Whether classic metal C parts such as screws or other mass-produced workpieces, complex die-cast aluminum parts for automotive manufacturing or finished electronic devices — all of these items have to undergo a quality inspection before they are delivered. Until now, this was a time-consuming process that was performed manually or visually.

This is where the innovative solution developed in Fraunhofer CCIT’s "Marquise" research project comes to the rescue: Instead of the resource-intensive inspection that has been carried out up until now, it provides a magnetic field-based check in a matter of seconds. "Marquise" identifies the presented object and automatically detects production faults and defects that are difficult to identify visually — and can do this even during the production process if so required.

Having reached the end of the research phase, the technology is now available for use in practice.

 

Two technologies, one objective: smart quality control

"Marquise" is based on the integration of two Fraunhofer developments: One of the components is the IndLoc® material detection system from the Fraunhofer Institute for Integrated Circuits IIS, which excites conductive objects using a low-frequency magnetic field. This generates measurable secondary fields that map a unique signature of the workpiece as a “magnetic fingerprint.” Both different conductive materials, the form of a workpiece and even minimal differences from the required standard are detected — in a real time-capable and highly precise manner.

“To continue to expand the use cases and enable IndLoc to be used to identify thousands upon thousands of objects, with Marquise, we are utilizing the combination of algorithms and AI,” says Tobias Dräger from Fraunhofer IIS. “By taking this approach, we are able to leverage the potential of large volumes of data for various fields of application.”

With this in mind, the IndLoc® material detection solution is accompanied by a second component in the form of a system for powerful AI-based data processing from the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS. Taking into account special characteristics of the registered signals and physical laws governing magnetic fields, the system can be trained to detect individual objects or object groups in a data-efficient manner. Machine learning models, modern neural networks and highly effective data processing ensure that the application also identifies the slightest deviation from the “learned” standard using outlier detection algorithms — even in areas that cannot be differentiated using conventional computer vision methods for object detection.

 

Efficient, precise and versatile

Marquise impresses with its speed and precision — and not only in comparison with manual inspection. The application is also the superior solution compared with other common inspection methods: Users do not need a large, difficult-to-operate device in a radiation-safe environment, as is the case when using x-ray technology. What’s more, in contrast to visual methods, for example, the measurement result is not affected by poor lighting conditions or dirt, and it does not matter if the objects are concealed, stacked or packaged. Magnetic field-based quality control can therefore be integrated into ongoing operations of entire production lines using a relatively small, automated intermediate step.

As "Marquise" has a simple, inexpensive structure, it can also be used in many different ways — for example, to test material compositions and alloys. “A huge number of other fields are also possible,” says Tobias Dräger. “For example, 'Marquise' could be used in medical engineering to provide support in operations and identify the correct instrument. Or, as a detector of conductive materials, it could be put to use as an alternative to complex x-rays or as a way of locating metal foreign objects in food or separating waste effectively.” As "Marquise" also “looks through” packaging, the application offers huge benefits even during inspections on incoming goods, outgoing goods and returns. Products can be identified and inspected efficiently based on their metal components without having to unpack them.

Despite its wide-ranging fields of use, "Marquise" is easy to implement without a long training period: The application is quick to install and easy to use — it guides the user through the process they require in just a few steps on the computer screen.

 

Rapid benefits for users

“'Marquise' can be geared to users’ needs with great flexibility,” says Helen Schneider from Fraunhofer IAIS. The structure of 'Marquise' can be adapted to meet the requirements of the particular use case, as inspecting a large die-cast aluminum part will need a different system configuration and data acquisition approach to that of an individual screw, for example. The features of the processed training data are incorporated into a constantly growing library in compliance with data protection regulations. This enables knowledge to be compiled step by step in order to identify objects and object clusters at an increasingly rapid pace and respond to new requirements. Helen Schneider explains: “Put simply, the algorithms we are using at the institute are trained on each new set of objects using powerful hardware and the software at the application site is then updated. The application is then ready for use — compared with other inspection methods, it quickly achieves the required time and cost savings while ensuring an excellent level of quality.”

Can you see potential for the use of smart quality control involving Marquise in your company? Then seize the opportunity to work with Fraunhofer CCIT and integrate this efficient solution into your processes at little expense and effort.

 

Expertise from project partners in magnetic field-based object identification

Alongside sensors and wireless technologies for Industry 4.0, the RFID and Inductive Sensor Systems research group at Fraunhofer IIS delves deep into scenarios involving communication, positioning and material detection based on magnetic fields. These include the development of hardware for signal generation and signal analysis, antennas and firmware and software components used to control hardware modules, as well as the preprocessing, analysis and visualization of data. In addition to contract research and in-house research, the development of the IndLoc® technology comprised funding projects and a dissertation.

Fraunhofer IAIS is one of the leading scientific institutes in the fields of artificial intelligence, machine learning and big data in Germany and Europe, and works in the domain of AI projects on image recognition and interpretation alongside numerous companies from research and industry. As part of the "Marquise" project, Fraunhofer IAIS developed informed machine learning algorithms and prepared the data processing for future wide-ranging and complex use cases.