Skip to main navigation Skip to search Skip to main content

Probabilistic Structural Health Monitoring in Composite Structures using Large Language Models

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

1 Downloads (CityUHK Scholars)

Abstract

Carbon fiber reinforced polymers (CFRPs) are widely employed in aerospace engineering due to their high specific strength and lightweight properties. To ensure structural integrity under cyclic fatigue loading, ultrasonic guided wave based structural health monitoring (SHM) techniques have been developed to detect internal damage, particularly delamination, which critically reduces stiffness. However, existing damage imaging algorithms are constrained by limited data utilization, reliance on manually tuned thresholds, and low imaging accuracy. To address these challenges, we propose a novel Large Language Model-enhanced Probabilistic Damage Imaging (LLM-PDI) framework for high-precision visualization of damage in composite structures. First, the proposed Time Series Large Language Model (TSLLM) extracts damage-relevant features from path-level time-series signals and specimen-level textual prompts. These features are then fused through context-based multimodal reasoning within the LLM component to generate back-projection weights for each selected path. Subsequently, an improved probabilistic damage imaging (PDI) algorithm employs the generated weights to back-project signal damage indices (DIs) into the inspection area, yielding pixel-wise damage intensity. After normalization and thresholding at 0.5, the damaged region is visualized as a binary map. The method was validated on seven CFRP laminate specimens with varying ply orientations, achieving a Dice Similarity Coefficient (DSC) of 0.8491 on the test set, significantly outperforming existing imaging algorithms.
©2026 IEEE
Original languageEnglish
Title of host publication2026 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) Proceedings
PublisherIEEE
Number of pages6
ISBN (Electronic)979-8-3315-8938-7
ISBN (Print)979-8-3315-8939-4
DOIs
Publication statusPublished - 25 Sept 2026
EventInternational Instrumentation and Measurement Technology Conference 2026: Smart Instrumentation for a Sustainable World - Nancy Congress Center - Centre Prouvé, Nancy, France
Duration: 25 May 2026 → 28 May 2026
https://i2mtc2026.ieee-ims.org/

Publication series

Name
ISSN (Electronic)2642-2077

Conference

ConferenceInternational Instrumentation and Measurement Technology Conference 2026
Abbreviated titleI2MTC
PlaceFrance
CityNancy
Period25/05/26 → 28/05/26
Internet address

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 52405164).

Research Keywords

  • CFRP
  • Guided Wave
  • Damage Imaging
  • Large Language Model

Fingerprint

Dive into the research topics of 'Probabilistic Structural Health Monitoring in Composite Structures using Large Language Models'. Together they form a unique fingerprint.

Cite this