Entwicklung von neuartigen magnetosensorischen Techniken und Überwachungsverfahren(Development of novel magnetosensor techniques and monitoring methods)
In recent decades, magnetic field sensors have evolved into mass products with an extremely wide field of application. Typical applications are their use in computer hard disk drives and mobile phones. Today, magnetic field sensors are available as cost-effective standard components and are constantly evolving in terms of their properties. Our group has been developing magnetic field sensors for over 20 years and has contributed to numerous technological innovations.
An important and promising field of application of magnetic sensors is monitoring and safety technology. Perimeter protection systems are of great importance in various infrastructural facilities including buildings, airports, power plants etc. The principle of a cable with integrated magnetic field sensors and on-board computing allows much cheaper perimeter protection compared to existing technologies for kilometer-long fences. Owing to improvements in chip fabrication, a variety of sensors are now inexpensive and can be adapted to suit the prerequisites for providing a reliable perimeter protection system. Furthermore, the cable offers a greatly expanded functionality, since not only vibrations of a fence can be observed, but also magnetic stray fields of objects with just a few ferromagnetic components of vehicles that drive over a cable located in the ground can be detected. Additionally, employing on-board computing, the problems of “single point of failure” and latency can be greatly improved upon.
Because of these benefits, a highly competitive new technology could emerge in the high-revenue and ever-expanding civil security technology market, which could outperform or complement existing or alternative technologies in multiple technological, economic, and environmental criteria.
Goal of the project
The overall goal of the EFRE-funded project is to develop a cost-effective, reliable and scalable cable demonstrator that allows integrated sensors to monitor extended fences and access roads in real-time with high accuracy. A single cable may typically have a length of several meters to several kilometers and it can be connected to extensive networks. The cables can easily be retrofitted to existing fence systems or integrated into new fence systems. On-board computing with smart algorithms differentiates noise and other disturbances from intrusion attempts, and communication between edge-devices enables a very high accuracy. Equipping the individual sensors to communicate not only with a central evaluation unit, but also with surrounding sensors enable the use of swarm intelligence methods to discriminate atypical magnetic field fluctuations of those caused, e.g., by typical environmental influences (wind, rain, fog, etc.) and lower the false alarm rates. Since the entire evaluation is computer-based, it can be easily and modularly adapted to respective applications.
The main task will be to find the best possible compromise between sensitivity, real-time capability, low false alarm rate and manufacturing costs. The scaling and modular interconnection of multiple cables with thousands of individual sensors will be technically tested on larger scales in field trials.
The benefits of such intelligent sensor modules extend beyond surveillance. The methods to handle incoming sensor data, which includes noise-filtering, isolating interesting patterns and making decisions about these patterns will be largely generic. This gives us the opportunities to broaden the application domain, including preventive maintenance in industrial equipment, environmental control, traffic management and many others.
Project Current Status
At this stage, we have a standalone system that can differentiate between different patterns of disturbance on a model fence. As a proof of concept, we have deployed the system to differentiate between the fence vibration patterns generated by two different types of balls (a basketball and a tennis ball) on the fence. A block diagram of the development phase of the system is show in Figure 1. The device acquires data, treats it and sends it to a computer for storage, analysis and model development. At the end of this stage, we have a Machine Learning model that classifies the pattern. The model is then deployed on a microcontroller to perform real time classification of patterns (Figure 2).
Figure 1: Overview of system development phase
Figure 2: System deployment phase
This project has received funding from the European Regional Development Fund.