The aim of predictive maintenance is to predict equipment damage before it occurs and, therefore, reduce downtimes to practically zero. Maintenance work can then be carried out exactly when necessary.
Whether predictive maintenance delivers what it promises, depends on the technology used.
The first step is to determine where the damage, which is to be prevented, can occur. “However, this is not automatically the ideal position for the appropriate sensor,” explains Stephan Menze, Product Manager MEMS and Optical Sensors at Rutronik. “The reason for this is because this position is not always accessible or offers enough space, sometimes there is also too much distracting noise in the vicinity.”
In most cases, the placement usually determines the sensor type: “If the sensor is mounted directly on the device or equipment, it must detect structure-borne sound. In this case, a shock, vibration or acceleration sensor should be chosen,” says Stephan Menze. “If the sensor is mounted outside the device or equipment, it must detect the airborne sound. In this case, a MEMS microphone sensor can be considered – unless the environment is too humid or dusty.” This is because microphone sensors always have an opening to pick up the sound waves and to reduce the sound pressure level. Dust or moisture can also penetrate through this. “Therefore, the sensor must be protected or alternatively, shock and vibration or acceleration sensors can also be used here.”
Microphone sensors are available in different frequency ranges. Stephan Menze: “The higher their frequency, the sooner they can register approaching damage. In the audible range it is often already too late for predictive maintenance!” From 16 kHz, i.e. in the ultrasonic range, sensors can already detect irregularities months before the damage occurs; in the audible range below 16 kHz, it is usually only weeks. If this is sufficient, for instance because the damage in question can be rectified quickly, the ideal frequency range of the sensor depends on the equipment type or the particular part of the equipment. “The faster the movement here the higher the frequency range of the sensor,” according to Stephan Menze’s rule of thumb. “For parts that move very slowly, an acceleration sensor mostly provides more reliable measurement values than a microphone sensor. Even better predictions can be achieved with combinations of different sensor types, especially if they are networked with one another. However, this only makes sense if the damage to be prevented or the maintenance effort is significant enough.”
Via the data collector to the internet …
Depending on the type of application, individual sensors must now first transmit their measurement data to a local data collector. For this connection between sensors and data collector, radio technology is generally cheaper, more flexible and more durable than cabling. “Microcontrollers with integrated radio interfaces and AD converters are ideal for this,” recommends Bernd Hantsche, Marketing Director Embedded and Wireless at Rutronik.
They are often already available with radio stacks that are tailored to the microcontroller, so that the customer only needs a few lines of code to digitize the analog values and transmit them to the data collector, where they are evaluated. “In this case, LTE offers sufficient speed to connect to the internet. However, if a reaction time of a few milliseconds is required, LTE is not sufficient. You should then choose 5G.”
Users are very flexible with the nRF52840 from Nordic Semiconductor: It supports Bluetooth 5, Bluetooth mesh, ZigBee and Gazell, an open-source stack for star topologies. Sensors can be connected to the data collector via its NFC tag. “However, the nRF52810 would be more favorable and perfectly suitable if only Bluetooth 5 or Bluetooth mesh are established as wireless technology from the outset. Depending on the environment, Bluetooth 5 has a range of more than one kilometer in long-range mode and can thus, under certain circumstances, replace sub-GHz technology,” continues Bernd Hantsche.