Trusted Sensor Technology For The Internet Of Things

“Data is the new oil” — Clive Humby, 2006

While this prediction relates to the value that can be generated from data, the focus here is on the tools at the oil well. Just as oil drilling platforms are expected to reliably produce crude oil around the clock, sensors are expected to reliably and continuously deliver high-quality data.

But sensors have long since evolved from simple measuring transducers to become computing platforms that communicate – sometimes wirelessly. This expands their range of potential applications and benefits enormously, to the point of integrating artificial intelligence in order to fuse and analyze data directly at the measuring point.

Conventional, discrete sensor designs have for many years been losing ground to microelectromechanical systems, which are more energy-efficient, more compact and less expensive. This means sensor systems can operate autonomously over a long period of time and can be integrated into a wide variety of environments. Well-known examples can be found in numerous consumer products such as smartwatches. MEMS-based intelligent sensors can also be integrated into the moving components of machine tools to monitor the production process.[1] Their application even extends to earthquake detection: the use of MEMS reduces the cost of a monitoring station by two orders of magnitude, enabling dense sensor networks for early detection of seismic waves.[2]

Intelligent ball screw project as an example for integrated intelligent sensors.

Many applications place high demands on the reliability of the sensor technology, especially in closed control loops, e.g. for automated flying, driving or robotics, where it must be ensured that the sensor signals are always trustworthy. Integrating artificial intelligence into the sensor enables real-time condition diagnosis and can prevent the failure of transducers or electronics from causing severe damage.[3]

The increasing integration of functions into sensor systems also brings with it the need to expand development and testing methods. Starting with the fundamental question of whether the measured variable has been correctly detected,[4] it is crucial that the algorithms used by intelligent sensor systems are correctly trained. They must also deliver reliable and error-free analysis during operation – even under changing conditions, such as temperature fluctuations. Given the complexity of the sensor systems and the wide range of relevant test cases, testing in the application environment is not suitable: the effort required to reach reliable conclusions as to suitability for use would be far too high and would make development uneconomical.

This calls for the further development of hardware-in-the-loop processes. The testing of the sensor element’s function must factor in stresses, such as those caused by temperature changes, as well as real-time simulation of operating scenarios and possible errors. This can accelerate the training and validation of intelligent sensors. Finally, these methods also make it possible to simulate a wide variety of operating environments with comparatively little effort, and thus to test the feasibility of new applications for intelligent sensors, right through to evaluating energy consumption in order to estimate the service life of wireless, stand-alone systems.

1. https://www.leistungszentrum-mikronano.de/en/projekte/sdsema.html
2. Cascone, Valeria, Jacopo Boaga and Giorgio Cassiani. “Small Local Earthquake Detection Using Low-Cost MEMS Accelerometers: Examples in Northern and Central Italy.” The Seismic Record 1, no. 1 (April 1, 2021): 20–26. https://doi.org/10.1785/0320210007.
3. https://penta-eureka.eu/project-overview/penta-call-5/trust-e/
4. Iqbal, Ali, Naeem S. Mian, Andrew Longstaff and Simon Fletcher. “Performance Evaluation of Low-Cost Vibration Sensors in Industrial IoT Applications.” International Journal of Automation Technology 16, no. 3 (May 5, 2022) : 329–39. https://doi.org/10.20965/ijat.2022.p0329.