Sensor technology in production: data acquisition and processing

In the course of advancing digitalization, the requirements for automated and networked factories can be met, which are only made possible by intelligent sensors.

Sensors are the sensory organs in industrial production. The data they record and evaluate is used to control machines and systems. This makes them the basis for successful automation. The ZD.B topic platform Digital Production & Engineering has set itself the goal of enabling SMEs in particular to use digital technologies. In addition to networking and knowledge transfer between digitization providers, users and science, this also includes providing information on topics, trends and technologies. This is where the webinar series "A Bridge between Science and Industry - From Research into Practice", which will address the topic of sensor technology in production in May 2022 - conducted in cooperation with the Strategic Partnership for Sensor Technology e.V.

Generating Added Value from Data

A wide range of types of sensor technology today makes it possible to capture a large number of data points in production. For many companies, this poses a major challenge: to select the relevant data points for their own use case, as well as the sensor technology required to capture them, and to process the resulting data in such a way that information can be derived from it and added value generated.

Cooperative sensor technology in intralogistics

In the first presentation, Prof. Dr. Hans-Georg Stark, Head of the Center for Scientific Services and Transfer (ZeWiS), Aschaffenburg University of Applied Sciences, explains cooperative sensor technology using the example of a project from intralogistics. Cooperative sensor technology is characterized by the fact that it works together with target objects. It is used wherever maximum safety and reliability are required, such as in the use of industrial trucks in heterogeneous warehouse systems in which autonomous and non-autonomous industrial trucks as well as logistics personnel are represented. Using a schematic diagram, vehicles with different levels of automation are shown communicating wirelessly with each other. This enables them to recognize each other's position. The data collected by the sensors is transmitted to a process computer, which takes over control of the vehicles. In addition to localization and control, other possibilities open up. For example, the wear conditions of vehicles can also be measured, such as the condition of a lift mast chain. Acoustic detection with corresponding frequency analysis can be used to derive the maintenance status and plan maintenance in advance. In simplified terms, this can be expressed as follows: a damaged part emits different frequencies than a qualitatively perfect one.

The Fast Fourier Transformation is used to determine this. It breaks down a signal into individual spectral components and thus provides information about its composition. This method is used for error analysis, in quality control and in the condition monitoring of machines or systems.

With increasing networking of industrial production, further data can be recorded. For example, distances can be measured or the loading states of vehicles can be determined. With modeling via mathematical graphs, all intralogistics processes are simulated. In this case, a virtual forklift fleet can be constructed using the open source software tool CARLA. Critical distances can be checked and thus avoided, and position and loading states can be determined with suitable sensors. In the digital model of a real intralogistics center, numerous logistics processes can be optimized through simulation. One example is the algorithmic determination of the number and position of charging stations for e-trucks, which can be used to improve energy balances and security of supply for the forklift fleet.

Using acoustics to ensure the quality of workpieces

The second presentation by Klaus Lutter, M. Eng., Coburg University of Applied Sciences, also establishes a link between theory and practice. He deals with the symbiosis of acoustics and machine learning for increased quality assurance. Relevant factors for quality assurance are found to be time, subjectivity of the test method and cost. True to the motto "More complex is always possible", a simple measurement principle is chosen. It is based on the assumption that test specimens or workpieces emit characteristic acoustic signals after impulse excitation. And that flaws, cavities or cracks or similar alter these signals. These signals are recorded via a microphone. A frequency filter is used to filter out the often low-frequency interference frequencies that are typical in industrial production. Via a principal component analysis, it is possible in the next step to reduce the recorded sensor data to the most important ones. This simplifies the complexity of the calculation and the data can be scaled accordingly.

Recognizing patterns in data sets

But how do you get to the actual machine learning now? Here, it is about recognizing patterns in data sets and deriving possible anomalies from them. Essential here is the quantity and quality of the data, which can be classified using various mathematical calculation forms. In the method used - Support Vector Machines - the classifier detects errors in test items with a high probability. Advantages of the method include the very short test duration and the high degree of flexibility, because different algorithms can be implemented and tested. And the whole thing can be automated if the measurement method is integrated into production. As a conclusion, Mr. Lutter states that a relatively simple measurement method can classify components and thus accelerate and improve the quality assurance process.

With sensor data through the digital transformation

The third speaker will be Dr.-Ing. Ulrich Lettau, CEO of iba AG. His presentation "Using sensor data from different sources consistently and comprehensively - the basis for successful digitalization" focuses on applications in the process industry, where sensors play a very important role. Whether filling levels, speeds or temperatures - actions can be derived from the sensor data determined, which are used to automatically guide the process via actuators. This is done via specialized computing units, the so-called programmable logic controllers (PLCs). In more complex processes, there are usually several specialized PLCs that exchange data with each other via bus communication. In addition, there is usually an operating and monitoring system that visualizes process variables and allows the operating personnel to intervene if necessary. The data available in the programmable logic controllers can be recorded in a highly cyclical manner together with data from external sensors in a measurement and analysis system and stored permanently. In addition, the sensor data acquired in this way can be further processed into characteristic values both online and after defined process steps have been completed. This makes the acquired sensor data particularly valuable not only for process automation, but also for many other applications. These include fault and malfunction detection, process analysis, quality documentation, but also condition monitoring, power quality, energy management and vibration analysis. Ultimately, these are all topics that contribute to the digital transformation. The characteristic values determined are used, among other things, to monitor limit values and can be transmitted to higher-level or cloud-based IT systems via connectivity solutions. For example, they can be used for AI applications where a large amount of historical data is needed.

On the part of Theme Platform Digital Production & Engineering , we will be happy to support you in identifying suitable scientific cooperation partnerships and in selecting appropriate funding opportunities for your digitization project.

Speaker contact details:

• Prof. Dr. Hans-Georg Stark, Head of Center for Scientific Services and Transfer (ZeWiS), Aschaffenburg University of Applied Sciences; e-mail: Contact by Mail
• Klaus Lutter, M. Eng., Faculty of Applied Sciences, Coburg University of Applied Sciences; e-mail: Contact by Mail
• Dr.-Ing. Ulrich Lettau, CEO, iba AG; e-mail: Contact by Mail