Overview of Preventative Maintenance

Maintenance strategies based on “prediction of failure” have always been the goal of reliability teams tasked with reducing unplanned downtime and overall maintenance costs. In fact, nearly every industry has assets where unplanned downtime is either catastrophic or extremely expensive.

For these scenarios, the traditional response of reliability teams has been to build mean-time-to-failure models whereby past failures of similar assets in similar production modes are compiled into a statistical model. Based on the risk of failure of critical components and the severity of each failure a preventive maintenance (PM) schedule is put in place to direct maintenance teams to replace or refurbish those components at regular intervals. If you’re reading this blog, you likely know the shortcomings of preventive maintenance schedules.

Shortcoming of Preventative Maintenance

• Lack of Failure Mode Data: For failure mode data to be useful, the data points must be numerous and comparable. Many high-value assets rarely fail which means very little failure mode data is available which decreases the accuracy of the model. Failure mode data points must also be comparable (same asset type, same production mode, same environmental conditions) and this is also unlikely.

• Needless Maintenance Cycles: When a PM schedule is put into practice, it is well known that components are being replaced far in advance of their useful life cycle. This needlessly increases maintenance costs and burdens maintenance teams who must perform the work.

• Staff Shortages: A recent study by the United States census bureau found that nearly one-fourth of the manufacturing workforce is age 55 or older and a global shortage of skilled labor means the days of reliability engineers single handedly keeping a site running are coming to an end.

• Unexpected Failures: Even when a PM schedule has been developed and is being followed, it is well known that unexpected failures still occur.

• Asset Failure Models Rarely Work: A comprehensive study by NASA showed that 82% of assets display a random failure pattern. This means that only 18% of assets display a failure pattern that is amenable to models based on hours or cycles of operation.

Conclusion

As reliability teams began to understand the obvious shortcomings of preventive maintenance, the advancements in IoT and the ability to install industrial sensors on every asset offered a new data-driven approach. The new data-driven approach finally solved the problem that everyone knew for decades: that each asset is different. With the introduction of real-time data, reliability teams could at last begin tailoring their maintenance strategy for each individual asset.

A few years later, with the data renaissance no longer in its infancy and databases starting to fill, reliability teams were surprised that they were no closer to the promised 100% uptime. We’ll discuss the reasons for this and more in the next blog, Unrealized Potential: Why pervasive sensing has been so disappointing.