It is no secret that unplanned downtime is the bane of existence for many manufacturers, and that simply relying on conventional run-to-failure maintenance schemes is the costliest method for plant maintenance overall. Many manufacturers have implemented preventative maintenance programs to reduce the occurrences of unplanned downtime, but there are shortcomings to this approach as well. This has led more manufacturers to attempt to implement predictive maintenance programs or even prescriptive maintenance programs. However, there are cost implications that will need to be factored into determining what makes the most sense for your facility.

What's in a name?

It's important to understand exactly what each maintenance program warrants to determine which plan makes the most sense for your facility.

Preventative maintenance: this type of program involves creating an equipment maintenance plan that is based on knowing the historical performance of the equipment to determine when (and what type of) maintenance is required. This is typically provided by the manufacturer of the equipment and is based on mean time between failure (MTBF) and mean time to repair (MTTR) data for major components. This (theoretically) allows for planned maintenance during a period that is least disruptive to operations. It is also a concept well understood by most manufacturers.

This is a great system if your equipment performance is average. However, there are many factors that could cause a premature failure resulting in unplanned downtime. Conversely, this system might result in premature maintenance, which means you could be performing maintenance tasks too often. Ideally, you would like to know the actual device health at any given moment to know precisely the best time to perform a maintenance task. This is where the concept of predictive maintenance comes in.

Predictive maintenance: this type of program involves the use of sensors to monitor actual device health in real time and make maintenance decisions based on the actual performance data rather than "average" data. The average data is still useful as a baseline, and then actual data is compared to the historical in order to make the best predictions on when maintenance should be performed. To be most effective you need to have sensors everywhere, and then be able to analyze the huge amount of data – in real time - that will be generated to glean appropriate insights.

Most modern machinery contains sensors that can be used for this analysis, and "add-on" sensors are readily available and relatively inexpensive. The greater expense is in the data storage and analytics software required. This sometimes results in a system that periodically collects data that is then fed into a system for analyzing. Depending on the measurement intervals, this could be just fine for many applications but is not as accurate as continuous monitoring.

Prescriptive maintenance: this type of program incorporates all the features of a predictive maintenance program, but then raises the bar by incorporating machine learning and artificial intelligence that can prescribe mitigation solutions for optimum results. The “prescription” will not necessarily be limited to describing what is or how to perform the maintenance task but will also offer suggestions on how to better run the equipment to extend the time between maintenance cycles and the overall life of the machine. As machine learning and AI develop, you will see much more of this type of program being implemented.

Which maintenance program is best?

While it seems that a prescriptive maintenance program will yield the best results, there are cost and complexity considerations that might be a barrier to implementation. However, with the prevalence of inexpensive sensors to sense just about anything, and the IoT and cloud computing making data analytics more affordable, it is easier than ever to improve the effectiveness of your maintenance program. The bottom line is this: the more data you can collect, the better decisions you can make. But keep in mind you need to be able to make use of all the data you collect.

Many organizations will have a mix of programs: predictive or prescriptive maintenance programs for the most critical assets and preventative maintenance programs for less critical assets. There is no “one-size-fits-all” solution. Choose a solution that’s best for your organization.