Event gets down to details of data-driven drilling and production
Exhibitors at the Data Driven Production Conference (DDDP), held in Houston May 30-31, were a diverse set of suppliers of machine learning, analytics, and other digital technologies, including blockchain, impacting the upstream oil & gas industry.
Some issues under discussion at the event included how data scientists and engineers will collaborate to apply machine learning; how sensing and control assets are best deployed at remote sites; and how remote connectivity and model-based technologies will be managed as they proliferate on engineering desktops.
Several suppliers at the event also said they’re using blockchain to streamline oil & gas industry business processes.Until recently, blockchain was primarily associated with the use of crypto-currencies. This is no longer the case. Andrew Bruce, CEO of Data Gumbo, said blockchain will play a significant role in the oil & gas industry in applications that include validation and execution of performance contracts.
A blockchain is a decentralized and distributed public digital ledger used to record transactions. What’s key is the record cannot be altered retroactively by any single party to the transaction.
For example, say party A is contracted to supply a ton of sand to party B. Rather than manage this transaction based on records kept separately within enterprise system A and enterprise system B, the blockchain ledger has both parties looking at the same “sheet of music.” Recorded in blockchain, product delivery can trigger immediate payment for services. Disputes about order expectations and performance are easier to resolve.
Other blockchains will track equipment location and usage with greater precision than is typical today. In the same way, the logistic steps involved in water hauling could be tracked and validated.
“Blockchain is transformational,” Bruce said. “It is not going away. With its use, the debate about who owns the data goes away because it’s all about transactions. You can think of it as a kind of commercial Internet. Coupled with smart contracts and data from IIoT devices, blockchain will let us know exactly where things are and what their status is,” said Bruce.
At the well site
San Antonio, TX-based WellAware is an Industrial Internet of Things (IIoT) company that says it has a “full stack, modular solution to simplify collection, management, and analysis of oil field assets.”
David Milam, a company vice president, said IIoT ease of implementation is enhanced by the comprehensive data services available today, whether from Amazon, Microsoft, or others, which also reduce the cost of remote connectivity.
“WellAware helps companies reduce operating expenses, minimize downtime, and achieve safety and regulatory compliance,” he said. “It does so by reliably collecting remote data, allowing management by exception, and providing actionable analytics.”
To ensure success when embarking on IIoT proof-of-concept projects, it’s important that operators and other users carefully match technology capabilities with business requirements so that return on investment can be demonstrated, Milam added.
On its website, WellAware mentions its use of heterogeneous (het) networks, which for wireless networking, means it allows use of different access technologies. For example, with a mobile heterogeneous network, a wireless LAN network can maintain the service when switching to a cellular network.Therefore, WellAware has a single network relationship covering RPMA, cellular, 900 MHz, and satellite. Aggregate data is derived from multiple existing systems, including SCADA.
Steve Sarracino, CEO of Activant Capital, a major investor in WellAware, in an interview with Oil & Gas Engineering, recently said, “There is substantial and increasing investment in analytics and machine learning for oil and gas, but, if you can’t get to the data, no amount of effort can produce good results. The challenge really is that of building a data supply chain. That’s what WellAware does.”
Sunnyvale, CA-based Falkonry says it solves problems with its machine learning (ML) system, which uses multivariate time-series data to discover patterns, recognize conditions, and predict critical operational events.
Machine learning is a kind of artificial intelligence that allows systems using advanced statistical methods, to learn, in the sense of progressively improving performance of a specific task, often by exposing it to examples of data associated with good and bad outcomes.
For example, by looking at equipment and raw materials together, said Crick Waters, a Falkonry senior vice president, users are better able to meet quality specifications. By defining and monitoring a range of operations that will lead to good quality, users can be alerted in-situ and ahead of time when meeting specifications is in doubt, leaving operators opportunity to make needed adjustments.
“The same kind of thing might have been done using advanced process control,” said Waters. “But the amount of time needed to develop the solution using advanced process control would be overwhelming. You might be looking at, for example, more than 60 parameters to predict an event as many as 12 hours in advance.”
One pressing issue in operational machine learning is how much data-scientist support engineers need to make regular, beneficial use of the machine learning. “To start, we schedule regular training sessions, as often as once per week. Users need to learn to think in terms of pattern recognition and to establish a practice using the system. Long term relationships lead to measured outcomes,” Waters said.
Falkonry’s defined methodology includes first identifying a list of possible projects that will lead to good outcomes. These can be validated by means of a brief pilot, a proof of concept exercise, and a trial that can extend up to nine months.
Industrial time-series data is acquired using data historians, for example, and connectors make it possible to use data gathered by OSIsoft Pi, as well as feed the machine learning results back into the system, concluded Waters.
Educational sessions at the event included owner/operators and service providers. In one session, it was asked what will be needed to move decision making related to drilling to automated systems, as is done today when it comes to flying jet airliners?
Panel members said what’s needed is a data infrastructure, including sensors, programmable logic controllers (PLCs), and networks; decision-making algorithms; and real-time physics-based models or digital twins equipped with machine learning or decision-making logic.
For drilling, sensors are moving ever-closer to the drilling bit. One speaker noted that while steering and vibration already can be controlled in an automated fashion, automated management of fluid pressures and dealing with cuttings remain in the not-so-distant future.
Legacy to cloud
Ron Victor is the CEO of Santa Clara, CA-based Iotium, provider of network-as-a-service (NaaS) to enable industry to securely connect legacy onsite systems to cloud-based applications that leverage analytics, machine learning, and predictive analytics applications at scale.
“The cost for every using company to build an IIoT infrastructure is prohibitive,” says Victor. “The cost of managing those devices and managing the users is equally prohibitive.”
In contrast, Iotium services include a cloud-managed infrastructure for installing patches and a built-in firewall. No VPNs are involved. The port structure is maintained, which simplifies management. Virtual machines are instantiated at the data source and its destination such that no distributed denial of service (DDoS) attack is possible. Containers are isolated, packet arrival is confirmed, and operational technology is isolated from the enterprise network.
“Information technology (IT) and IIoT are different disciplines,” Victor said. “Small IT departments tend to understand the need for our expertise better than the larger departments, which tend to persist with the idea that they can build everything themselves. But they soon come around to our point of view.”
Arundo Analytics, Houston, TX, said its expertise is in “edge state streaming and analytics, rapid cloud deployment of machine learning models, and enterprise-scale models management.”
Cody Falcon, vice president of product, Arundo Analytics, said what all it comes down to is a focus on machine learning deployments at scale.
The challenge for data scientists, said Falcon, “Is they can build the algorithms needed for complex machine learning but have little idea how to execute.”
That’s why, Falcon said, it’s important that Arundo is staffed with engineers schooled in the oil & gas industry. “The work of the data scientist has to be infused with domain expertise,” said Falcon.
Another issue with models is that while they may work well when introduced, over time they can reflect less and less well the actual oilfield conditions. The fact is, models must be trained and retrained.
Falcon said he’s been surprised at the disparities among companies when it comes to how prepared they are to tackle machine learning. “Some of the largest companies are struggling to operationalize machine learning. Smaller companies need services,” he said.
Predictive maintenance seems to be the low-hanging fruit when it comes to applications, Falcon said, although a fuller, defined application set will emerge with impact versus feasibility studies uncovering a range of possibilities.
In another learning session, conference participants discussed some of the challenges the oil & gas industry will face as it adopts these new technologies. These challenges include time synchronization, sensor calibration and location, telemetry, application of general standards to oil & gas specifics, interoperability, and contractual language.
“We’re moving from Excel-like to Lego-like systems,” said Ragu Gandikota, an executive with MindMesh. “In the future it will be seen we’ve moved from an older tool set to a newer tool set. At the moment, within the oil & gas industry, that transition is ongoing.”
Kevin Parker, senior contributing editor, Oil & Gas Engineering, email@example.com.