A Deep Learning Framework for Vibration-Based Assessment of Delamination in Smart Composite Laminates

Area of Science:

• Materials Science
• Structural Health Monitoring
• Artificial Intelligence

Background:

• Delamination is a critical defect in composite materials, often caused by manufacturing flaws or operational stresses.
• Current assessment methods primarily rely on high-frequency guided waves and mode shape analysis.
• There is a need for advanced techniques to reliably detect and quantify delamination in smart composite structures.

Purpose of the Study:

• To present a deep learning framework for assessing delamination in smart composite laminates based on structural vibrations.
• To simulate and analyze the impact of delamination on the dynamic response of composite materials.
• To develop a robust method for distinguishing intact from delaminated composite structures.

Main Methods:

• An electromechanically coupled model was used to simulate inner and edge delaminations in piezo-bonded composites.
• Random loads were applied to healthy and delaminated structures, and transient responses were converted into spectrograms.
• A convolutional neural network (CNN) was designed to analyze spectrograms and classify structures as intact or delaminated.

Main Results:

• The CNN achieved high accuracy: 99.9% (training), 97.1% (validation), and 94.5% (testing) on unseen data.
• The model demonstrated effectiveness in distinguishing between intact and delaminated composite laminates.
• Analysis of the confusion chart provided insights into delamination severity and detectability for different scenarios.

Conclusions:

• The proposed deep learning framework offers a promising approach for structural vibration-based delamination assessment in smart composites.
• The study highlights the potential of CNNs in analyzing vibration data for defect detection and characterization.
• This method can enhance the reliability and efficiency of structural health monitoring for composite materials.