Intelligent Acoustic Sensors

AI-based analysis methods support process monitoring and the early detection of significant events. Sensor technology that mimics human sensory perceptions is also used here. For example, in order to make technical use of the high information content of the acoustic sensory channel, it is necessary that the sensor technology detects and evaluates sound fields in a (similar) way as humans do.

Experienced personnel know which noises to pay attention to when evaluating machines, for example, subjectively blocking out interfering noises. Therefore, the sound field is often listened to from different positions. By combining what is heard with empirical values, the person can very quickly grasp the prevailing situation. In practice, however, due to a variety of factors, this approach can sometimes only be implemented with a high level of personnel effort, costs and, if necessary, elaborate safety measures for the employees.

Automated event monitoring by means of acoustic sensor technology must prove itself in terms of performance and costs compared to previous methods. The performance is directly dependent on the quantity and quality of the sensor-based measurement data. The sensor technology must therefore deliver the best data quality with manageable installation effort and low operating costs. Furthermore, it should carry out the data acquisition largely autonomously in order to save personnel resources.

The potential to integrate sensor technology into mobile platforms (e.g. vehicles, robot systems, or drones) requires not only the automation of the measurement process but also small hardware dimensions and possible battery operation. This means that measurement positions that are difficult to access can be taken dynamically or events that extend in time and space can be analyzed (remote maintenance). Likewise, the possibility of measurements independent of the time of day and personnel is particularly suitable for long-term observations.

An acoustic sensor system is characterized by:

• Automation of measurement data acquisition
• Self-calibrating acquisition in a short time
• Intelligent signal processing for analyzing the recorded sound field
• Real-time analysis of the data quality and feedback into the measurement process
• Selection of measurement points and adaptation of sensor characteristics based on previous analysis results
• Preparation of data for further processing via universal interfaces/platforms (usability)

The technologies developed at Fraunhofer IDMT in the field of intelligent acoustic sensor technology enable very powerful sound field detection and event analysis.

Depending on the application, sensor concepts are developed that consist of special sensors, dedicated electronics as well as measurement, pre-processing, and analysis methods. The developments are carried out under the general conditions of an efficient, low-effort, and cost-effective realization of automated event monitoring. Characterizing features  for this are, among others:

• Use of low-cost MEMS sensors
• energy- and data-saving operation (data recording/transmission/processing only for relevant events)
• improved measurement routine through automatic conclusions from previous analysis results
• Automotive sensor platform (vehicles, robotic systems or drones)
• Use of universal interfaces between sensor and data processing unit