Subscribe Find out first about new and important news
  • Share via email
  • Subscribe to blog alert

IoT and predictive maintenance

man is working on a robot Source: Bosch

I’ve often wondered why I need to change my automobile’s motor oil every 3,500 miles or 3 months, whichever comes first. Maybe I’m one of the few people left in the world who still enjoys changing his own oil. But it’s often concerned me that maybe a mileage- or time-based maintenance schedule wastes lots of valuable resources like oil; aluminium and metals for fabricating the new oil filter; plastic for making the new jug of oil; and energy for the recycling facility that takes my dirty motor oil, strips all the impurities from it and recycles it.

Maybe I don’t really need to change the oil that often. Maybe the metrics which determine when I should change the oil – in this case either vehicle mileage or time – aren’t the best predictors of my car’s optimal maintenance schedule. What if there were sensors on my red 1966 Ford Mustang convertible (V8, 289 cubic inch engine, 2-barrel carburettor, for those curious) that would tell me when I need to change the oil. And what if those sensors were collecting data from the oil filter, the valves, the pistons and the exhaust to find anomalies that are better predictors of when I should change my oil to prevent deterioration of my car.

Predictive maintenance is one such IoT/M2M solution that helps lower operating and capital costs by facilitating proactive servicing and repair of assets, while allowing the more efficient use of repair resources – both human labor and replacement products. See Fig. 1.

Traditional maintenance scheduling versus predictive maintenance for assets Source: Analysys Mason
Figure 1: Traditional maintenance scheduling versus predictive maintenance for assets. Source: Analysys Mason, 2013

IoT predictive maintenance is enabled by a three major solution enhancement over a traditional maintenance schedule

  1. Capturing Sensor Data – There are sensors on the asset being monitored. These assets provide a constant flow of data – most of which are all within standard ranges of tolerance and trigger no alert or error. However, a very small percentage of those data falls outside standard ranges and trigger alerts that might indicate a reason to undertake maintenance.
  2. Facilitating data communications – The data flowing off these sensors are connected to a central processing facility using some WAN- or LAN-based connectivity platform. The choice of connectivity technology is dependent on several factors including security requirements, level of asset mobility and expected integrations with other assets.
  3. Making predictions – The captured data is continuously evaluated – for example, in a business intelligence system – based on expert knowledge or data from past events.  The application relies on a series of rules that are based on expected and observed data associated with the assets being monitored. If the application can make better predictions, it can suggest a maintenance schedule which better matches the progressive deterioration of the asset. An enterprise need not over-invest in maintenance labor and parts. Predictive maintenance allows an enterprise to match its cash outlay with actual prevention of asset wear-and-tear.

Data aggregation and analysis is a key piece of a predictive maintenance solution. We wrote about data aggregation and analysis in “M2M management software: supplier and product review.” Predictive maintenance requires the collecting and systematic analyzing of large quantities of data to help predict faults and errors outside the normal tolerance range. These anomalies could be symptoms of larger problems with the asset and might trigger a maintenance response.

As a result, predictive maintenance can:

  • Lower maintenance costs by matching a maintenance activity with actual symptoms of future asset failures
  • Save vital resources by reducing the need to purchase replacement parts for asset components before they fail
  • Reduce out-of-service time for an asset by determining when maintenance must be completed in order to minimize asset failure

IoT predictive maintenance allows for more services

In addition, predictive maintenance allows enterprises to innovate through new revenue streams including enhanced warranty and maintenance services and also strengthen their competitive advantage through a differentiated offering. And last but not least, using predictive maintenance will help manufacturers to increase customer satisfaction from fewer warranty claims.

Share your thoughts about predictive maintenance for proactive servicing and repair. Is this a valuable solution? What other applicability can we find for predictive maintenance?

More information on IoT and predictive maintenance

From remote to predictive maintenance: How IoT refines a classic M2M concept.

Customer case study on predictive maintenance: It’s late in the evening, and the video system breaks down in a sensitive, security-critical building. The technician responsible for such issues clocked off hours ago. How can the security problem be detected and fixed as quickly as possible?


  • 19. June 2013 at 9:49

    @Steven, thanks for your interest – there are a few sources that I can share: Check out the More on manufacturing on our blog. Also, we provide free webinars for predictive maintenance, check out our website for next dates.

    Also, there is a rather new report by acatech about Industry 4.0 in Germany.

    If you need more help, please contact me directly.

  • 19. June 2013 at 8:07

    Nice article. I’ve been reading a lot on this topic recently and I would like to get some insights on the best practices of implementing Predictive Maintenance with M2M technologies in the manufacturing sector. Any help or leads to articles/books is very much appreciated.

  • 28. May 2013 at 18:40

    @Archie – Thanks for the comment. Why do you think the financial issues often derail deployment?

  • 28. May 2013 at 18:34

    Industrial engineers have heard the merits of predictive maintenance discussed time and again by colleagues, counterparts and solutions vendors. The advantages of a more intelligent, proactive operating model are no longer in question, but stubborn financial anxieties continue to delay investment.

  • 27. May 2013 at 13:38

    @ Barney — Thanks for the comment. No doubt doing things always by the textbook can be prohibitively expensive. What do you recommend in order to make the process a little less expensive? Are there particular areas of RCM where you recommend not going by the textbook? Certain things that are possible to exclude? Thanks

  • 25. May 2013 at 9:32

    Therefore, the goal of RCM is to determine the critically equipment in any process, and based on this information, designed a customized preventive/predictive maintenance strategy for the organization. RCM initiatives however involve a tremendous amount of resources, time, and energy. Thus the process is an extremely time consuming and expensive too especially when done according to the textbook.

  • 17. May 2013 at 16:18

    @ Edna — I agree. Are you involved in predictive maintenance or the automotive sector? There is so much valuable data created by sensors already in modern automotive vehicles. Capturing and analyzing that data has tremendous potential for environmental savings, product improvements and quality enhancements. What do you think?

  • 17. May 2013 at 4:40

    Oil sampling is a great example of predictive maintenance. Not only can you monitor the condition of your motor oil, but you’ll get a list of the amounts of metal and contaminants in the oil, basically giving you a health report on the engine.

  • 6. May 2013 at 8:30

    @Cameron — thanks for the comment. This is extremely interesting stuff. Do you have some examples of where CPAS deployments have taken place? What types of equipment, where in the world, who is benefiting from them? Please let the readers of the blog know, because you raise some very interesting examples. Thanks!

  • 5. May 2013 at 7:57

    To evaluate equipment condition, predictive maintenance utilizes nondestructive testing technologies such as infrared , acoustic (partial discharge and airborne ultrasonic), corona detection, vibration analysis , sound level measurements, oil analysis , and other specific online tests. New methods in this area is to utilize measurements on the actual equipment in combination with measurement of process performance, measured by other devices, to trigger maintenance conditions. This is primarily available in Collaborative Process Automation Systems (CPAS). Site measurements are often supported by wireless sensor networks to reduce the wiring cost.

  • 5. March 2013 at 22:34

    Thanks for the comment, Dattaraj. Fleet studies are an interesting topic. When you think about feeding information back into the design process, there can definitely be improvements in the planning process, cost savings, quality enhancements, services additions, etc.

    What are some of the impacts of feeding this information back into design from your perspective?

    Thanks again.

  • 5. March 2013 at 3:45

    Nice article showing the value of IoT in Prediction and helping planning of maintenance cycles. The data collected and analyzed is of great value and can be used to mine more Knowledge on Assets. For example if components at a site show a specific failure pattern, that could be attributed to certain factors in operation profile, environment, etc. Moreover this lets us do fleet studies pn Assets and feed the information back into Design. Lots of possibilities and new opportunities of this Technology.