Artificial intelligence and machine learning technologies are helping manufacturers move beyond time-based maintenance practices, aiding more proactive practices and predictive maintenance. When a maintenance team detects anomalies in their measurements, vibration for example, that means there’s already equipment degradation. AI/ML can help identify those anomalies in asset data earlier, before deterioration starts.
These advanced technologies are also paramount to breaking down siloes in the manufacturing system and processes as a whole. Here, experts at Aspen Technology share eight ways these advanced technologies will impact PdM strategies in the future.
Looking beyond asset failure to process improvement
Asset performance management includes asset failure analysis, but also monitoring product quality and production efficiency through process analysis. “Not only do I want to look at predicting the failure of the equipment, but I want to look at the process,” says Don Busiek, senior vice president and APM general manager at AspenTech. “Am I producing a product with the right level of quality? Is it within my quality metrics? What if I change the variables? What if I change how much I want to produce? What is the impact on the quality? That’s also part of what we are solving with our APM solutions. How do we improve maintenance costs, drive down unplanned downtime and ensure the right amount and quality of the product gets shipped?”
Collecting and historizing worker knowledge
The manufacturing workforce is changing rapidly, and technology should aid that transition, collecting worker knowledge and making it widely available for newer generations.
“Part of the challenge of the maintenance base is the transformation that’s happening with the industrial workforce, and making sure the software works for all employees (digital, maintenance, reliability, process, operations, etc.),” Busiek says. “How do you take the knowledge of the retiring worker that’s been doing maintenance for 30 years and capture it? How do you deliver that knowledge in a digital format that accelerates an employee’s ability to do their job? How do we make lives easier and deliver more business value? How do we provide a digital solution for all roles in a format that is tailored for each of them?”
Taking the place of physical inspections
Many companies are still using traditional time-based maintenance practices, where the maintenance staff inspects equipment at a certain interval and performs actions such as measuring equipment vibration. Oftentimes, the maintenance team performs routine maintenance on the assets whether they need it or not, but technology, like machine learning, can help make maintenance and asset reliability more proactive. For example, using sensors to monitor a wide range of variables, such as vibration, temperature, location or performance, can then trigger an inspection or a work order depending on the criticality of alert from the solution, providing more lead time on issues or avoiding equipment degradation altogether. “When you add sensors to the asset and then you apply machine learning on top of it, now we are detecting patterns without doing the physical manual inspection,” says Nithiya Parameswaran, vice president of product management at AspenTech.
Helping prioritize maintenance planning, more proactive less reactive
Using artificial intelligence may extend the life of some equipment with more prescriptive upkeep, but assets failures can still happen. The advantage with technology comes with planning. “If you know 60 days ahead of time that something is going to give way, you can plan for it. It gives you that time to plan and you’re not reacting or running around trying to get spare parts. This planning not only ensures the right spares are available, but it also ensures that properly certified and trained employees are available to perform the work. Essentially, planning allows companies to dramatically reduce their maintenance cost and unplanned downtime,” says Pratibha Pillalamarri, senior product marketing manager at AspenTech.
Moving beyond the self-optimizing plant
The endgame for machine learning technology in industry may be self-optimizing or fully autonomous assets, processes, and plants running full scale on the internal data, self-correcting and self-fixing. AI could even take that one step further. “We start to get into this concept of not just a self-optimizing asset and a self-optimizing process, but a self-optimizing plant, which requires companies to leverage all of the data that’s available, deriving insights, and making those insights available to all appropriate personnel,” Busiek says. “It’s pulling data from a variety of sources, both inside and outside of the plant. These data sources could range from information about the equipment to parameters like the weather and humidity. The data is then sanitized, analyzed, validated and then made available. To me, I think that’s where the industry could go: using AI/ML to perform repetitive tasks and anticipate future performance of an asset, process, or a plant. This frees up time for plant personnel to do more complex and value-added tasks. Machine learning will help create the next wave of industrial efficiency.”
Influencing choices for capex spending
Machine learning technology can predict failures from pattern signatures in the data, often spikes in measurements, which can be missed with time-based inspection. Technology can tell facilities ahead of time when an asset will fail, if they don’t take action. It can also look further than the individual assets at the entire process and the interaction of assets together, which can greatly inform capital expenditure decisions. What asset failures create a choke point? What assets are most critical? Will a sound maintenance program be enough, or does the facility need backup equipment? “That kind of analysis can also have an impact in terms of your capital expenditures,” Parmeswaran says. “You can run simulation models and figure out what are those choke points.” Then, plan capex around that critical infrastructure.
Curating vast amounts of information with generative AI
Tools for generative AI like ChatGPT have bounded into everyday landscape, and businesses and industries are both trying to figure out how to use it to their benefit, and also how to protect against misuse. Generative AI may even have a place in industry and maintenance practices. “If you ask ChatGPT, how do I predict a failure in a pump? Actually, it does come up with answers,” Parameswaran says. “There is a vast amount of information that’s available. But every company operates their equipment differently and with different business objectives. So one cannot simply take a generative AI answer as the gospel truth. Companies must first ask questions like, how do you curate it? How do you collect it? How do you validate it? How do you then make some sense out of it? And, how do you align it with how you operate the equipment and your objectives?”
Getting quick wins
Most organizations can see the benefits of digitalization and AI-based automation, but the challenge of reinventing entire practices and infrastructure at scale is overwhelming. Manufacturers can start small and scale slowly. “Most of these companies have the data and they’re trying to find ways to unlock their data, get insights from that data, and solve those problems today. Whether it is energy efficiency or improving maintenance practices, that data is there or can be made available. And what they want is quick wins—how to use the data, unlock insights, and do so with their existing resources, with the existing facility and people,” Pillalamarri says. “That’s where ML is helping these customers. You’re able to add to what they have today without having to stretch, without having to invest in something new. You’re looking at the data and trying to make adjustments to your process, to your maintenance practices, and thereby meeting your profitability goals or your sustainability goals.”