Predictive Maintenance in Manufacturing: Benefits, Trends and Examples

The manufacturing industry is probably the biggest application for PdM. This approach emerged back in the 1990s although it hasn’t gained immense popularity until recently. This happened because of the introduction of the Industrial Internet of Things (IIoT), Machine Learning, Big Data, and Cloud Computing, which made PdM more available and affordable to manufacturers.

Types of maintenance in manufacturing

As manufacturers struggled to keep up with the roaring demand, they tried different maintenance strategies. One of them was reactive maintenance, which embodies the mindset of “don’t fix it until it’s broken,” yet it ends with an increase in unplanned maintenance and downtime.

Here we can also speak about corrective maintenance, which idea is to identify and repair a failure so the equipment could perform as earlier. The two concepts are very similar. The only difference is that reactive maintenance is about fixing the equipment when it’s completely broken, whereas corrective maintenance means the equipment is still running but noticeably deteriorating.

Unlike reactive or corrective maintenance, preventive maintenance allows scheduled maintenance activities to perform before the problem occurs. It increases the longevity of the equipment, but it’s also very costly. And it still doesn’t eliminate the risk of equipment or its components breaking down randomly.

Condition-based maintenance expects equipment monitoring, and based on the data collected, maintenance activities are performed right before the failure occurs. It’s similar to predictive maintenance, which doesn’t let equipment degrade. Based on the sensor data and trends, the latter predicts degradation and failures as early as possible to extend the equipment lifecycle.

Industrial predictive maintenance seems a better solution to manufacturing companies due to its ability to minimize downtime and cut maintenance expenditure. Besides, tight deadlines and profit margins in today’s manufacturing sector made unplanned downtimes even more unwelcome.

The survey of 300 industry experts organized by Frenus GmbH Darbi College proved that 80% of the respondents consider PdM essential in the manufacturing industry. 77% of them also believe that PdM is a real need to stay competitive in today’s business environment. (See more detailed results of this questionnaire below.)

Benefits of predictive maintenance in manufacturing

The list of the benefits of PdM in manufacturing is exhaustive, so we’ll focus only on the most important:

• Better ROI: In the case of PdM, your return on investment (ROI) is huge. First, you save on maintenance costs since you don’t perform regular maintenance tasks that may be unnecessary. Secondly, PdM solves the problem of unplanned downtime and reduces a planned one. Repairing tasks take less time because data scientists already know where to search for faulty components and why they’ve broken down. Thirdly, manufacturers can order new parts or machinery in advance. Thus, they don’t waste time waiting for them or spend extra on stocking in case of planned maintenance.
• The high lifespan of machinery: Deloitte states that PdM increases machine uptime by 10 to 20%. First of all, the system is tracking the asset performance, so it will notice a problem before the equipment reaches the stage of serious damage. Secondly, in PdM, the useful life of machinery is long because the equipment or its parts work out all its possible time instead of being replaced regularly (as it happens in preventive maintenance).
• Reduced waste: Manufacturers can also increase their bottom line revenue by reducing waste. Sub-optimal operation (when your equipment is on the verge of breaking down) usually leads to wasteful production, such as in raw materials, energy, machine time, and labor costs. PdM can warn businesses about this issue before it even happens.
• Better performance: PdM reduces the time needed for repair as well as the frequency of repairs in general. Therefore, manufacturing organizations can operate more efficiently. Over some time, factory conditions should improve, while equipment should break down more rarely.
• Improved operator safety: Getting early warning signals about faulty equipment can prevent injuries in manufacturing. Big data analysis helps to eliminate safety risks on daily basis, plus to identify potentially dangerous conditions by keeping track of machinery for longer periods.
• Asset protection: While fixing an equipment problem, manufacturers can sometimes affect other parts of the machine. PdM can help you notice any abnormal behavior after a repair so you can solve it early.

Basically, we can say that any PdM system is composed of four parts:

1. Installed condition-monitoring sensors that gather and send further real-time performance data as well as machine health info. It’s possible due to IoT technology that combines the connection between machines, software solutions, and cloud technology
2. Data collection pipeline that helps with collecting data from sensors in a raw format for further data analysis and processing
3. Predictive ML models that are trained on the historical data and fed with real-time sensor data to obtain failure predictions
4. Analytics and monitoring software that helps with data analysis, system events monitoring, and scheduling human-machine interactions.

In addition, there are two common approaches to predictive maintenance that manufacturing companies can choose from if they want to implement PdM:

Rule-based predictive maintenance

The idea of rule-based PdM, which is also known as condition monitoring systems, lies in relying on condition monitoring sensors. Those sensors continuously collect data about equipment and then send alerts when a specific rule has been activated, in line with predefined protocols.

A special feature of rule-based AI systems includes close cooperation between production teams, engineering, and customer service departments. This cross-department collaboration is necessary to understand the direct and indirect causes that eventually lead to equipment breakdown.

Once a manufacturer knows those causes, it can create a virtual model of its connected system where they outline the behaviors and inter-dependencies between its different IoT elements. As a result, if the temperature increases above the predefined level in a smart factory, then the system sends an alert message to the team.

From the mentioned above, rule-based PdM delivers some level of automation. However, it’s still too dependent on the team’s understanding of what components or environmental events to monitor.

Predictive maintenance with machine learning

A different approach is to build AI predictive maintenance in manufacturing based on machine learning algorithms. In this scenario, ML-based systems define their own set of rules proceeding from data inputs. They take the correct data and work based on a probabilistic approach.

Precisely, an ML model uses and learns from the data generated from IIoT sensors historically and in real-time. This way, the model knows the equipment’s normal behavior and is able to detect anomaly data and events. It also helps to find correlations and make predictions for the production team to take action timely to remove the potential defect.

Here’s one more great thing about ML-based predictive production maintenance: the model can dynamically adjust to new data and make sense of what’s happening in real time.

If you are wondering about the predictive maintenance implementation process, below we explained how to do it step-by-step.

Like any other relatively new technology, predictive maintenance for manufacturing has caused lots of hype about its adoption. Still, there are a number of specific reasons why manufacturers are looking in the direction of PdM. Below read seven trends of PdM in manufacturing:

1. Plug and Play technology

One of the bottlenecks for introducing advanced technologies in manufacturing, including a predictive maintenance approach, is its dependence on legacy equipment. This problem is especially relevant to large companies where much of the machinery is not equipped with connectivity to send real-time info.

We’ve been working with Intelliarts for eight years. Originally, we were reaching out just for help on the engineering side. But we found a true partner. We think of them as true members of our team, and depend on them for really the day to day of the business.

Bob Dilan

PnP (Plug and Play) devices can be an optimal solution in this case. These include ready-to-use computer equipment that can connect legacy machines to computers. As a result, PdM becomes available to manufacturers, without the need to replace older equipment. Aside from this, PnP technology doesn’t require any specific knowledge to be installed, so anyone can do it.

2. AI and machine learning

3. Supply chain cooperation

Another popular trend is to extend predictive maintenance of machines to supply chain management, which unlocks new benefits to manufacturing companies. For one thing, PdM won’t only monitor the asset lifespan but will consider production schedules and choose the most optimal time for maintenance activities. For another, the system can help manufacturers order new parts for replacement automatically.

4. Thermography checks

Since heat is one the earliest indicators of equipment issues, manufacturers are starting to use infrared PdM checks on temperature. This becomes possible with thermographic analysis that expects using infrared scanners to detect wear and rusting of machinery that remains invisible to the naked eye.

5. Digital twin coupled with PdM strategy

A digital twin is becoming one more mainstream technology in the manufacturing industry, helping businesses to streamline their operations. If a manufacturer has a detailed virtual version of its company, they can test processes before their introduction, make plans about new equipment installation, and so on.

To make use of this technology to its fullest, you can combine digital twins with predictive production maintenance. For example, Tech27 tells how a digital twin paired with PdM helped to save an oil and gas production plant as much as $360,000 due to predicting a potential plant outage. The two approaches make a perfect pair when there is a target to predict and when high-quality operational data are available.

6. VR and AR

Next to diagnostics, VR and AR are also used for personnel training, remote assistance and equipment monitoring, improved documentation and knowledge management.

7. PdM As A Service

Last but not least, PdM as a service is one more interesting trend that draws the attention of manufacturers. This hybrid approach combines the effort of a dedicated service team and software so to make PdM instantly accessible to manufacturers.

This approach is particularly valued by original equipment manufacturers. OEMs produce parts and equipment for other manufacturers while PdM as a service enables them to collect real-time data from their clients’ equipment and, thus, improve their operation.

Industrial predictive maintenance is a must for those manufacturers that want their assets running smoothly and without costly downtimes. Here are the most common use cases why manufacturing companies choose predictive maintenance:

• Real-time asset monitoring: Predictive maintenance is a great way to extend the useful life of their machines. Modern plants usually have expensive equipment while PdM can optimize the use of machinery via real-time condition monitoring.
• Timely diagnostics: PdM helps to detect and diagnose faults timely so manufacturers avoid breakdowns and idle time for machinery. What’s more, the technology warns technicians before the failure even happens.
• Maintenance cost reduction: With PdM, businesses stop performing scheduled maintenance activities, which always brings extra expenses. At the same time, they spend less on diagnostics.
• Reduced waste: PdM also helps to optimize the use of resources, machine time, energy, and labor costs.

Learning from peers is better than entering the uncharted waters of PdM without researching the practical experience of others. For this reason,  let’s take a look at predictive maintenance examples of three global manufacturers that have already implemented PdM into their operations.

1. Frito-Lay snack food manufacturer

PepsiCo’s subsidiary Frito-Lay has a positive experience of introducing a robust program of PdM technologies. According to its manager’s presentation at the Leading Reliability 2021 conference, PdM helped the company reduce planned downtime to 0.75% and unplanned one to 2.88%. The manufacturer also improved its operations in a range of ways thanks to PdM.

Specifically, the technology warned and helped to prevent the failure of a PC combustion blower motor. If the company hadn’t responded as early as it did, this failure could have caused the shutdown of the entire potato chip department for an unknown amount of time.

One more example includes the early detection of increased acid levels in oil samples, which most likely warned about oil degradation. Unless it hadn’t been fixed, this problem could have led to downtime in Cheetos Puffs production and low sales.

2. Mondi packaging & paper manufacturer

Another interesting use case for predictive maintenance in manufacturing includes the achievements of Mondi manufacturer that produces paper and packaging products. The company has implemented PdM specifically to avoid abnormal shutdowns of its plastic extruder machine in Munich’s plant, Germany. A single failure of this equipment cost the manufacturer as much as €50,000 in cleanup and lost revenue, according to Rainer Muemmler, the speaker at PAW Industry Virtual Conference.

By estimations, the introduction of PdM allowed Mundi to save from €50,000 to 80,000 mainly due to reduced operating costs and less waste generated by the machine.

3. Alumina manufacturing facility

Noranda Alumina LLC specializes in alumina product production. The manufacturer implemented PdM in 2019, and since then, the technology has helped the company save approximately $900,000 in bearing purchases as well as reduce the downtime, as this was mentioned at the Leading Reliability 2021 conference.

One more acquirement of this alumina refinery included the impressive growth in the grease completion rate from 67% in 2019 to 92% in 2021, which the manufacturer also attributes to its implementation of PdM.

Most manufacturers may find a transition to predictive maintenance rather overwhelming. To make it smoother, follow the next strategy:

1. Analyze historical data. Before you go all in predictive maintenance, you’d better know your current state of affairs. Check your occurrences of downtime, their frequency, and scale; review defects, their types, and costs; finally, look closely at reactive and preventive maintenance activities you completed for each asset.
2. Identify priority assets. Use the insights above to get a clear understanding of the most critical assets to your operations. Your next step as a manufacturer is to install condition-monitoring devices and IoT sensors on this equipment. No need to install all available devices on the market — a thermal imagery camera, vibration meter, and oil measurement could be your starter pack.
3. Start a pilot project. The same as with assets, you don’t have to try the uncharted waters of PdM with a big and costly project. Start with less and initiate a pilot project with one or two most suitable assets and machinery.
4. Move on and scale. Once your pilot project proves successful, you’re ready to scale up the effort. The good news is that PdM can be scaled up in quick tempos, especially if you pair up with professionals. From 6 to 12 months, your company can grow from several connected machines to a smart factory.
5. Have a detailed plan ahead. We also recommend having a clear vision of what you want to achieve. Get a detailed plan of implementation, a realistic budget and time expectations. Also, think about the goals of your PdM project, try to make them measurable, and track the progress.
6. Seek professional help. Not to waste your company’s resources, get a professional ML team on board to help your company implement PdM successfully. You can also ask expects to analyze the weak points of your PdM journey if the results of your pilot project weren’t as promising as expected.

Sure, a manufacturer could meet a range of challenges along the way to building a smart factory. Here are the key stumbling blocks to PdM:

1. Outdated equipment and machines. Sometimes, it happens that industrial machines are too old to be equipped with smart sensors. In this case, a manufacturer has to figure out how to make them compatible, for example, by updating or replacing the machines.
2. The data issue. Having no experience with PdM, manufacturers might be stuck on what data to collect and how to gather and store them right.
3. The lack of expertise. Manufacturers also need a tank of ML knowledge to make sense of their messy historical data; choose the most appropriate ML algorithms out of dozens available; and train and fine-tune the algorithms in the most efficient way.
4. Digital mindset and culture. Having the right technology isn’t enough for a manufacturer’s success. Your corporate culture should have data-driven decision-making, focus on innovative technologies, and cross-functional collaboration between the IT department and other teams, including operators on the factory floor. Staff education is an important puzzle a manufacturer has no right to ignore if they want to implement a PdM solution.

If you read this far, you probably already wonder whether and how predictive maintenance could be useful to your manufacturing organization. To research this question more thoroughly as well as to get to know how to shift to PdM smoothly, you are welcome to read our Predictive Maintenance White Paper.

Optimizing a maintenance strategy is difficult but important for manufacturers. The most reliable way to do it is to implement predictive maintenance. PdM will help you define the best time to do maintenance activities, warn you about any asset problem and potential downtime, and improve the useful life of the equipment.

We at Intelliarts love to help companies to solve the challenges with data strategy design and implementation, so if you have any questions related to Predictive Maintenance in particular or other areas of Machine Learning — feel free to reach out.

1. What challenges should manufacturers be aware of when implementing predictive maintenance strategies?

Predictive maintenance implementation could be challenging. Keep in mind the next difficulties: the need to collect and manage massive amounts of data; the employee skill gap and, thus, the need for personnel training; privacy and security concerns.

2. What is the ROI (Return on Investment) for predictive maintenance in manufacturing?

The ROI largely depends on the plant performance, data availability, and the industry the manufacturer operates in. Still, we can say about ten times the cost on average.

3. What are some common misconceptions about predictive maintenance that manufacturers should be aware of?

Be aware of the next misconceptions about predictive maintenance: PdM is too expensive and, thus, it’s for large manufacturers only; deploying IoT sensors is enough for PdM; PdM systems are too complex; PdM is a danger to human jobs; PdM is one-size-fits-all-solution.