Impact Damage Localization in Composite Structures Using Data-Driven Machine Learning Methods

Area of Science:

• Mechanical Engineering
• Materials Science
• Artificial Intelligence

Background:

• Traditional impact source localization methods struggle with real-time detection on plate structures due to material property uncertainties.
• Accurate impact localization is crucial for assessing damage and ensuring the structural integrity of plate-like systems.

Purpose of the Study:

• To propose and validate a data-driven machine learning method for real-time impact source localization on plate-like structures.
• To demonstrate the method's effectiveness across structures with varying material properties.
• To provide a robust technique for impact damage detection in composite materials.

Main Methods:

• An experimental platform using piezoelectric transducers and an oscilloscope was developed for impulse response testing.
• Impact source locations were generated by releasing a steel ball onto test plates.
• Eigenvalues of arrival times from time-domain signals were extracted to create machine learning datasets.
• A Back Propagation (BP) neural network was employed to analyze sensor arrival time differences and predict impact locations.

Main Results:

• The proposed machine learning method successfully predicted impact source locations on plate-like structures.
• The technique demonstrated high accuracy and robustness, independent of specific material properties.
• Effectiveness was validated across three distinct plate-like structures with different material characteristics.

Conclusions:

• The developed data-driven machine learning approach offers a reliable solution for real-time impact source localization in plate structures.
• This method eliminates the need for prior knowledge of material properties, enhancing its applicability.
• The research provides valuable experimental methodologies and testing techniques for impact damage assessment, particularly in composite structures.