Proper maintenance on plant equipment is vital to keeping the plant running smoothly, meeting customer commitments, and controlling costs. Any equipment breakdown or unscheduled maintenance downtime is an expensive disruption that no plant can afford. So the focus of maintenance activity and planning is always aimed toward prevention, as much as repair.
Preventive maintenance is common practice in plants and factories around the globe because it just makes sense; and it works. But simply replacing parts and tools at or near the end of their expected useful life ignores the fact that even “identical” parts and tools are not really identical. Each individual item will perform and behave in its own way. Operating conditions and the specifics of how it is used will also affect its actual useful life. Nevertheless, it still makes a lot of sense to repair and replace before failure shuts down production.
Is Knowing the Useful Life Estimate Enough?
Typical preventive maintenance would have you replace a specific part or tool after, say, 10,000 cycles, if that is its expected useful life. If the useful life estimate was the average life for that particular part, however, half the time the part will still have some useful life left when it reached that point whereas the other half of the time, the part will fail before it reaches that milestone. True, that’s a fifty percent reduction in failures, but that’s not really good enough.
To be safe, the replacement point would have to be set at less than the average – let’s just arbitrarily pick 8,000 cycles – meaning that even more useable part or tool life will be replaced and discarded in order to prevent failures on the line. In reality, this replacement recommendation would be based on statistics and engineering studies, but the principle applies. Replacement according to a schedule prevents breakdowns at a cost (wasted part or tool life), albeit a cost that is acceptable compared to the cost of a breakdown.
IIoT, Simulation and Machine Learning
There’s good news. The Industrial Internet of Things (IIoT), simulation and machine learning now offer a truly effective predictive maintenance approach that is far superior to simple preventive maintenance. Instead of using average or standard useful life estimates, predictive maintenance measures actual conditions and performance, largely from IIoT sensors and machine controls as well as other data input from the plant floor, builds a model of part behavior, quality and failures, then models (simulates) future behavior to get a statistically-based estimate of when the specific part or tool will need adjustment or would fail.
Avoid Failures and Non-Performance
Predictive maintenance relies on measurement of various factors of the part itself and the environment including such things as temperature, humidity, vibration, pressure (stress), operating speed, material being processed, and more. It builds profiles of each part type part and its collective history of wear and stress under varying conditions, then uses that profile to project the likely timing of end-of-life or the need for adjustment or repair. Each individual part has its own profile, establishing its position on the lifeline model for its type, and projecting its continued wear until it reaches the point where it needs attention.
Predictive maintenance is much more effective in planning maintenance activity to avoid non-performance (repair or adjustment needed) and outright failure compared to preventive maintenance, based on averages. Companies using predictive maintenance repair or replace parts and tools no more often, perhaps a bit less often in some cases, while enjoying much lower rates of breakdown and failure.
It should be evident that the data that feeds predictive maintenance comes from operations management arena, specifically Manufacturing Execution Systems (MES) and operations management systems (MOM). In turn, the results of predictive maintenance – including the impact of maintenance activities on the plant and equipment schedules – is important input for MES and MOM. The integration of operations and maintenance from both a systems and data standpoint, as well as operational processes and procedures, enables both sides of the organization to work together to maximize output and efficiency.