An effective predictive maintenance (PdM) system helps prevent 75 percent of potential failures, saving 12 percent more in maintenance costs each year than a preventive strategy alone.
In fact, this benefit is a major factor driving the rapid growth of the internet of things—half of manufacturers have already implemented PdM, with another 40 percent planning to by the end of 2018 (full content available to Gartner clients).
But establishing the correct setup to gather and analyze data lays the foundation for a more efficient maintenance system that can learn to spot patterns and prevent expensive breakdowns.
If you start small, track the right data and setup alerts correctly, you’ll experience the greatest benefits PdM can offer.
The primary PdM steps include:
Perform ABC Analysis on Your Critical Assets
To effectively prevent failures, you have to understand them. ABC analysis is a popular method used to determine the failures for each asset that have the greatest negative impact on your company’s operations.
The analysis takes into account three factors, on a 1-10 scale, for each failure mode:
- Occurrence: The frequency of the failure
- Detectability: How difficult the failure is to detect
- Severity: The impact of the failure on production or operations
Starting with your most important machines, record any common failures your maintenance team has discovered already and enter them into this tool, which will produce a risk priority number (RPN).
From there, it’s simple—just prioritize the assets with the highest RPNs. You should quickly recognize the machines that truly need close care and those you shouldn’t waste valuable labor time inspecting as often.
Establish Data Streams and Detection Alerts
With your most critical machines identified, you have a head start in finding the best method to collect and leverage the condition data used to make failure predictions. Take a look at your most damaging failures in the list and, for each, ask: “Which symptom of the failure is detectable? Which is detectable early?”
Let’s say you have an AC motor like the example in the tool, with various subcomponents that could wear out and cause a failure. Above, worn bearings is one cause of failure based on an observable degradation, but you may also discover that vibrations in the motor occur around the same time. This is a good opportunity to determine which comes first.
The result should be to measure normal operating vibration levels and implement sensors that alert you when they begin to increase. By inspecting the machine at this point in time, you can avoid bigger, more expensive failures.
Sensors exist for nearly any kind of asset and condition, from detecting moisture to ultrasonic sound waves. You can use these sensors to eliminate unnecessary repair work orders based on a calendar and maintain machines only when they indicate a problem.
Modern PdM software can even leverage artificial intelligence or machine learning, which allows the system to understand and predict patterns data anomalies create. This more advanced technology is leading the way for a greater understanding of why machines fail.
The best part? Sensors are getting cheaper every day—for example, a water presence sensor can be implemented for just a couple hundred dollars each. When an asset breakdown can cost thousands to remedy, it’s a waste of money not to invest.
Set Up Alerts to Generate Detailed Work Orders
The final step to make sure this technology catches failures early is to alert the appropriate people at the right time. That’s the distinct feature of condition-based predictive maintenance.
When the data stream from a sensor exceeds the thresholds you established, a computerized maintenance management system (CMMS) or enterprise asset management (EAM) suite can automatically generate a work order with the asset and its condition.
Maintenance software providers allow you to customize the alerts in a few ways:
|WO status||Notify specific people when work orders are generated, when they start and are closed out.|
|By asset||Prioritize work orders for critical assets and send alerts only to the technicians who work on those machines.|
|How it’s delivered||Assign work orders to groups or individual workers through the software, via email or by text.|
Right alert at the right time—got it. But your technicians need the right information too.
We recently shared some best practices for work orders, which included adding failure codes, root causes and solutions. These details help technicians in the field understand the problem with context so they can make repair decisions based on the true cause of failures instead of assumptions.
When completing work orders moving forward, you can tweak failure codes and improve solutions as you gain more experience and learn the quirks of individual machines.
How Can I Make Sure I Catch Anomalies?
Identifying critical assets, getting the right data and generating work orders cover the key components of a predictive strategy. But establishing the most accurate data thresholds to trigger work orders requires intimate knowledge about the anomalies that occur when streaming asset data.
Here are some best practices to be able to recognize anomalies in the data:
- Start with a few priority assets first. You’ll be better prepared to understand the data of a machine you’ve worked with for a while—start there to improve your asset data literacy in the beginning.
- Improve failure codes and solutions over time. Put an emphasis on consistently recording information during inspections and repairs to discover more nuance about asset data.
- Maintain your regular PMs as you ramp up. Your PdM strategy will improve over time, but meanwhile, stay on top of emergency repairs by maintaining your scheduled PMs and inspections to avoid breakdowns as you grow.
If you’re ready to adopt a new maintenance system, we have profiles, user reviews and demos for more than 130 systems. And for more guided help, call (844) 689-4876 for a free consultation with one of our advisors who can narrow down your best options in 15 minutes or less.