Predictive maintenance - how AI and IoT are transforming machine maintenance

The Internet of Things (IoT) and artificial intelligence (AI) are changing the way we maintain equipment to save time and money and keep systems running. Read on to learn more about the different types of maintenance, including reactive maintenance, preventive maintenance, predictive maintenance and prescriptive maintenance. You'll also read how technology is driving innovation in the world of machine maintenance.

It's important to maintain the machines we rely on. But sometimes it's inconvenient. Just like when you have to miss your car for a few hours for maintenance. Sometimes it's not only inconvenient, but it also costs money. For example, when an entire production line has to be shut down to perform maintenance on equipment. But you know what's worse? When the production line has to be shut down unexpectedly because of equipment failure due to lack of maintenance. If you are in charge of that line, that downtime need not last longer than is strictly necessary.

What is reactive maintenance?

Reactive maintenance is maintenance performed after a component has failed. If the equipment is not mission-critical and can be easily replaced, there is nothing wrong with this approach. However, if the asset is an essential part of the infrastructure, equipment failure can potentially lead to collateral damage.

Suppose you run a restaurant and make heavy use of a walk-in freezer. If you don't notice that the freezer is malfunctioning until the food begins to thaw, unplanned expenses could include wasted food, overtime costs and additional costs for emergency work. Not to mention the cost of closing the restaurant until the problem is fixed.

What is preventive maintenance?

Preventive maintenance is maintenance that you perform on a set schedule to keep something in good condition (and avoid the need for reactive maintenance). In effect, you replace or repair something before it breaks down. You can perform preventive maintenance based on data or usage.

Data-based preventive maintenance versus usage-based preventive maintenance

• Date-based preventive maintenance
  • Preventive maintenance is strictly based on the calendar. For example, replacing the HVAC air filter every 90 days.
• Preventive maintenance based on usage
  • Preventive maintenance may also be recommended based on usage, such as tire replacement every 65,000 kilometers.

Preventive maintenance can also be recommended based on a combination of time and usage. For example, you are probably already familiar with the recommendation to change the oil in your car every 6 months or 10,000 kilometers, whichever occurs first.

The cost-effectiveness of preventive maintenance

Preventive maintenance can help you avoid reactive maintenance and the associated costs of collateral damage, but it is not always the most cost-effective form of maintenance. In some cases, maintenance is performed when it is not needed. In other scenarios, especially if the equipment is in particularly harsh conditions, preventive maintenance according to the maintenance schedule comes too late. In that situation, the equipment may still fail unexpectedly. This hard-to-find balance has led to the development of new maintenance strategies aimed at giving equipment adequate attention on an ongoing basis.

What is predictive maintenance?

A predictive maintenance model uses data to predict when equipment is likely to need maintenance. By collecting data from the equipment and other relevant data points, technicians can identify declines in efficiency that may indicate the need for maintenance. Thanks to advances in Industry 4.0 and the industrial internet of things (IIoT) has vastly improved the way data is collected and how it can be used to ensure equipment availability.

Data for predictive maintenance

Thanks to the Internet of Things, data can be read from almost any location. These can be sensors on machines that monitor energy consumption, pressure, temperature, noise and/or vibration, for example. Even cameras can be used as visual sensors. A predictive maintenance model can take into account many variables related to an individual piece of equipment's unique conditions, usage, environment and how these factors change over time.

Using AI for predictive maintenance

Using artificial intelligence (AI), a predictive maintenance model can evaluate both historical data and thousands of current data points to identify patterns and trends. These trends can be used to predict when a failure may occur so that maintenance can take place to prevent the failure.

In our walk-in freezer example, a predictive maintenance program had likely identified an impending failure by observing an increase in energy consumption and vibration. These data points can indicate long before temperatures rise and food begins to defrost that maintenance on the condenser or other components is required.

What is prescriptive maintenance?

Prescriptive maintenance takes predictive maintenance a step further. In addition to using data to predict failures, it also prescribes the most effective maintenance actions. Data in decision making includes machine data, but can include other data such as maintenance schedules, demand forecasts, parts availability, weather and other variables identified by the user.

Prescriptive maintenance can also use digital twins and/or digital simulations to digitally experiment with adjustments in the industrial process to determine the best plan of action. For example, the digital simulation may determine that part life can be extended by slowing down a machine so that the end of life coincides with the delivery of a new part or the availability of the maintenance team.

Industrial computers to support a maintenance strategy

An effective maintenance plan can have positive side effects across the enterprise, including improved product quality, reduced downtime, a safer work environment, higher return on investment (ROI) and lower maintenance costs. Whatever maintenance strategy you choose, OnLogic has a hardware platform to support it.

IoT gateways

In a cloud-based solution IoT gateways the bridge between assets and the cloud. These computers are generally located near the IoT devices and are used to collect data. Depending on needs, they can consolidate, filter, analyze and/or perform computing tasks before sending data to the cloud.

For example, industrial computers from the Helix 300 series for example, optimized for the IoT. It is available with mobile connectivity and dual LAN to collect and deliver data where and when needed.

Edge computers

If you have an AI maintenance solution on the edge implements, you can enable on-site security with low latency. For example, the Helix 600 10th generation Core i processing with expansion for additional storage, I/O or graphics cards.

Have more questions? Our team is ready to help you. Please visit contact with us.