Real-Time Data Analysis, Predictive Maintenance Adds to Safety Profile
By Gregory Hale
In today’s technology-driven manufacturing ecosystem where artificial intelligence (AI) appears to fall into play in just about every discussion, there is little doubt there could be a huge opportunity for use in functional safety moving forward.
Facing a serious crunch of safety experts leaving the industry and not enough coming in to fill the void, plus a more sophisticated digital enterprise coming online, it only makes sense AI could help fill the gap.
“With the complexity of the design of systems and lack of resources, manufacturers have to adapt to this,” said Ahmed Elbatanony, functional safety offer manager at Schneider Electric. “There’s always a resistance, always questioning values and questioning usage and the real need for this and the investment. But eventually they are buying into it and more and more.”
That follows along with research trends as the functional safety market looks to reach $13 billion by 2031 from $7.63 billion in 2024, according to a report from The Insight Partners. One of the reasons for the growth trend is the need for advanced safety measures because of Industry 4.0 technologies, such as artificial intelligence (AI) and the Internet of Things (IoT). The report added incorporating AI and machine learning in safety systems will enable real-time data analysis and predictive maintenance, enhancing fault detection accuracy to allow for better decision-making in predicting potential failures before they occur, allowing for proactive maintenance and reducing downtime.
History Lesson for AI
“AI can keep a plant safe by helping by helping to pull together the information of how it is running and using that data to go back through history to see what happened in the past when things started to go in a certain way,” said Sheri Sammons, global functional safety director industrial automation at Schneider Electric. “AI can say, ‘hey, this happened before, and this is what happened,’ and it can then give people additional information on how and what could be causing issues in their plant.
“It’ll also help if you take some of the data in about the assets; it’ll help you understand if your assets are starting to have failures or getting close to failures based on past information. It can pull that together and say, ‘in the past, you had a failure on this device when it started to do this.’ That’s where AI can come in. We can use it in a way that gets more definitive answers than what we have today.”
To that end, right now AI for safety is not delving into the everyday process, but rather, providing information, research and data that can help move the plant forward; not necessarily saying there is a problem and go hit the shut off button.
“The next step would be once that it starts understanding how the plant should run, you could tell it if it gets to that state again, shut it off,” Sammons said. “That would be the next step is using that information and having AI do that. I don’t think we’re ready for that today. I see that could definitely happen in the next 10 years, maybe quicker.”
Prepare for Using AI
Showing data from a historic perspective from the plant is key, but manufacturers need to take the old school, dust covered paper manuals and digitize them. They could also use AI to see if they have enough data to ensure they have the proper information to set up and use the technology.
“Every company now must find a way to contain and maintain the knowledge of the data lake or data reservoir that AI compilers or AI agents can take to benefit their team,” Elbatanony said. “Instead of going through manuals and operating procedures and safety data sheets, I can have this already plugged in into my AI agent. I can search for it quickly. It will be easier to find with AI. It’s just a challenge with the accessibility and security of your data and how to keep it within your own environment and then applying it to safety. I think you’ll see on the front-end for sure, helping to identify the potential hazards.”
One of the advantages of AI when it comes to safety is it can take the subjective nature out of decision making. It can just provide the facts and keep human opinions out of the discussion.
“There was a study done by one of the big companies in Europe where they took 10 super specialized people in their company and a complementary 10 others,” said Matt Henderson, NAM Business Development Leader – Triconex Process Safety/Turbomachinery at Schneider Electric. “They all had a deep bench of diverse skills, but they were equal to each other. They gave them a problem statement and sure enough, they came up with different results.”
Culling Data
“Humans do things based on experience, whereas AI can only do what it knows from its instruction set. It will learn on its own and get smarter and smarter all the time. Humans don’t do parallel processing. We are only limited in the speed at which we can think. We’re unbelievable machines, but when you take a billion human inputs over hundreds of years and aggregate everything, you can do unbelievable things,” Henderson said.
On the other hand, when doing safety assessments, a team must come together but oftentimes they can introduce their own bias from past experiences whereas AI can eliminate that.
“One of the reasons you have all the players that you have on a team is to bring in their experience,” Sammons said. “Because one person’s experience may be different than another’s. In a way, you want those biases because you want to hear what they’ve seen and how they think about what they see.
“To me, the analytical part is where AI is very valuable because it takes a person too long to do all those counts, to pull all that information together. I think AI is a value add, and I think it can really help level everyone and confirm some of the details people know,” she said. “You still need to have the different perspectives and the different biases coming into the room just so you could have those discussions and understand what’s going on from a people standpoint.”
Train and Educate AI
Who is to say AI isn’t biased?
“AI by itself could be biased as well because it’s an algorithm,” Elbatanony said. “It’s machine learning, and it can learn from itself. And changing the way it answers and have multiple layers of answers every time. We’re going to have to train and educate AI and wait for it until we say it is just not biased. Think of a workshop where we are doing an assessment. We have the team. We can have the AI agent where we ask it questions, and then it will give an answer based on certain facts. You better define the question and the facts and numbers well, and then the AI algorithm will run it for you and give you good answers.”
In the end, humans are not going anywhere when it comes to safety in manufacturing, but also having all the proper data in context and at your fingertips is vital.
“AI can be valuable if it’s using data and the actual data for how the plant runs, because people have opinions on what’s causing problems and how safe things are,” Sammons said. “They may not actually have all the data in front of them when they make those statements. Because sometimes people have experienced things and they come up with something and think, ‘oh, well, I have to do it this way because one time, this is how it worked.’ You have to understand the risk, you have to understand what’s in place to protect you, and what the steps are to make sure that you’re you keep a safe state.”