Decision fusion for damage localization in CFRP laminate using Lamb wave and acoustic emission

Abstract

Carbon fiber reinforced polymer (CFRP) structures are particularly vulnerable to barely visible impact damage during service, requiring advanced methods for assessment. This study presents a novel damage localization approach that synergistically combines Lamb wave (LW) and acoustic emission (AE) techniques through decision-level fusion. Experiment using low-velocity impacts generated complementary active LW and passive AE datasets, which were processed using Hilbert transform and envelope extraction for feature extraction. A deep learning regression model was developed to predict damage coordinates, augmented by an optimized weighted-average fusion strategy. However, due to signal distortion and sparse sensor coverage, single-modality methods suffer from degraded performance in certain regions, particularly near structural edges. Featurelevel fusion struggles to reconcile the nonlinear differences between LW and AE signals, often leading to information redundancy or loss. To overcome these limitations, this work proposes a novel decision-level weighted fusion framework that leverages the complementary strengths of LW and AE while preserving their individual signal characteristics. Comparative analysis shows that the localization errors in the x and y coordinates are reduced by 41.49 % and 38.36 %, respectively, compared to single-modality methods. In the edge regions, the errors in the x and y coordinates are reduced by 18.04 % and 14.44 %, respectively. The methodology demonstrates significant potential in practical implementation in CFRP structural health monitoring systems, offering enhanced reliability for impact damage assessment.