Manufacturers, Do not Believe these Myths about Predictive Maintenance

Manufacturers, Do not Believe these Myths about Predictive Maintenance

  • By

    Umesh Chavan, Solution Architect

  • /
  • Posted

    25 Dec 2018

Manufacturers, Do not Believe these Myths about Predictive Maintenance

Until very recently, the manufacturing industry has been aware of only two kinds of maintenance practices. Preventive and Reactive. Preventive means taking care of the machine on a regular basis so as it is regularly maintained and its longevity increases and reactive, as the name implies, happens when there is some (often major) breakdown. With the advent of the Internet of Things (IoT) and Artificial Intelligence / Machine Learning (AI/ML), another kid on the block is coming into the picture known as predictive maintenance. As is the case with anything new, predictive maintenance is also being surrounded by a lot of hype and at the same time looked at with a lot of skepticism. Through this blog, will attempt to assist folks in taking a rational view about the various myths surrounding predictive maintenance.

  • One brush to paint the complete landscape – People tend to believe that when it comes to predictive maintenance, there is one single underlying AI model that can be applied in all the similar type of equipment running in a setup. However, it does not work that way. For example, even two pumps of similar specifications which may be fulfilling the same requirements become very different once off the production line. Each and every asset and machinery is subject to its own set of quirks and conditions. In such a scenario, a 30,000 feet high-level recommendation simply does not fly. The behaviors of the machines get greatly altered during its life and resultantly show different patterns. While the underlying machine learning algorithm may be same for similar types of machines, the training data will be different for each machine type.
  • You can reap the benefits of predictive maintenance from Day 1- It is a common misconception that as soon as one has put together a system for predictive maintenance one can see the benefits. Predictive maintenance is very different from the "plug-and-play” software. The system needs to mature by collecting and training on huge amounts of data. Its penchant for anomaly prediction improves as the amount of data it sits upon increases. Understanding the whole process, identifying the correct equipment to consider, and getting the right people involved takes a lot of time and analysis. As we often recommend, start small while thinking big. To zero down on the starting point, asses each process and equipment on the basis of prioritized use case, operational readiness, and business criticality.
  • You don't need expertise - Predictive maintenance is a heady concoction of IT skills, domain knowledge, and data science expertise. It needs a lot of readings and drawing inferences from the data. So, it is pretty much evident that the already existing workforce has got to be reinforced with people having these skills. While subject matter expertise and deep understanding of the machines is important knowledge to have, to benefit from predictive maintenance, you need data science and data analytics experts in the team.
  • Preventive is better than predictive - This is the most common argument that one gets to hear. Each maintenance methodology has its pluses and minuses and a good maintenance program should be combination of preventive, prescriptive, and predictive methodologies. With IIoT systems, one can define preventive maintenance based on Running Hours of asset rather than fixed period and this brings about huge cost savings.
  • Equipment is not important and is under warranty - As was stated in Zen and the art of motorcycle maintenance, even a small nut or screw can cause serious disruption in the way the whole bike works. Similarly, if an equipment is there in your plant, it is important and a vital piece of the whole grand design. Even if it is under warranty, employing predictive maintenance has its own benefits. First and foremost, you start feeding the data to your model which captures even minor data changes – these data changes can point out issues that are sure to rise. This can assist the organizations in dealing with those issues during the warranty period which otherwise would have gone unnoticed for long.

The primary consumers for the predictive maintenance system will be the floor technicians and the system has to be easy to use and appealing to them.

Predictive maintenance will gain ground as data quantity and quality improves. However, it remains to be seen how fast organizations jump on the bandwagon. But providers of IIoT solutions are definitely preparing for wide adoption of predictive maintenance and used properly, it will give huge benefits to industries.

Be a System Integrator with AIPL....

Try PlantConnect!

Thank You!

We will get back to you soon!