Scary questions and woolly mammoths are best approached and tackled by groups. That’s why several dozen engineers gathered this week at the Honeywell User Group meeting in Madrid to answer, “How can AI software drive operations excellence?”
The group was moderated by Michel Teughels, senior product solutions manager at ExxonMobil, and Nadia Merzaa, solutions consultant, Honeywell. “Digital transformation includes cloud computing, Industrial Internet of Things (IIoT), mobility, big data and artificial intelligence (AI), but for an operating company is an AI strategy needed?” asked Merzaa. “The answer seems to be yes and sometimes no.”
Honeywell’s overall AI strategy focuses on enterprise analytics and insights, which are enabled by safety and operational excellence, sustainability, competence and productivity, end-to-end optimization and asset reliability, all protected by a cybersecurity program. “Honeywell has adopted an AI strategy to stay ahead of the curve, but oil and gas and other process industry users must decide if AI can help their businesses and outcomes. We envision AI as a way to accelerate your achieving those outcomes, and this is where AI can play a big part.”
Small steps to start
Teughels reported that ExxonMobil’s goal for AI is autonomous decision-making, but it’s pursuing it incrementally. “We’re not planning to spend a lot of time yet on using AI for very advanced analytics. As an operations company, we want to use AI to solve individual problems,” explained Teughels. “So, we’re using AI as a cognitive advisor, but we’re doing it in small steps, and progressing toward autonomous answers and decisions we can trust. This will give us confidence in AI, but we’re still going to have human operators monitor it.”
Although ExxonMobil seeks to run its worldwide facilities in a standardized manner, as if they were all one plant, Teughels added the biggest hurdle in using AI is the company generates a large volume of data that may not be consistent enough to be useful. “AI produces lots of data, but it’s not clean, so we’ll need to add naming conventions and other context,” added Teughels.
Some cautious toes in the water
The overall mood of the AI roundtable’s other members was cautious. Though most haven’t adopted any AI functions yet, several reported that their companies are reflexively jumping on the AI bandwagon to gain what see as competitive advantages before examining how they can apply it and machine learning in practical and useful ways.
“We’re an oil and gas operating company, and our management wants to use AI and ML right away, but we think we need a more strategic approach about how to get there,” stated one attendee. “We don’t want to develop AI and simply test it on ourselves.”
Another roundtable member reported that his production unit uses AI for maintenance and to examine piping photos to check for damage. “We’ve developed a tool that can check for corrosion,” he said. “This helps us determine the right places to build scaffolding, so we don’t have to waste hundreds of thousands of dollars building it in the wrong places like we did previously.”
A third roundtable attendee added that his company is about to stand up advanced process control (APC) on a unit. “However, this program can’t learn yet,” he explained. “We’re building the model, but this isn’t where AI can help.”
A fourth member agreed that while APC applications have been available for years, they are typically limited by the algorithms used to populate them. “This is where AI could help because its technology is faster and potentially more capable,” he added. “AI could be leveraged for new operating regions and help provide feedback we can use for optimization.
“We see AI as helping with predictive maintenance,” he added. “We previously looked at it for condition monitoring because we needed to get the compressors performing at better than 60% to 70% efficiency. Now, we want to get real-time health data of items like bearings, but APC can’t do it alone. AI should be able to help, so we’re making a big push in that direction.”
A fifth member reported that his company runs numerous compressors and could use AI to assist in analyzing all the data they generate. “It’s been a pain to schedule and have to sit on many maintenance issues,” he said. “We think AI can help with them.”
Two more roundtable members questioned AI’s precision and reliability. “I don’t think AI is precise enough for closed-loop control or other operations,” he said. “It may be precise enough someday, but I don’t know when that will be.”
The other member noted that AI’s performance depends on the reliability of its data sources. “If you add the wrong data, then AI could be like having a virus,” he cautioned.
Seeking a strategy
Teughels added that all this input about AI shows that each player needs to research and develop a strategy that will be useful for them. “A couple of years ago, we asked if we needed a cloud computing strategy,” said Teughels. “Now we need an AI strategy because without one we’ll be all over the place, especially because we’re often shorthanded. This is why we need to narrow our focus on AI and apply it in small steps. If it shows it’s OK for populating a dashboard, then that will build our trust for using it elsewhere. We’re also using AI to clean and validate data, so the next step could be asking it where to zoom in.”
