AI driven predictive maintenance (PdM)(opens in new tab) has become a buzzword in the maintenance world, but not all PdM is created equal. While AI-driven PdM has set a new standard for performance, many solutions on the market today still rely on traditional, non-AI methods. These techniques have value, but they don’t offer the intelligence or automation many organizations assume they’re getting.
In this blog, we’re touching base with our subject matter experts to draw a clear line between what’s actually predictive and what’s just marketed as such.
The long history of non-AI predictive maintenance
Predictive maintenance didn’t begin with artificial intelligence. As Mike Cooper(opens in new tab) highlights, “for decades, maintenance teams have used tried-and-true methods to anticipate failures before they happen. AI is relatively new for maintenance teams; there are a lot of traditional forms of prediction that aren’t always talked about.”
The traditional forms Mike mentions include:
Condition-based or threshold monitoring This is when an alert is triggered for a variable like temperature or vibration that exceeds a predefined limit.
Statistical trend analysis This is when you are spotting anomalies by analyzing changes in equipment performance over time.
Physics-based models An example of this is using mathematical models to simulate how and when components are likely to degrade.
OEM guidelines and lifecycle models This is when you are estimating failure timelines based on historical data and manufacturer specs.
Each of these approaches laid the foundation for PdM, and they’re still being used today. Mike adds that “the catch with these four approaches is that they’re not truly predictive in the ways AI can be, and that’s important for the users of a CMMS to understand.”
AI driven predictive maintenance isn’t what you think
After speaking with some of our industry experts, one thing has become abundantly clear: many CMMS providers are claiming predictive AI capabilities, but it’s mostly statistical or threshold-based logic.
What we see in the market is the bold claim that a CMMS has AI-driven predictive maintenance capabilities, and that’s simply not true.
Mike goes on to explain that true AI-driven PdM uses machine learning (ML) to identify certain patterns. For example, it can:
Recommend optimal maintenance timing based on real-world behavior and not just predefined rules
“When predictive maintenance is truly AI-enabled, it changes how teams think, work, and measure success. It doesn’t just detect problems; it helps prevent them altogether, with fewer false alarms and more confident decision-making,” says Mike.
But what form can AI-enabled PdM take? What does it look like?
“For one thing, AI-enabled means it can look at specific, measurable, machine details,” Mike adds. For example, these can be:
Remaining useful life (RUL): Instead of guessing when an asset might fail, AI calculates how much time is left before a breakdown occurs.
Failure rate: AI models continuously learn from operational data to estimate how often similar failures occur.
Mean time to repair (MTTR): With earlier warnings and clearer fault diagnostics, teams can reduce the time it takes to fix issues.
Unplanned downtime reduction: AI helps maintenance teams prioritize interventions before failures happen, cutting costly disruptions.
Condition indicators: By analyzing data like vibration, temperature, and pressure, AI detects subtle shifts that precede failure, often weeks in advance.
Non-AI vs. AI based PdM approaches
Below is a table that outlines the non-AI and AI based predictive maintenance approaches:
Does predictive maintenance really work?
In short, yes. But it needs to be implemented correctly in order to work effectively for maintenance teams.
“Since PdM uses data from sensors, ML models, and historical trends to predict things like failures, that data needs to be accurate, and not only does it need to be accurate, but it also needs to be available,” Mike highlights.
Success for PdM really depends on the data quality and availability, the technological maturity (i.e., AI/ML models are trained and validated properly), and of course integration with operations (i.e., PdM is embedded into the workflows and decision-making process).
Why is predictive maintenance so important today?
Predictive maintenance is important because it changes maintenance from a reactive or scheduled task into a strategic, data-driven process. For industries like manufacturing, energy, and transportation, PdM can lead to thousands of dollars in savings and productivity gains.
Improving safety (e.g., preventing hazardous failures in critical systems)
Increasing operational efficiency (e.g., keeps production lines running smoothly)
Not all predictive solutions are made the same
In an industry crowded with predictive claims, it’s easy to assume all solutions offer similar value. But when predictive maintenance doesn’t use AI, you’re still doing most of the heavy lifting like manually setting thresholds, analyzing trends, and reacting after the fact.
With a true AI-powered system, your maintenance strategy becomes proactive, adaptive, and continuously improving.
If your PdM strategy still relies on old-school rules or statistical models, you’re missing out on the full potential of what AI can do, not just for assets, but for uptime, productivity, and peace of mind.
That’s where solutions like Fiix CMMS AI features stand out.
Predictive maintenance (PdM)(opens in new tab) is a proactive maintenance strategy that uses data analysis tools and techniques to detect anomalies and potential issues in machinery and equipment before they fail and lead to unscheduled downtime. Unlike traditional maintenance approaches—such as reactive maintenance(opens in new tab), which addresses equipment only after it fails, or preventive maintenance(opens in new tab), which relies on fixed schedules—PdM aims to predict and prevent failures through continuous monitoring and analysis.
PdM spans various industries, including manufacturing, energy, transportation, and healthcare. Human-machine collaboration in PdM involves the integration of advanced technologies, such as sensors, machine learning (ML) algorithms, and data analytics, with human expertise and decision-making. This collaboration leverages the strengths of both humans and machines to enhance maintenance processes. Machines handle data collection and analysis, providing insights and predictions, while humans interpret these insights, make informed decisions, and take necessary actions.
In this blog, we’ll explore the collaboration between humans and machines in PdM, its benefits and challenges, and future trends and representations.
The role of machines and humans in PdM
Machines and people have been working together since the great revolution, and they will continue to collaborate across sectors with new and exciting changes. As we discussed previously, machines are advancing to the point where they can self-diagnose(opens in new tab). What does this kind of evolution mean for the relationship between people and machines?
Jason Afara, our Director of Solutions Consultants, shares his thoughts on the machine and human dynamic in PdM:
“Right now, the critical thing to remember in the relationship between human and machine, at least in the context of predictive maintenance, is this … machines collect the data, analyze it, predict problems, and automate tasks, while the person has the job of interpreting the data, making the decisions on what to do with the data that’s been analyzed, implement any changes, and always look to improve,” says Jason.
Below are some common examples of the role of machines in PdM:
Data collection: Sensors and IoT devices continuously monitor equipment, collecting data on temperature, vibration, and pressure.
Data analysis: Machine learning (ML) algorithms process and analyze vast amounts of data to identify patterns, anomalies, and potential failure points. This ultimately assists in the decision-making process.
Predictive analytics: Advanced predictive models forecast equipment failures and maintenance needs based on historical and real-time data.
Automation: Automated systems can trigger alerts, schedule maintenance tasks, and perform certain maintenance activities without human intervention.
“The role of the machine in PdM is also advancing in how we can react. Previously, if something didn’t sound right with a machine, it would have been flagged and fixed by a technician who had been there for a while, or it would have been seen during a preventative maintenance schedule. Now, we don’t necessarily need to interact with the machine. We get the data right to our fingertips in something like a CMMS, and that data will get to us quicker and with even more detail,” Jason adds.
Jason indirectly references the evolution of the human-machine collaboration, something covered by experts and authors in the manufacturing and automation sector [1]. As machines advance and quicken with the quality and amount of data they can capture, people will evolve to make faster decisions.
Jason highlighted a strong point earlier regarding the role of people in human-machine collaboration that we need to circle back to. “We can’t have successful PdM without people,” Jason adds. People are essential in PdM for tasks that require critical thinking, contextual understanding, and complex decision-making.
The human role in PdM includes:
Interpretation of data: Engineers and maintenance professionals interpret the insights and predictions generated by machines, considering factors like operational context and business priorities. This task isn’t easy, and different people can look at data differently. It’s important to collaborate with your team and test your thinking.
Decision-making: Humans decide on the appropriate maintenance actions, balancing cost, safety, and operational impact. They also get support from other departments to gain access to the machine.
Adaptability:Maintenance professionals bring their expertise, problem-solving skills, and speed when it comes to challenges and managing the rapid evolution of PdM technologies.
Implementation: Skilled technicians carry out maintenance tasks, repairs, and adjustments that machines are not equipped to handle.
Continuous improvement: Human expertise is vital for refining PdM processes, incorporating feedback, and enhancing the accuracy of predictive models.
It goes without question that people and machines need to collaborate to execute PdM and enhance maintenance processes. Let’s review a case study that highlights the success of this kind of collaboration.
A practical example of human-machine collaboration in PdM
In 2017, the Massachusetts Bay Transportation Authority (MBTA) commuter rail system experienced the highest number of mechanical failures in the U.S., primarily involving locomotives [2]. In response, the MBTA created a program to upgrade its legacy locomotives, incorporating AI and machine learning to predict maintenance problems before they lead to breakdowns.
Ryan Coholan, the MBTA’s chief railroad officer, recognized the potential of data analytics to improve commuter rail performance. The MBTA previously collected diesel locomotive oil samples and conducted primary analyses, but an approach was needed to relate sample values to maintenance problems. Coholan enlisted Mike Jensen from 4Atmos Predictive Analytics to explore using machine learning models to predict engine breakdowns. The MBTA had extensive historical data from oil samples, which was used to train the model.
The trained machine learning model analyzed elements within the oil to predict engine issues. It could identify high-probability engine problems within the next fifteen days, providing specific predictions like turbocharger or injector failures based on oil composition. The model’s success led to the implementation of regular oil analysis, now conducted every ten days. The project, named Project Velocity, integrated AI-driven oil analysis into the MBTA’s daily operations. The MBTA established an in-house oil lab, enhancing its maintenance capabilities and attracting interest from staff.
Benefits of the MBTA’s PdM model
The predictive maintenance model provided substantial benefits:
Cost savings: Predicting and preventing failures before they occurred saved significant time and money.
Operational efficiency: Preventing on-track failures avoided passenger inconvenience and improved service reliability. Awareness of data at near real-time, remotely from the machine. Ability to change context based on the available data.
Investment and practices: The success in predictive maintenance prompted a $40 million investment in locomotive rehabilitation and changes in maintenance practices.
The program dramatically increased the mean miles between locomotive failures and improved on-time performance for the commuter rail system.
Benefits and challenges of human-machine collaboration
As we’ve seen in the MBTA’s project, human-machine collaboration in PdM can be very effective and beneficial. Still, it is important to highlight the challenges of this kind of collaboration. Below is a table to illustrate some of the benefits and challenges of this collaboration:
Benefits
Challenges
Increased accuracy: Combining machine-generated insights with human expertise enhances the accuracy of predictions and maintenance decisions.
Data quality and integration: Ensuring high-quality data from diverse sources and integrating it seamlessly is complex and requires robust systems.
Reduced downtime: Proactive maintenance based on precise predictions minimizes unplanned downtime and extends equipment life.
Skill gaps: There is often a gap between the technical skills required for PdM technologies and the existing capabilities of the workforce.
Cost savings: PdM reduces maintenance costs by preventing failures and optimizing maintenance schedules.
Change management: Implementing PdM requires a cultural shift and buy-in from all levels of the organization, which can be challenging.
Improved safety: Early detection of potential issues ensures a safer working environment by preventing hazardous equipment failures.
Cost and investment: The initial investment in PdM technology and training can be substantial, posing a barrier for some organizations.
By addressing the challenges and capitalizing on the benefits, industries can achieve greater efficiency, safety, and cost-effectiveness in their maintenance practices.
Future trends of human-machine collaboration in PdM
As we reviewed earlier, the future of human-machine collaboration is an exciting topic; when data becomes faster to access, people evolve to respond quickly, too. Some trends shaping the future of PdM include the integration of IoT technologies and Industry 4.0, as well as the adoption of emerging technologies like digital twins.
“IoT devices like sensors are collecting real-time data from equipment, but these devices are going to allow us to monitor data down to the second, not down to the hour, or even several days of data,” says Jason.
It’s not necessarily a psychic prediction to say that the future of IoT devices and Industry 4.0 will involve faster, better, and more responsive data readings on machines. We know from other technological evolutions that product development focuses on things like speed of data transfer and lowering the cost of manufacturing, and therefore, a lower price to purchase said technology. In the later point, reducing the cost means more access, which means more manufacturing facilities can implement smart factories.
A confident prediction is the fusion of physical equipment with digital systems, which enables real-time interaction and autonomous decision-making. Machines equipped with cyber-physical systems (CPS) can self-monitor and communicate their status, optimizing maintenance schedules. Additionally, the use of digital twins more proactively will prove to evaluate the outcomes of various maintenance actions, leading to more informed decision-making. Digital twins and machine learning will likely work with artificial intelligence (AI) to automate anomaly detection, learn even more in-depth about a team’s maintenance process, and evaluate current and ongoing strategies.
Human-machine collaboration in PdM represents a transformative approach
The evolution of human-machine collaboration in PdM reflects a continuous shift towards more proactive, efficient, and intelligent maintenance practices. From the early days of reactive maintenance to the present era of AI-driven predictive maintenance, the collaboration between humans and machines has been instrumental in enhancing equipment reliability, reducing costs, and improving operational efficiency. As technology continues to advance, this collaboration will only deepen, leading to even more innovative and effective maintenance solutions.
Recent Comments