Predictive asset management provides new insights into equipment status to increase process uptime, avoid disruption and optimise asset utilisation. Through regular monitoring of production assets, predictive asset management helps users avoid costly, unscheduled downtime, reduce maintenance costs and more efficiently plan scheduled maintenance, all with the aim of increasing plant productivity and maximising capital outlay. Predicting and resolving problems before they occur also brings obvious safety benefits.
Predictive maintenance is the next step in the constant evolution of maintenance best practices. Reactive maintenance is a fix-on-failure approach. Preventive maintenance is servicing at fixed intervals or cycles. Condition monitoring allows maintenance informed by regular assessment and anomaly detection. Predictive maintenance takes the next step in this sequence by providing additional information on probability of failure and predicted time to failure, which allows accurate and effective maintenance scheduling.
“This specific knowledge empowers producers to extend the time between shutdowns with confidence and enables them to plan exactly which maintenance activities to prioritise, and which to postpone without risk of failure,” Paul Gogarty, asset lifecycle product manager, ABB, explains. “This means duration of shutdowns can also be reduced, providing more calendar time for production and profitability.
Growing interest
One reason predictive maintenance is becoming more popular is technology-related. The growth of the Industrial Internet of Things (IoT) has been fundamental to the development of predictive maintenance, as it allows greater opportunities to harness, share and use data drawn from a wide range of sources. The more data that can be gathered, the better scope for making predictions.
“Additionally, the advent of IoT provides new opportunities to increase asset reliability and availability while reducing maintenance costs and avoiding expensive unplanned disruption,” Gogarty continues. “While the desire to increase production and reduce costs may not be new, the methods for achieving them in this case are, which is a powerful driver of industry interest.
“Across industrial automation, sensors and detectors have been in place and in use for more than 20 years. However, when the IT industry started to make advances with Big Data analytics, we saw the possibilities to use increasingly higher volumes of data gathered on asset condition to create deeper analytics for asset management. The computing horsepower is now available to accelerate data analysis for practical and effective use.”
Managing the data
A key challenge to adopting a predictive asset management and maintenance approach is the marshalling and integration of data from a variety of sources. Machines, devices and sensors need to be set up for data gathering, through wired or Bluetooth methods, and the data needs to be collected, analysed and presented in a way that allows meaningful conclusions to be drawn and practical action taken. Additionally, in some cases where specific guidance is required, the data needs to be seen and analysed by experts.
“The biggest challenge to adoption is ensuring that a business is ready and receptive to embrace predictive asset management,” Gogarty adds. “The data and knowledge this approach generates is powerful, but for it to be effective, a predictive maintenance approach must be integrated fully into the existing processes of the company. Maintenance teams may still have systems and processes built around reactive or preventive maintenance; these will need to be updated and even disrupted for the new predictive maintenance approach to work effectively and to truly deliver incremental value.”
The path to predictive
Before a predictive asset management or maintenance contract can begin, the business will need to conduct a site audit to identify which assets will be included in the predictive maintenance program.
“First, an on-site assessment by experts and site maintenance specialists evaluates the precise maintenance needs,” Gogarty explains. “A long-term solution is then designed based on business goals and available technologies. The next stage sees the deployment of online condition monitoring technologies to provide continuous data on the assets. The service provider remotely measures and optimizes maintenance performance and presents results in periodic reports. Using a service agreement ensures the application is always updated and working optimally, so that customer maintenance teams can focus on their daily maintenance work.”
When assessing assets to be monitored, the criticality of the assets to the plant’s operation is taken into account, discerned by performing a criticality analysis in which equipment and processes are analysed for what would happen if something went wrong, and how that would adversely affect plant performance. “The predictive maintenance program is introduced in a phased manner according to the asset criticality ranking,” Gogarty says. “This makes for a smoother and easier transition and helps in-house maintenance teams with gradual rather than wholesale adoption.”
Gogarty believes that when looking to introduce a predictive maintenance solution, it should be viewed as an operational improvement project rather than an IT project. “Planning and budgeting for the technology and ensuring that it fits the plant’s OT/IT architecture, is important, but it’s more important to justify the investment directly to the improvement of production availability and operational cost reduction,” he says. “Ensuring user adoption is also critical.
“Many an investment has lost or delayed its return on investment because the users’ readiness to adopt and adapt was not considered. Design your project with the user in mind, and work with suppliers who have a defined approach for engaging with users. If a predictive maintenance project is to be successful, it needs the support from all relevant users, including maintenance, operations and IT. Without buy-in from all affected and/or interested parties to such a significant business process change, success will be slow.”
The need for data and digital twins.
For successful APM implementation you not only need data, but data in context. “The knowledge graph is about understanding an asset so you can predict failures before they occur,” Craig Hayman, CEO, Aveva, says. “The data that makes up the digital twin you are now putting to use, having it do something. Remember, the premise of digital twin was always, if I have a digital twin of an asset, I can do things with that digital twin without touching the asset. I have a copy of the physical asset digitally. With that now I can run all sorts of things around it without having to keep making individual connections to the physical asset. Now we are starting to see use cases for this and APM is a hot area; probably one of the best use cases for the digital twin.”
It is still early in the APM maturity curve, but Hayman is certain that its adoption will grow rapidly as understanding evolves. “My example is that when cars first had computers, sometimes you turn it on and you get a red check engine light,” he explains. “You would have to take it to the dealership to figure out there was nothing wrong with your car; it was a false positive.
“Historically that is the case with APM. People were only running small data sets, and they were only looking at one or two variables, so there was a lot of false positives. People thought there was a problem with an individual asset. Now, we are much better at looking at not just one pump, but 300 pumps, not just a one variable of two variables, but 30 variables around the pump. What has happened is because we have much more data, we are able to dramatically reduce false positives and get better at predicting when something is going to fail.
“One thing is that the technology has got a lot better on APM. The second thing is that we have started applying strategy around it. I think this is like going to the doctor. You know that there is a bunch of things you could do to improve your health, but there is probably a priority list around them. The same thing applies in and around APM. Inside a company there are thousands of asset types, perhaps tens of thousands of asset types, perhaps millions of assets. You need help to prioritise these; you need a strategy to understand which asset types you should go after? And in which order. And I think that is where the process of prioritisation becomes quite relevant.”
A key part of prescriptive strategies is building failure modes for high value assets. But given the nature of these assets where failures are rare how difficult it is often difficult to access data to build these models. Within APM there are classes of equipment. One class that APM performs well on is rotating equipment. “Anything that rotates we are very good at,” Hayman adds. “Our system can learn about that equipment. One customer had a gas turbine and we were monitoring it. We were looking at all sorts of variables around it. We do not manufacture gas turbines, so we do not know much about gas turbines. But we were looking at all the data around this gas turbine and the system learned how that turbine worked using historical data and current operational data. And it detected in an anomaly. It did not know what? But it knew that there was something different today about this turbine versus every other day.
“Back to the car example you get in your car, you turn it on, it sounds a little different today. You are not sure what it is, but it sounds a little different today. That was the situation with the gas turbine. We told the company, who called the manufacturer of the turbine and were told it was okay. In fact, they said it was just settling in to its environment. It then ran for another couple of weeks, and we went back to the customer after we found more anomalies that made us think it really was going to fail. We told the company and they cordoned it off from other workers and ran it in a reduced mode. During the next production window, they took it offline and found metal parts inside the turbine that were not supposed to be there.
“That turbine was indeed about to fail. We had been able to predict that it was going to fail, and the customers saved somewhere between $25 to $35 million on preventing this outage. What’s the lesson here? I don’t know that anyone was right or wrong in the situation? I think the lesson is that the more data you have, even if you do not know what that asset type is, you can learn something about it.”
That is much the same way that humans learn. The more data you have, the better educated you are about something. “I think that really applies in APM,” Hayman concludes. “Just one other point here, which I think we have learned as well is the data does not have to be data that sits in our systems. It could be data the customer has been collecting for years and not doing anything with. We will start looking at that data as well to build up a pattern around a pattern around the piece of equipment.”
One thing is for sure APM use is growing. According to Grand View Research the market size valued at $13bn in 2018 and is expected to register a CAGR of 9.15 per cent from 2019 to 2025. Look out for APM coming to an asset near you.
Five drivers for asset performance management
1. Loss of experience means less on-site skill and expertise as staff knowledgeable in asset reliability and maintenance retire, especially in the more developed economies.
2. Market pressures drive the need for higher production availability, which in turn drives the demand for maximum asset availability.
3. Margin pressures force producers to continuously find opportunities to eke out every bit of production capacity while reducing maintenance and operational expenditures.
4. Aging assets and slimmer capital budgets push producers to find new ways to extend asset life.
5. Asset complexity increases as components become more digital, driving a need to access higher knowledge levels to understand and address asset issues.
The Factory that never sleeps
Tenaris is a leading global manufacturer of steel pipe products for the world’s energy industry. Its Dalmine site runs 24/7 to produce some 800,000 tonnes of products a year.
To ensure uninterrupted production, the company implemented a planned, preventive and predictive maintenance program focused on reducing the frequency of repairs of 460 electric motors that drive the site’s rolling mills and other vital production equipment.
A wireless gateway automatically collects data from the motors and transmits it to a dedicated portal for storage and analysis. Tenaris engineering technicians can assess the condition of all the motors being monitored. Additionally, if there are any faults and anomalies, the system sends an alert to the maintenance team via e-mail.
“With an infrastructure of this size, the efficiency and timeliness of maintenance are vital to the profitability of the company,” sys Ettore Martinelli, Tenaris maintenance engineering director, says.
