One of the truly great results of the growth of (Internet of Things) is the fact that machine data can be leveraged to limit the operational costs and impact of downtime, both planned and unplanned. This is otherwise known as predictive maintenance.
Predictive maintenance—or PdM, for short—is a method for anticipating maintenance requirements in machines on a factory floor. By analyzing operational data from the machines, patterns emerge that will allow operators to develop an understanding of failure modes and predict when maintenance will be required on any given unit, allowing for it to be planned during less-costly times.
In the past, manufacturers would rely on reactive maintenance and other maintenance strategies, otherwise known as the “if it ain’t broke, don’t fix it” method. You can well imagine that servicing machines only when they broke down was a huge cost, both in terms of unplanned downtime and the potential impact to other parts of the machine, as well as the quality of the output for the time that the part was failing.
Over time, companies sought to move away from reactive maintenance and to implement preventive maintenance strategies. But preventive maintenance relied on averages and didn't reflect the current or real-time condition of equipment.
While preventive and predictive maintenance is a step-up from reactive maintenance, there is a clear winner. Implementing predictive maintenance relies on specific information pulled from each machine to detect potential problems. An example would be vibration analysis.
A model that uses a baseline to collect predictive maintenance data for a machine will be able to detect changes, such as an increase in vibration in a specific part, which could be caused by damage or the introduction of a foreign object. Deviations from the baseline allow operators to predict a need for maintenance before the problem becomes serious, resulting in equipment failure.
Unlike preventive maintenance, a predictive maintenance program uses the data generated by equipment and sends the data to the cloud. Like the production and business process data, real-time condition-based maintenance can be analyzed to detect patterns and trends on the machine's health or the life of its parts or tools and to lower maintenance costs.
Often, OEM equipment may state an expected lifecycle for a part or tool. However, this is based on averages from across the industry. It does not consider light-duty applications where parts may wear slower than expected or heavy-duty applications where they may wear faster.
Advanced analytics in production and maintenance monitoring platforms use trends and condition-based maintenance to plot the actual time-to-wear or predict failure modes. Predictive maintenance programs can be developed around these insights to reduce downtime and control costs.
Robust predictive maintenance programs can use the platform to send alerts to maintenance staff when a failure is imminent, reducing downtime from waiting for technicians to arrive after the fact. They can also work with factory production information, such as expected changeover times to make repairs during scheduled changeovers or planned downtime for cleaning.
Scheduled alerts can be extended from technicians to MRO stockrooms so parts can be staged to avoid more downtime. And the parts can be automated for reorder against established min/max part counts, which are also prescribed by the system. As part of its overall maintenance strategy, each company can tailor its predictive maintenance techniques to its own needs.
From a cost-saving point of view, the advantages of a predictive maintenance strategy include optimized planned downtime and minimized unplanned downtime. A predictive maintenance program will also optimize employee productivity and equipment lifetime. Using advanced analytics and machine learning algorithms, preventive maintenance cannot be understated!
Planned downtime can encompass everything from machine cleaning and oiling to the replacement of parts that are known to fail on a regular basis. This kind of preventive maintenance reduces the risk of unplanned downtime. Just like taking care of your computer and sweeping it for viruses or keeping other appliances clean in your home, you’ll get more efficient and better-quality output from a well-serviced machine.
Thanks to the data collected in machine operations, predictive maintenance can be scheduled regularly and at times that will have the least impact to order production. There is also the added benefit that adequate maintenance of this nature will invariably extend the life of mechanical equipment that would be difficult and costly to replace. Maximizing uptime and the life of a component in a predictive maintenance program will ultimately result in significant cost savings.
According to a Wall Street Journal post, “Unplanned downtime costs industrial manufacturers an estimated $50 billion annually.” Using predictive maintenance to limit this cost is critical in highly competitive manufacturing industries.
In as much that scheduled preventive maintenance can ensure that machines run smoothly most of the time, monitoring machines digitally collects reams of data that, when analyzed, will show patterns on any given machine. This kind of pattern detection, based on historical data, can help to identify a machine that is likely to experience an outage and for which maintenance can be planned proactively.
Being able to monitor a machine’s efficiency, output, and quality over time will reveal data that will identify when a machine requires maintenance, as noted above, but will also help identify when a machine is reaching the end of its life. Reactive and preventive maintenance cannot do this.
As machines age and depending on their level of use, the maintenance schedule will change, which can be managed through predictive maintenance. Parts of the machine will respond to production stress differently over time. The eventual increase in maintenance that is predicted through data patterns will reveal when a machine is reaching a tipping point on cost vs. performance. The need to eventually replace large parts of a machine or the entire unit is made manageable by forecasting that need and planning for it, both from a cost/budget and time/effort point of view.
There are many ways that predictive maintenance optimizes employee productivity. Firstly, let’s just look at the cost of the labor itself. When repairs are scheduled, the amount of time needed for repair is reduced because of a smaller number of component replacements instead of entire equipment replacement. Also, the frequency of repair for critical failure of equipment will be reduced, and the number of “critical callouts” will be greatly reduced.
From the employee’s perspective, predictive maintenance will lead to reduced breakdowns and accident-avoidance systems. These can alert or even halt equipment when there is a danger to a worker, dramatically improving factory conditions and minimizing worker injuries.
Furthermore, downtime and operations with sub-optimal parameters not only impact output but also employee morale. It is stressful to rush to solve problems when they arise. Predictive maintenance minimizes such instances.
The advantages of predictive maintenance we’ve covered above, in the end, all have the same goal: increasing the bottom line. With less maintenance on good components and quicker repair of faulty components, repairs can be more effectively handled, thereby reducing repair time. One of the most comprehensive studies on the potential of industrial analytics like predictive maintenance was conducted by McKinsey in 2015. They uncovered the opportunity for the following improvements:
Since planned maintenance is based on a schedule, there will be cases when maintenance tasks will be performed when they are not needed. Predictive maintenance can prevent such inefficiencies.
Sub-optimal operation that is not detected can result in wasteful production. Raw material, energy, labor costs, and machine time get wasted in such instances. Predictive maintenance systems can uncover issues that can result in waste before they arise.
Once data collection becomes automated, new insights on process optimization opportunities can be uncovered daily through advanced analytics.
There are numerous types of predictive maintenance technology used in a robust data-driven predictive maintenance program. This condition monitoring equipment can be used to create a predictive maintenance solution for an operation. These technologies include:
Heat is almost always a by-product of a manufacturing environment. But it is often predictable for each machine or job type being run. PdM platforms can map these heat patterns by machine or job and analyze spikes in temperature to determine approaching problems. Infrared thermography can monitor and measure temperature in equipment such as motors, bearings, or other friction surfaces. It can also help uncover "hot spots" in electrical cabinets and detect insulation failure. Infrared thermography measures the temperature and displays the unit as a picture of the entire unit being measured. The monitoring platform can store and analyze these images to detect problems and identify trends under specific conditions.
Some predictive maintenance monitoring platforms can use high-frequency signals to determine the equipment's condition. The same principle can be applied using airborne sound. By capturing this acoustic noise, breakdowns can be detected. Advanced machine learning algorithms are used to improve the predictive capability over time. These analytics can be combined with other monitoring technologies to drill down and uncover anomalies before they occur.
All manufacturing and milling equipment vibrates. And this vibration per machine and job type can be plotted to determine a healthy range on a curve. Vibration analysis helps predictive maintenance engineers to learn what subtle and significant changes mean. They can assess wear rates and failure points as the machine learning algorithms become more accurate over time.
While predictive maintenance platforms can use sound, vibration, and temperature to assess both the health and possible failure of equipment or parts, another tool that captures what is going on inside the machine is oil analysis. By measuring oil purity, debris contents, contaminants, and oil composition, technicians can identify, plot, and predict the cause and develop strategies for correcting them. The data from oil analysis can be sent to the analytics platform and combined with other monitoring data for a clear picture of machine health.
Now that it’s clear that predictive maintenance is an assured way to avoid unplanned downtime and incur higher manufacturing costs, the question is: how do you implement a predictive maintenance plan?
First, get to the crux of the problem you’re trying to fix:
Then you need to assess your existing status or create a baseline of data on machine performance. For this, you can use your own standards, OEE standards, or other industry standards. Review each machine to see what the historical performance levels have been: how often it’s been down, what components fail regularly, how often maintenance is currently scheduled, and so on.
Second, examine the historical data for patterns and what metrics will indicate a problem, what deviations from the baseline should flag an operator, and so on.
Finally, once you are using these patterns and the data relative to your baseline performance measurements, you need to institute a process for continually updating the data and reviewing it to ensure that it continues to reflect the current status and will flag deteriorating patterns that clearly signal a need for maintenance. This is the key: you can’t predict what you can’t analyze. Accurate data is essential!
Minimizing unplanned downtime, at least as it relates to the functioning of the machines, is a huge cost saving and will prevent delays to market that will also impact the bottom line. In today’s manufacturing environment, predictive maintenance is not a “nice to have.” It’s a necessity.
47 Pleasant St, Suite 2-S, Northampton, MA 01060