Every “thing” needs a sensor. However, in the industrial IoT environment, it shall be able to do more than just spit out physical measurement parameters.
Sensors have played key roles in automated vehicles and the Internet of Things for some time. According to a recent Roland Berger study, between 2015 and 2020, sales will double to 30 billion units. But the artificial “sensory organs” are also very much involved in the implementation of Industry 4.0 concepts. Not only as “measuring devices”. With their own microprocessors, microcontrollers, or DSPs (Digital Signal Processors), they also carry out complex signal processing and thus become important data sources. They only earn the title “intelligent” or “smart” when they carry out these additional tasks, which in this case, for a change, has nothing to do with artificial intelligence. Not yet.
It is also important that they can be integrated easily into complex production systems via digital interfaces (e.g. CAN, Modbus, CANopen, IO-Link, Ethernet). The current industrial sensors are suitable for this only to a limited extent. In many cases not smart or flexible enough, they use too much energy and are too expensive.
Intelligent sensors as MEMS
One solution is MEMS (microelectromechanical system). For example, the research project AMELI 4.0 (PDF) led by Bosch is also attempting to make the key technology fit for industrial sensor applications. It has long been a standard in consumer electronics (e.g. accelerometers in smartphones) and in cars (e.g. electronic stability control, ESC). However, in the industrial environment it still lacks the necessary robustness and performance.
Other solutions also need to be found for energy supply. Industrial MEMS sensors operate without cables or batteries. They harvest energy, for example from machinery vibration or waste heat and convert this into electric current.
At present, the main task of these systems is condition monitoring (of machinery). This should make unplanned downtime a thing of the past and repairs and maintenance can be carried out when the time is suitable.
Shifting “intelligence” from automation systems into field devices also makes automation systems more efficient and more powerful, because intelligent sensors and actuators communicate with each other directly.
Users want plug and play
In the past, condition monitoring has often foundered because of the high engineering expense. Intelligent sensors could be the solution. For example, they would pass on their results to a cloud service platform where they could be compared with a database in order to initiate actions, such as alarms or messages, as required.
However, a survey carried out by Namur (international user association of automation technology in process industries) showed that even smart sensors cannot fulfill all the wishes of users. For instance, almost 70% of those surveyed want automatic integration into the system architecture, followed by automatic calibration. Incidentally, half the participants expect that within the next five years only a quarter of the sensors could already be intelligent sensors. This should rise to 50% in the coming ten years.
The recently published study from Roland Berger about smart sensors more or less provided an answer with three ideal-typical strategies for sensor producers:
Strategy 1 “Analog Now”: This is the strategy for companies that see themselves as future metrology experts. Because of the cost pressure, the aim should be to achieve higher production volumes and standardized sensors.
Strategy 2 “Smart Plug & Play”: According to this, sensor producers position themselves as technology leaders for local analysis capabilities. For this, they should have systems that combine measurement data with smart software algorithms. The systems they offer pre-process data and, in the best case scenario, even have self-learning capabilities, for example via Deep Learning.
Strategy 3 “Sensor Fusion”: As digital innovators, sensor producers offer complete solutions. They should be able to gather large volumes of data from many different measurement data sources and process this in networked systems through to the cloud. Standardized platforms could help reduce costs and make products more suitable for the mass.