QualiBolS – Development of an AI-based acoustic inline monitoring system for the quality of stud welded joints

Metallbolzen auf eine Platte geschweißt mit Messmikrofonen darüber.

Description and Project Goals

Starting position for stud welding

In construction projects such as bridges, tunnels and buildings, bolted connections play a central role in securely attaching steel and metal components to concrete structures. Bolts are welded to steel girders and delivered to the job site as prefabricated components to be set in concrete. Ranging in diameter from 2 to 25 mm, these bolts transfer forces and moments from the steel to the concrete, making them safety critical components. Incorrectly welded studs can lead to structural failure and accidents.

A typical defect is a weld bead on one side caused by the deflection of the arc (blowing effect). Other problems include porosity and cracks in the weld metal. Stud welding requires a lot of experience, especially to avoid the blowing effect, which is achieved by grounding and balancing weights. Since this is often not done on every stud due to time constraints, the weld quality varies.

Current quality control for stud welding is based on destructive testing (bending the stud) and visual inspection by trained personnel. It is complex, expensive and time consuming.

The project in brief

The QualiBolS project is developing an acoustic inline quality monitoring system for stud welded joints that enables non-destructive and non-contact 100% inspection. By monitoring the sound generated by the arc, conclusions can be drawn about the quality of the welded joint. Using MEMS microphones and AI algorithms, the arc sound is recorded and automatically analyzed. For the first time, the AI-based evaluation makes it possible to distinguish whether the weld is OK or not during the welding process. Welding sounds can also be used to differentiate the type of welding defect (classification). With the help of the determined correlations, recognized welding defects can then be corrected. Process parameters (e.g. welding current, voltage, welding time) are also included in the evaluation to increase the detection accuracy.

Innovation for the industry

The goal of QualiBolS is to achieve a detection accuracy of at least 96%. The new inspection system will then enable 100% non-destructive testing of stud welds for the first time.

 

Responsibilities of Fraunhofer IDMT

  • Development of a measurement setup for the investigation of the acoustic situation during stud welding (requirement profile, test plan for the generation of training and test data)
  • Development of AI algorithms for quality control of stud welded joints (classification strategy for welding defects, data analysis)
 

Member of ZIM-Network »AkuPro« (German speaking)