates resources to maintain systems.

In partnership with a leading oil and gas consultancy Bitstrapped determined that to develop predictive maintenance for oil and gas applications, an organization would need:

1. A cloud based solution
2. Machine learning capabilities
3. Ability to operate machine learning with minimal effort
4. Easy to use tools and systems to enable more users and teams to benefit from predictive maintenance

Bitstrapped led the development of multiple software applications that met these requirements for a variety of clients including architecting and operationalizing the machine learning models. The end-to-end solution employed:

• Fully managed Google Cloud Platform products to minimize client overhead, limit need for physical on-premise infrastructure, provide pay-for-use efficiencies, enable teams anywhere globally to access, minimize operational overhead or maintenance required by the IT teams
• Machine learning models that could predict a variety of maintenance outcomes spanning multiple assets within the oil and gas facilities
• End to end machine learning automation so teams were not required to manage and evolve the technologies themselves
• Graphical user interfaces where teams could update the datasets that powered the models, configurability to control assumptions and inputs that changed how the models processed data, and visualizations that gave teams insights into maintenance schedules, simulated outcomes, and predictions via graphs

The cloud-based software solutions that made use of machine learning, provided the organizations with a variety of impressive results. The organizations remain customers today.

• Organizations saw the time required to run analyses to determine maintenance decrease 10x, from multi-stakeholder, multi-step analytical processes spanning 1 week, to a maximum of 20 minutes
• Organizations experienced an increase in predictive maintenance abilities across the organization, where once only a hand full of expert statisticians and data scientists were conducting research and experiments on predictive maintenance, to over 100 analyst and executive types accessing and consuming the predictive tooling and insights — a 5x increase in adoption
• In select cases, organizations saw as much as a 60% decrease in costs associated with maintenance through reductions in processing costs of using on-premise server infrastructure for their analyses, a reduction in the frequency at which maintenance was conducted under preventative, scheduled maintenance programs