From fitness trackers that fit perfectly on a wrist to intelligent virtual personal assistants listening at the ready to answer random questions and make grocery shopping lists, the presence of artificial intelligence (AI) is rapidly increasing. AI is pushing human and machine closer together, and businesses are looking to it to help increase productivity and improve profitability while enabling a safer workspace for employees.

AI is the next step for the oil and gas industry to take as it walks deeper into the forest of alarms, monitors and sensors. Data are this century’s oil. E&P recently spoke to Philippe Herve, vice president for oil and gas solutions at SparkCognition, about the future of data, cybersecurity and AI.

E&P: How does one ensure that the data being used for analytics are good, accurate and being used for the right purpose?

Herve: Generally speaking, we work with the data delivered to us by our clients, which means the quality of data is up to the client. A machine learning algorithm is only as good as the data it’s been given. If the only data available are of poor quality, the model will not be any better.

While we cannot change the quality of data, we can and do work closely with our clients to make sure this problem does not come up. 

If we find that there is an issue with data quality, we alert the client so they can take corrective action, allowing us to ensure that the solution we deliver is always of the highest caliber.

Of course, even the best dataset will never be without its flaws. With any data, the first step in building a model is cleaning the data, inputing missing values, scaling datapoints, converting all data to the same format and a single scale, and rebalancing if there are any issues with sample size.

At the end of the day, though, our most important role in terms of data quality is keeping our clients informed of the base level of quality necessary to deliver good machine learning software.

E&P: What safeguards are in place to protect data in terms of cybersecurity?

Herve: It’s true that making use of collected data involves a great deal of transferring data between different assets and different levels of informational technology and operational technology. But machine learning also offers powerful protection against hacking and cyberattacks.

Cognitive endpoint protection can safeguard every endpoint in a network, shielding systems against malware, viruses, worms, trojans, ransomware and more. This includes protecting the commercial off-the-shelf systems that are otherwise a critical weak point that can be exploited to gain access to industrial systems.

Our own cloud-based cognitive engine, DeepArmor, uses a multilayer filtering process to detect threats. The first layer of protection includes file reputation analysis and application control, thereby quickly identifying known malicious and anomalous files. Once known files have been filtered, DeepArmor examines the DNA of unknown files to develop a threat confidence score for each file. A machine learning solution like it could have caught even the infamous Stuxnet worm that sabotaged Iran’s nuclear program.

DeepArmor can operate out-of-band, meaning it does not go through programmable logic controllers [PLC] to access the rest of the system. This is critical to protecting industrial systems, as it allows DeepArmor to detect anomalous operations occurring, regardless of any false information being transmitted by PLCs.

E&P: How can a company convert the data it has collected/ continues to collect into meaningful operational outcomes? How is AI helping convert data into dollars?

Herve: Oil and gas operators would be able to record everything in an ideal world. In practice, rigs only have so many sensors and those sensors cannot catch every single datapoint. Still, there are far more data generated by the sensors on an oil rig than humans can meaningfully use and understand. The scale of data and operations is beyond human capabilities.

This problem is only further complicated by the interconnected nature of oil and gas assets. A failure in one asset can have far-reaching consequences across the entire rig. A change in the state of shale shakers will affect the mud pumps. Separating out these streams of data and achieving a meaningful understanding of events and causality on the rig is no easy feat. Cyberattacks also are increasingly taking advantage of this confusion as subtler attacks often camouflage themselves as normal machine failure.

Data fill in the gaps in understandings of assets and systems, and provide technicians with a bigger picture of operations as a whole, allowing them to infer new insights. But this can only happen if a company has the right tools to interpret the data. Cognitive analytics provide a systematic way to make sense of the massive volumes of data collected across the entire oil and gas value chain.

This greater understanding of data has a measurable value for oil and gas companies. Data that have been analyzed and interpreted by AI solutions can predict when an asset will fail or ensure that resources and personnel are in a place where needed. Data can provide a window into safety issues on a rig and how to prevent them. When combined with natural language processing, AI can present its findings to human operators using a naturalistic communication interface. AI uses data to significantly reduce operating costs, increase worker and equipment safety, and optimize all processes on a rig.

In essence, what we do is take data collected from a rig, remove the clutter and extract the value. That value has a massive impact on a drilling operation and on the industry as a whole.