Integration of Deep Learning Network and Robot Arm System for Rim Defect Inspection Application.

Area of Science:

• Industrial Manufacturing
• Artificial Intelligence
• Computer Vision

Background:

• Automated inspection is crucial for maintaining quality in large-scale manufacturing.
• Defect detection in automotive components like aluminum rims requires precise and efficient methods.

Purpose of the Study:

• To develop an automated system for detecting surface defects in forged aluminum rims for electric vehicles.
• To enhance the accuracy and speed of defect identification in industrial settings.

Main Methods:

• Utilized an eye-in-hand robot arm with a camera for 3D image acquisition.
• Employed deep learning, specifically convolutional neural networks (CNNs) and Generative Adversarial Networks (GANs), for defect detection and data augmentation.
• Developed a graphical user interface and a defect detection algorithm based on YOLO (You Only Look Once).

Main Results:

• Successfully generated additional training data using GANs and DCGANs.
• Achieved faster detection times and higher mean average precision (mAP) compared to existing methods.
• Demonstrated the system's accuracy and efficiency in identifying four types of defects: dirt spots, paint stains, scratches, and dents.

Conclusions:

• The developed AI system provides an effective solution for automated rim defect detection in industrial applications.
• The proposed method significantly improves the quality control process for electric vehicle components.