When looking at ways to improve plant operations, consider adding predictive maintenance (PdM) to your maintenance strategy. To help you think about PdM and implementing a PdM solution at your organization, we have compiled some ABCs of Predictive Maintenance.
Before a catastrophic failure occurs, there are signs of degradation. By analyzing large amounts of data and identifying deviations, organizations can detect potential and hidden failures before they become a catastrophic failure. Without technology to help, digesting large amounts of data and manually analyzing the data—even if done in a spreadsheet—is difficult, if not impossible, with available resources. So when an organization decides to implement a PdM solution, an organization can rest assured that its solution will continuously monitor the equipment and alert them to anomalies in real time, saving time, energy, and money.
Plants have access to more and more data. But what should you do with all that data? The value in big data is not having it; the value is in how you use it. As a result, when considering using data more, the first question to ask yourself is: what challenge am I trying to solve? If a major challenge is unexpected failures and unscheduled downtime, consider a PdM solution. By leveraging predictive maintenance solutions that capture, analyze, and visualize data, organizations will be able to focus on what is important based on current equipment health.
A type of predictive maintenance, condition-based maintenance (CBM) helps an organization prioritize equipment maintenance based on its status. Rather than replacing a valve when it does not need to be, CBM helps organization focus on equipment that is having issues now. To do this, CBM utilizes data from the equipment like vibration and temperature. When reviewing CBM options, there are highly specialized types including vibration analysis and oil analysis to monitor critical assets as well as more generalized solutions that monitor the entire plant and systems.
When technology helps organizations shift their mindset, organizations see improvements in their operations and how they do business. For maintenance, using a PdM strategy, organizations can shift their maintenance and reliability teams from being reactive to proactive. This does require a culture change. Therefore, be intentional when incorporating PdM technology into your organization. Think about the required buy-in and training needed to make your digital transformation effective and lasting.
Data needs context to be helpful. To add an additional layer of context, PdM solutions using machine learning and advanced pattern recognition can provide expected values of equipment operation. When you compare real-time values to expected values based on the current operating environment, you focus on the equipment that deviates from what is expected. Traditional setpoint alarms are static—does the temperature sensor exceed X°F? In contrast, by comparing to expected values, you will have a dynamic comparison of where the temperature is versus where it is expected.
The goal of predictive maintenance is to limit the risk of functional failures. By reducing functional failures that could have been avoided, plants improve safety and reduce costs. When failures do occur, plants are in a reactionary mode. They are no longer running the plant; the failure is running the plant. In contrast, by detecting potential and hidden failures in advance, organizations can address the issue before it turns into a functional failure. The goal of PdM then is for organizations to limit their time spent reacting to functional failures, so organizations can spend more time planning and optimizing.
In any plant, the systems, subsystems, and equipment relate to each other. Rather than looking at a piece of equipment or sensor in a vacuum, organizations can improve operations by looking at information holistically. By grouping related sensors and equipment together, an organization can more easily determine root cause of the issue, whether it is a sensor failure or equipment failure. Grouping also mimics how humans think. When we think about a car, we typically think about the car itself, not the individual components or systems. Both data historians and predictive maintenance solutions have followed suit, grouping data by assets not sensors.
In order to take advantage of predictive maintenance, you need a strong data foundation. Garbage in, garbage out applies. To ensure you have the right data when you need it, invest in a good data historian. A good data management solution is fast, reliable, accurate, intuitive, and easy-to-use. It should also allow for the centralization of data, so that all layers of the organization have access to data for data-informed decision-making. With a centralized data historian and trust in the quality of equipment data, your organization can take advantage of more advanced analysis tools like predictive maintenance.
The purpose of technology in the process industry is to improve operations by leveraging actionable intelligence. To take advantage of predictive maintenance, organizations need to focus on insight. Insight goes well beyond data and information. For example, information alone may be that a turbine bearing vibration is 81.6 um. Insight is that this vibration, as well as other related vibrations, are behaving abnormally. With insight, organizations are better able to identify the cause, for example an improper bearing seating. With accurate insight from data about equipment health, organizations can then prioritize maintenance and reduce costs.
Assets and equipment are necessary for production, but also represent a safety hazard. The process industry has created a safety culture, but unexpected failures potentially put members in harm’s way. With predictive maintenance, organizations can detect hidden and potential failures in advance. By reducing functional failures, organizations reduce the risk of accidents on site. For any organization, output and revenue are of course important, but it is the safety of its employees and surrounding area that is a core focus for the organization.
Organizational knowledge is key to an organization’s success. Your people represent your difference. This know-how is valuable and should not be lost. As the workforce changes, technology can help an organization maintain its know-how and can assist your experts. The first way PdM solutions help is by becoming your organizational memory. Fix an anomaly and then save that information into your PdM solution. The second way PdM solutions help is by freeing up your members’ time. Let your PdM solution work for you by monitoring and analyzing equipment data 24/7.
Machine learning enables organizations to transform their data into information. As a part of artificial learning, machine learning (ML) is further divided into 3 categories: unsupervised, supervised, and reinforcement machine learning. Advancements made in ML and AI have enabled organizations to take their ocean of data and turn into actionable intelligence. When evaluating PdM solutions, you do not need to understand fully the ML algorithms but should verify its value. One way of doing this is to ask for examples of catches found or case studies using the PdM solution.
One-way PdM solutions work is by creating models of your equipment and plant based on your data. There are different types and ways of modeling including empirical modeling. By starting with normal, fault-free historical data, a PdM solution can create models of your equipment and plant. The beauty of these models is that they are based on how your plant and equipment operates, not an ideal version only found in your OEMs testing lab. With the models, your PdM solution can generate expected values based on your current operating environment.
Check back in the coming weeks for N – Z! And feel free to share your own ABCs of PdM on our LinkedIn page.