Acoustic-based Non-Destructive AI Inspection 2026/09/29

Introductions

Traditional methods for detecting internal defects in metal products, such as magnetic particle inspection, destructive testing, and manual microscopic inspection, are inefficient and rely heavily on operator experience. Late defect detection leads to increased rework and scrap, raising production costs and hindering carbon reduction and sustainability goals, ultimately affecting competitiveness.

Technical Advantages and Features

• AI-based Acoustic Visualization and Recognition: Acoustic imaging technology for capturing acoustic fingerprints of metal materials. Utilizing AI deep learning and attention mechanisms to analyze impact sounds, the system automatically identifies internal defects with an accuracy of up to 95%.
• Automation and Standardization: Standardized learning processes improve and optimize human experience-based judgment. Improved efficiency allows transformation from sampling inspection to full automated inspection. Applicable to various components, enabling non-destructive testing.
• Cost Reduction and Data Optimization: Intelligent sample selection reduces labeling costs by over 70%. Detection accuracy improves by 13% compared to manual inspection. Systematic data management enables dynamic optimization of the model.

Industrial Applications and Business Opportunities

• Industry Applications
  • Metal manufacturing industry and semiconductor equipment component manufacturing industry.
  • Intelligent Factory.
• Application Examples
  • Successfully implemented at Taiwan’s largest powder metallurgy manufacturer for electric vehicle brake component inspection. It improves detection accuracy from 85% to over 98%, significantly reducing rework and scrap losses. It also reduces approximately 433 tons of CO2e emissions annually, effectively supporting the industry’s transition toward smart manufacturing and sustainability.