Friction Stir Welding Real-Time Monitoring of Acoustic Signals During Friction Stir Welding

Related Vendors

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

SIMTOS (Seoul International Machine Tool Show)

In the FSW-AcoMon research project, the project participants have developed a modular measuring system that records all acoustic signals from the friction stir welding of aluminum alloys and records them synchronously with the machine parameters. This provides the basis for AI-supported real-time monitoring.

The research project FSW-AcoMon, conducted by Fraunhofer IDMT, the Department of Manufacturing Technology at the Technical University of Ilmenau, and RRS Schilling GmbH, aims to identify process deviations in friction stir welding of aluminum alloys at an early stage, before they lead to scrap or costly rework.While traditional quality controls typically come into play only after production, acoustic monitoring allows for direct observation of the manufacturing process—providing an opportunity to increase process reliability and reduce downtime.

Flexible Measuring System for Different Production Conditions

The core of the previous work is a measurement concept in which microphones can be flexibly positioned directly on the welding spindle or at the joining site. At the same time, process data such as speed, feed, force, and tool position are synchronized and recorded. The modular sensory approach allows the microphones to be quickly adjusted to different machines and welding conditions—an advantage for later industrial application."The modular design of our measurement system enables precise representation of different machine environments and process conditions. This lays the foundation for a scalable industrial application of acoustic monitoring," explains project leader Olivia Treuheit from Fraunhofer IDMT.The data captured during friction stir welding form the basis for the development of AI models that can automatically provide indications of process deviations and wear on the FSW tool. This allows manufacturing processes to be analyzed more precisely and adjusted early if necessary.

Recognize Tool Wear with Results

The initial measurements already provide early indications of wear on the FSW tool and process deviations. An AI model for classifying tool conditions was able to identify heavily worn tools with an accuracy rate of over 99 percent. Additionally, a multi-class classification was tested, in which the model distinguishes between six levels of wear ranging from 0 percent to 40 percent. This model already achieved 80 percent accuracy, demonstrating that finer differences in tool wear can also be recognized."In our laboratory environment at TU Ilmenau, we can thoroughly analyze and assess the measurement results and understand how acoustic signals relate to different welding conditions. This provides the consortium with the scientific basis for the targeted application of AI models that reliably evaluate the process and the tool condition," says Professor Jean Pierre Bergmann, Head of Manufacturing Technology at TU Ilmenau (Germany).

The results achieved so far are promising for industrial applications and serve as an important indication that acoustic monitoring can be well integrated into existing manufacturing processes and provides reliable insights into the process and tool wear.
 In the second phase of the project, the focus will be on testing and evaluating the suitability of additional AI models and the integration of acoustic signals and machine parameters. The goal is to create a robust, retrofittable system that monitors friction stir welding of aluminum alloys in real time, thereby enabling continuous quality control.