Wireless

Preventative maintenance a key benefit of industry 4.0

17th August 2020
Joe Bush
0

Equipment down-time is a major issue for any manufacturer. For that reason, most will maintain a schedule of maintenance inspections and tests based on elapsed time or usage to determine the health of each machine. Though it is effective, organising maintenance according to a strict schedule is inefficient. Cliff Ortmeyer, Global Head of Technical Marketing at Farnell, explain the importance of preventative maintenance.

The heart of the problem is the inflexibility of any fixed schedule. To avoid the high costs that come with unexpected down-time, maintenance schedules are typically conservative.

They err on the side of having more inspections than are necessary to avoid unscheduled failures. A study on enterprise asset management by the ARC Advisory Group claimed as much as half the money spent on maintenance performed using fixed schedules is wasted.

Optimising maintenance

The cost of traditional maintenance is largely down to two factors. One is that of the personnel who need to move from machine to machine and from site to site to carry out the inspections. The other is the cost of having to take machinery offline to carry out each inspection. Although schedules try to avoid busy times, this is not always possible or easy to predict, with the result that manufacturers need to build additional capacity into their lines to cope with the loss of productive machinery at peak times.

Predictive maintenance, on the other hand, offers the opportunity not just for substantial cost savings by only performing manual inspections when needed, but in reducing the costs associated with both scheduled and unscheduled down-time. Several experts, including Steve Sands, Product Management and Marketing Manager at Festo, see predictive maintenance as a key stepping-stone in the transition to Industry 4.0 technology. As Sand says, this is because “you can see a clear return on investment” from its application.

Real-time data collection

With predictive technology, the maintenance schedule is no longer fixed but driven by the actual wear and degradation of the equipment. Predictive maintenance goes hand-in-hand with Industry 4.0 because implementations will use common components: embedded smart sensors and a combination of local and cloud-based computing resources. Vibration, temperature, electrical current and other sensors can be used to detect potential problems, supporting software models that run locally or remotely and conduct analysis in real-time.

Martin Walter, Vice President of Industry at Schneider Electric, said: “Imagine you are monitoring mechanically moving parts like robots and linear slides. By monitoring the motors on those devices, you can start understanding how quickly they are wearing. The more of that you do, the more issues you can understand and the parameters around them and with that, how to maintain them.”

Sara Ghaemi, Key Account Management Team Leader for Automotive and Industrial Systems at Panasonic, agrees with Walter’s comments. She expects the availability of cloud computing and the additional capabilities it brings in terms of powerful machine learning to be vital in driving the development of effective predictive algorithms. “We already have forms of predictive maintenance but it is going to get much more intelligent because there will be much greater access to the data needed through the IoT and greater understanding of the data through Artificial Intelligence.”

The role of AI

“AI is going to play a very important role in this part of the market and there’ll be many more advances here that will make the use of these tools easier for the end engineer or the company,” added Ghaemi. Historically, there have been attempts to implement predictive maintenance, however they have typically relied on human expertise as the volume of data and ability to interpret it were limited. The combination of AI and access to detailed longitudinal data streams collected using the IoT provides the ability to train the models rather than rely on manually-coded algorithms. A further advantage of AI is that it can call on many more sources of data than is possible with hand-tuned models.

Very often, there will be signals embedded in sensor data that domain experts do not have the time to explore and code into an algorithm. A machine learning model can easily find correlations between sensor modalities that reliably pinpoint issues and provide an estimate of time to failure. In doing so, they provide manufacturers with the ability to develop more effective and dynamic repair schedules. Thomas Dale, Engineering Manager at Omega says the ability to use previously unrealised correlations will make predictive control and monitoring more powerful.

Festo is among the companies building AI into their maintenance solutions. “About two years ago, Festo bought an AI company called Resulta which we’ve been integrating into development areas. This is driving the programmes that run anomaly detection and can then transfer this information to the outside environment. Another advance for us is a tool called Smartenance which is like an additional smart maintenance tool that enables the customer to build up a maintenance log on tablets or devices,” Sands explains.

Smartenance provides the essential link between anomaly detection and AI as well as ensuring valuable human insight is not lost. “If you start doing anomaly detection using AI, you need to feed that into somebody’s maintenance schedule. This can be sent to the person trained on the machine,” Sands adds. “The person trained on the machine then needs to think if something goes wrong, what caused it and how was it fixed. You need to keep a human in the loop to work alongside the AI to check the algorithm and adjust it accordingly so that the AI learns and progresses.”

Walter says Schneider has embedded AI into a number of its products to help monitor their condition. The use of the IoT to connect sensors from many different vendors, such as temperature sensors from Omega, with those from Festo, Schneider and others, provide the ability to manage the maintenance programme more effectively. For example, long-term analysis may demonstrate stresses that reduce reliability that can be ameliorated by changing production flows.

Local and cloud processing

Intelligence can be distributed, with machinery within cells adapting to flows without necessarily involving master controllers. This is the concept behind Molex’s Industrial Automation Solutions (IAS) 4.0 portfolio. “We’re going to allow each individual area to control itself and its own safety, and report that information to the other controllers so that they can react to changes,” said Jeff Barnes, European Distribution Corporate Account Manager for Industrial Products at Molex.

Manufacturers and integrators can incorporate customised edge devices that have embedded AI. One solution is to use the Brainium AI software running on Avnet’s SmartEdge Agile platform. The Brainium solution offers the flexibility to run some AI algorithms locally while other more complex functions, and those used to support long-term planning, are passed to the cloud for processing. Schneider’s EcoStruxure, for example, uses the large amount of data that can be contained within a cloud repository to support large-scale analysis and data mining. As processing power continues to reduce in cost, it is likely that manufacturers will take advantage of increased amounts of local intelligence to reduce the time to insight.

New service models

Jeff Barnes, European Distribution Corporate Account Manager for Industrial Products at Molex, sees the combination of predictive maintenance and Industry 4.0 technologies as enablers for new service models. One is a shift away from capital expenditure to funding models based on output, where machine builders take a more active role in guaranteeing up-time and so maximise both theirs and the customers’ revenue streams. “The builder would want to check the machine as they are likely to be on a high penalty clause if the machine fails,” he said. Predictive maintenance means they can avoid relying on hard schedules and react quickly to potential issues.

Pricing based on the output of machinery can make better use of the hardware than traditional sales models. For example, if the customer needs different capabilities, Barnes explained: “The builder can get the machine back and refurbish it into doing something else instead of ending up on the scrap heap.”

As a result, predictive maintenance can mean much more than simply reducing the cost of inspecting machinery. The more companies move to adopt predictive-maintenance solutions, the greater the level of understanding they will have of how their manufacturing systems operate. When used to its full potential, predictive maintenance and the technologies behind it will increasingly drive operational efficiency and long-term success.

Featured products

Product Spotlight

Upcoming Events

View all events
Newsletter
Latest global electronics news
© Copyright 2024 Electronic Specifier