Brought to you by Mouser Electronics
Innovation through predictive maintenance

In the past, businesses had only two ways to manage the maintenance of their machines. The first, “run until it fails,” was expensive and disruptive. The second, scheduled maintenance, was a calendar-based or use-based system that advises equipment maintenance after a set number of days or hours in operation. Of course, scheduled maintenance was the better choice; downtime could be planned when it was less disruptive.

However, servicing equipment every three months or 3,000 hours regardless of its condition could lead to unnecessary costs. Thus, predictive maintenance (PdM) was introduced. At its core, PdM involves continuously monitoring equipment using various sensors installed on the machines to gather real-time performance and machine health data. Then administrators can use that data to infer when a piece of equipment is in jeopardy and what to do about it.

John Bernet, Application Specialist at Fluke Reliability, a manufacturer of test, measurement, and monitoring products and software, has seen monitoring equipment get smaller as artificial intelligence (AI), remote analysis, and software handle processes. There has been a shift from building tools that can only collect data to creating a data ecosystem where data is collected and used for analysis to provide meaningful answers to the equipment operator. This means design teams and electrical engineers need to understand the process of getting the information from the sensor, turning it into data, and processing it in a way the software can use.

Explains Bernet, “It’s less about the data and more about the measurement and the answer you get from the data. If you know that measurement is high, that doesn’t mean a thing…. Maintenance teams have less and less time, and they’re being asked to do more and more…. They need to know what’s wrong, how to fix it, and how to get the equipment up and running.”

But before we dive into how the data is used, let’s look at the collection of that data.

Exploring the Sensor Spectrum for PdM

PdM relies on various sensors to monitor equipment and collect data on critical parameters. The selection of sensors depends on the specific asset being monitored and the data type required for analysis. Some commonly used sensors in PdM:

• Vibration sensors measure the mechanical vibrations of rotating machinery, such as motors, pumps, and turbines. They detect abnormal vibrations that may indicate misalignment, bearing wear, or other mechanical issues.
• Temperature sensors monitor the temperature of equipment components to detect overheating or abnormal temperature variations. They are commonly used in motors, transformers, and electrical connections to identify potential faults.
• Pressure sensors measure fluid pressure in hydraulic systems, air compressors, or pneumatic equipment. They detect abnormal pressure levels that may indicate leaks, blockages, or other issues.
• Accelerometers measure acceleration forces acting on equipment. They often monitor machinery vibrations, shocks, or impacts that can lead to premature wear or failure.
• Current sensors measure the electrical current flowing through conductors or components. They can monitor motor currents and power consumption and identify irregularities that may indicate electrical faults or imbalances.
• Oil analysis sensors monitor the condition of lubricating oil in machinery. They detect contaminants, such as metal particles or water, and changes in oil properties that may indicate component wear or degradation.
• Ultrasonic sensors detect high-frequency sound waves emitted by equipment. They are used for leak detection, bearing analysis, and friction or mechanical anomalies.
• Infrared sensors, also known as thermal sensors or thermal cameras, measure the surface temperatures of equipment. They can identify abnormal heat patterns, hotspots, or thermal gradients that may indicate electrical faults or insulation issues.
• Humidity sensors monitor the moisture levels in equipment or environments. They detect excessive moisture that can lead to corrosion, insulation degradation, or other moisture-related problems.
• Optical sensors use light or laser technology to measure distance, position, or alignment. They are used in alignment checks, position monitoring, and detecting changes in equipment dimensions.

The specific combination and selection of sensors depend on the asset being monitored, the parameters of interest, and the requirements of the PdM program.

Once the data is collected, cloud-based platforms provide infrastructure, tools, and services for storing, processing, analyzing, and visualizing PdM data. They offer scalable and secure environments for data-driven insights, machine learning (ML) models, and collaborative workflows. Organizations can choose a platform based on their requirements, compatibility with existing systems, scalability needs, and security considerations.

Despite the technology available, a direct visual examination of assets is often the first line in identifying visible defects, corrosion, or signs of degradation. All these methods aid in assessing the condition, detecting wear or contamination, evaluating structural integrity, and identifying potential failures in rotating machinery and other critical assets.

Unraveling the Challenges of PdM

Instead of relying on a calendar or raw data to determine a maintenance plan, PdM solutions can continuously read the machine’s condition. Then, based on the situation, engineers would pinpoint the problem and recommend a solution. However, engineers must ensure the accuracy and reliability of the models and the algorithms the system uses for diagnosis to avoid flagging problems that don’t exist or, worse, missing failures entirely.

While AI has proven to be a valuable tool in PdM, it’s important to understand that human expertise and equipment and system knowledge remain essential. Engineers and maintenance teams play a vital role in interpreting AI-driven insights, validating predictions, and making informed decisions based on AI-generated recommendations.

Bernet explains that in the past, electrical, vibration, and temperature measurement tools were all about calibration and accuracy. However, we’re now at the point where we can trust the data because the device is doing what it is supposed to be, and we can begin to find the patterns.

For example, Bernet says that Fluke has been analyzing rotating machines, motors, pumps, fans, compressors, and blowers for vibration for close to 40 years. Based on those patterns, they know what imbalances and misalignments are. That track record is essential because the accuracy of the diagnosis is based on the experience of the person that analyzed the data, built the algorithms, and created the rules that make the patterns. Only then can you trust the analysis.

Predicting the Future of PdM

As PdM expands throughout industries, engineers will first need to become expert communicators. They will not only understand how to get answers from the tools they make, but they will find ways to get those answers to the person who needs them. Bernet says it comes down to sharing information from many different test tools (e.g., vibration, electrical, thermography) and then passing along the information in a way that is easily accessible, understandable, and practical.

Secondly, different ways of collecting data are becoming more necessary. “[There is a] need for wireless remote sensors on machines because we don’t have the labor to walk around and take measurements. There aren’t enough people around and not enough time. It’s a problem in every industry,” says Bernet. “The other problem is access.”

Whether machines are located high above the shop floor, behind a panel, in a remote location, or inaccessible for any reason, Bernet says systems need wireless remote sensors, which must be tethered and talking to a common software portal. Hence, the information is accessible no matter where staff are.

Third, tools must be able to work together no matter the device, the manufacturer, and which group is using it. Bernet says that all the information should go into a familiar user interface that links that with other systems.

“In the past, it was all about collecting data, putting it on paper logs or graphs, and storing it away,” offers Bernet. “Today, we need to find a way to get that data, share it amongst everybody, and keep it electronically.”

Organizations should use all the available tools—including smartphones and tablets—and make the information accessible to everyone. That is, provide the information to anyone who needs it to keep the business running, whether they are company employees or third-party experts.

Empowering Engineers for a Predictive Future

It’s essential for engineers to understand the entire process of PdM, and not just focus on the machine. While obtaining and processing data from the sensors is important, the emphasis should be on analyzing the data and answers derived from it.

As the integration of AI, remote analysis, and software in measurement tools increases, it is crucial that engineers continually familiarize themselves with how to use these technologies effectively. But they must also understand how to extract information from the tools and share it effectively with stakeholders, so they know precisely what needs to be done.