An Integrated PCA-ICA Approach for Early-Stage Damage Detection

The combination of a model-free approach with Blind Source Separation (BSS) has emerged as a robust strategy to detect structural damage in the field of Structural Health Monitoring (SHM). Because of its inherent flexibility and adaptability, the model-free approach is well-suited for practical applications involving complex real structures, where accurate modeling is challenging. Meanwhile, BSS effectively isolates individual source signals from mixed signals, even in noisy environments. This study introduces an innovative SHM methodology that combines Principal Component Analysis (PCA) and Independent Component Analysis (ICA) to extract reliable damage-sensitive features: PCA is used for data normalization and dimensionality reduction, followed by ICA, which enhances vibration signatures indicative of structural damage. To quantify differences between signals collected at different time intervals as damage indicators, relevant metrics are introduced. The efficacy of this technique in early-stage damage detection is evaluated using a Finite Element model-based dataset. A bridge structure is numerically modeled and its structural response under random environmental load conditions is evaluated, both in healthy state and with simulated damage scenarios. The proposed method enables the tracking of temporal changes in the introduced damage indicators, thereby possessing the capability to detect damage characterized by gradual degradation.