Neural network approach to holographic nondestructive testing.

Area of Science:

• Optical Engineering
• Artificial Intelligence
• Materials Science

Background:

• Holographic interferometry is a powerful technique for nondestructive testing (NDT) of technical components.
• Automatic defect detection in holographic interferograms remains a challenge, often requiring manual analysis.

Purpose of the Study:

• To develop and evaluate a neural network approach for the automatic detection of defects in loaded technical components using holographic interference patterns.
• To define translation- and rotation-invariant features for robust defect identification.

Main Methods:

• A neural network model was trained using a dataset generated by simulating interferograms, guided by experimentally measured samples.
• Translation- and rotation-invariant features were defined based on the maximal local slope of intensity and pattern partitioning.
• The method was tested on practical holographic interference patterns from loaded technical components.

Main Results:

• The developed neural network approach successfully demonstrated the feasibility of automatic defect detection.
• The defined invariant features contributed to the robustness of the detection method.
• Practical results confirmed the effectiveness of the proposed technique for holographic NDT.

Conclusions:

• The neural network approach offers a viable solution for automated defect detection in holographic nondestructive testing.
• The method provides a scalable strategy applicable to various holographic NDT tasks.
• This work advances the automation capabilities in the field of component integrity assessment.