predictive maintenance

Is predictive maintenance possible for older devices without sensors?

16 November 2017

Internet of Things

At the end of October, around 200 customers, professors, students, IT experts and other guests gathered in Ghent for our event, ‘Co-Thinking about the Future’, to consider the technology of the future. You will find a number of ideas and topics discussed during the event in this series of blogs.

02_Foto blog maintenance.jpg

Industry 4.0 isn't outside waiting the door; we've already invited it in and offered it coffee. It enables us to use sensors, the Internet of Things (IoT) and artificial intelligence for predictive maintenance: forecasting when machines need servicing. And this has lots of benefits, but what do you do if your machines are a bit older?

Predictive maintenance and preventing downtime

The Internet of Things allows machines to communicate with each other, and gives us the ability to collect huge amounts of data to run a whole battery of analyses on. The information this provides means we can know exactly how our devices are functioning, or when and why they break. We can discover patterns in component life cycles, and plan servicing accordingly.

Does the data show that part A normally gives up the ghost after a year, or that pipe B is as good as blocked after six months? Then you can take preventive measures to replace part A after eleven months instead of waiting until your machine fails. Predicting defects means you reduce the risk of broken machines and can keep the chance of downtime for your production to a minimum. You streamline your routine servicing because predictive maintenance gives you a better overview of what needs checking or repairing.

Make older devices smart with add-on sensors

You'd like to get started with predictive maintenance, but the devices in your factory are mostly older models without many sensors. So what now? We can make machines like this 'smart' nowadays by adding external sensors. You can place temperature sensors on water pipes, for example. So a naturally analogue object like a tube can suddenly be made smart: it tells you when the water is freezing and warns you if you need to check for leaks. Unfortunately, this doesn't mean that you can place any old sensor anywhere without first thinking about it. If you do, you'll quickly notice that your data is inadequate or poor quality.

Think about your sensors and data

One of the main pitfalls with predictive maintenance for older machines is collecting the wrong, poor quality or insufficient information. You need to add sensors to an old device yourself, so you also need to consider the kind of data you want to collect via the device. The way machines operate is after all affected by many factors. Keeping a record of when a device no longer works is the minimum you can do for predictive maintenance. At least then you can already estimate a number of expected interventions.

It's better to gather data from various sensors to be able to find out why the device isn't working. This information makes your predictions much more accurate. But because you're installing the sensors yourself, you need to think carefully about exactly what you need. Drawing up a clear and well-considered data strategy, not just for collecting but also for processing, is therefore essential. If you notice later that you need info about the humidity in the production hall but haven't installed a suitable sensor, for instance, you can only try to derive this information from correlations.

Secure external sensors

Your production lines used to be securely protected in a factory. Anyone who wanted to sabotage your production had to be able to sneak in. But the arrival of the Internet of Things means your factory is now in constant digital connection with the outside world, which makes securing Internet of Things sensors and machines a big challenge. It's quite simple for new and large devices: you can ensure watertight protection during the design stage.

Security for add-on sensors in use on older machines is a very different story, however. These small devices don't always have the computing power necessary to provide impenetrable protection. Hackers once worked out how to hack electronic price tags in a supermarket, for example, and changed all the prices to 99 cents. But even these sensors can be comprehensively protected with a detailed policy based on IP security. You can also introduce operational security: program commands that prevent sensors from performing certain actions, e.g. code to ensure that a sensor linked to a blast furnace cannot switch the heat protection off.