Sensor Technology
Predictive maintenance precedes success, here's how it can help you improve
By Maria Alejandra Salazar, product sales manager Analog & Sensors, Rutronik
O
nce upon a time machines were maintained on a schedule and some companies still use that method. But nowadays, thanks to machine learning, wireless sensors and cloud AI, machinery maintenance is far more flexible, adaptive and smart. The result? Predictive maintenance that avoids unexpected damage, reduces downtime and saves money. So how does this work and is it something you can take advantage of right now?
Predictive maintenance sensors One way to sense what is going on in a machine is to use audio data, not only in the traditional audible range, but also from ultrasonic sensors; the higher the frequency, the earlier damage can be detected. That means avoiding wider damage and allowing for time in parts ordering and maintenance planning, saving on costs. Which sensor is needed depends on the machine. For example, gas pipes require ultrasonic sensors; if a leak is detected through audible noise it is already too late for predictive maintenance. By contrast, an air guidance system damage is usually due to an imbalance which can be detected in a range of about 2 kHz. In the case of very slow moving parts a new type of sensor is required, like movement.
Converting the data for use Once the information like noise or movement has been detected, it needs to
be used. The normal, healthy running machine needs to be recorded in order to establish a base range so that the sensors can look out for deviations.
Once the anomaly is detected it can be used in two ways: edge or cloud computing. The edge method, which is local, uses a closed system which only requires an internet connection in the case of updates. The advantage here is a response to any changes within milliseconds. By contrast, cloud computing deploys AI and swarm intelligence where the system learns from lots of machines all over the world at the same time. That means that detection prediction is far better as all variations can be considered. The swarm also offers a limitless amount of power with its huge memory capacity and can constantly update the algorithm used, so there is always the optimal analysis in place. This can use machine learning, or even deep learning, to improve and ultimately offer the best possible protection for the machine. Of course, this saving needs to be weighed against the ongoing costs of using longer range sensing over local and multiple sensors where applicable. Once a decision has been made the sensors have to be placed.
Sensor position importance To find the ideal installation position certain factors must be considered, including ambient noise levels, space availability and
interference. Once the ideal position has been established the sensor type can be chosen, usually either audio or movement based. For audio the MEMS microphone sensors with a specific frequency range are ideal, like the STMicroelectronics and Infineon sensors. For movement detection the use of shock, vibration or acceleration sensors can be used.
then be easily calibrated using a laptop and USB port. If the sensor nodes should only perform basic connectivity features the Bluetooth 5.1 qualified nRF52811 SoC can be used.
Using the data
The data is then converted into digital information, which can be evaluated
By combining these and even more sensors (e.g. temperature, humidity or pressure) a clearer picture can be painted for greater accuracy. This will, of course, increase the costs, but might be worthwhile in applications such as difficult- to-access offshore wind farms – to avoid unnecessary engineer calls - or machinery that can cause greater damage, like belt failure or faulty production.
Data transfer is smarter than ever Thanks to huge advances in wireless technology and its infrastructure, it's now easier and more affordable to monitor machinery from afar and via the cloud. Machine sensors can transfer their raw analogue data to a microcontroller. This data can be shared through cables but it's generally cheaper and easier to do it wirelessly. The Nordic Semiconductor nRF52840 System-on- Chip offers various mesh protocols including Bluetooth mesh, ZigBee and Thread. NFC can also be used to offer a simple way to pair up the sensors with the data collector. All these sensors can
28 November 2019 Components in Electronics
locally. This can also be connected to a larger system using LTE, which is fast enough to allow for newer sensors to connect directly to the cloud without the need for a hub. However, for a time-critical analysis, 5G is ideal as this has low latency that will offer the feedback within a few milliseconds that's required. In the case of LTE the latest NB1 and M1 – aka Narrow Band IoT and LTE-M - are ideal for predictive maintenance. Both are supported by the Nordic Semiconductor nRF9160 System-in-Package (SiP) featuring an ARM Cortex M33 microcontroller, sensors and actuators. This allows for much smaller amounts of data to be sent as information is generated on site thanks to its powerful computing speeds. That means energy use is optimised and data consumption is kept low. You can even connect LEDs using the 32 GPIOs to offer a visual on-site warning system. Buttons and switch relayed can also be added in, allowing for entire system shut off if a problem is detected.
rutronik.com
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56
