Supervised Detection of Connector Lock Events with Optical Microphone Data

Area of Science:

• Manufacturing Automation
• Robotics
• Signal Processing

Background:

• Automating connector plugging and locking is difficult due to positional variations and external disturbances.
• Quality assessment of these automated processes is challenging for traditional image-based systems.
• Existing methods struggle with process stability and reliable failure avoidance.

Purpose of the Study:

• To develop a robust method for automated quality assessment of connector locking processes.
• To leverage acoustic properties for reliable identification of successful connector locking.
• To improve automation stability and reduce failures in manufacturing assembly.

Main Methods:

• Utilized highly sensitive optical microphones for acoustic data acquisition.
• Applied various neural network architectures, including multimodal approaches.
• Conducted experiments under both laboratory and real manufacturing conditions.
• Analyzed inherent acoustic connector locking properties for signal identification.

Main Results:

• Achieved classification performance with accuracy levels close to 90%.
• Successfully distinguished connector locking signals from other machining events.
• Demonstrated the effectiveness of acoustic analysis in complex manufacturing environments.
• Multimodal neural network architectures yielded the highest classification performance.

Conclusions:

• Acoustic analysis combined with neural networks offers a promising solution for automating connector locking quality assessment.
• The proposed method significantly enhances process stability and reliability in manufacturing.
• This approach provides a viable alternative to challenging image-based quality control systems.