Full-Scale Structural and Nonstructural System Damage Detection Using Wavelet-Based Fast Independent Component Analysis

Author

Sabal Bhandari, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2025

Degree Name

Master of Science

Department

Civil Engineering

First Advisor

Tezcan, Jale

Abstract

Structural Health Monitoring (SHM) has become an essential part of civil engineering, providing tools to evaluate the safety and performance of structures throughout their lifespan. Experience from past earthquakes shows that while complete structural failures are uncommon, one or more structural elements may sustain damage during moderate to severe shaking. In addition, nonstructural components such as ceilings, partitions, and mechanical or electrical equipment frequently experience failures that cause high financial losses, downtime, and safety risks. A persistent challenge in SHM is that accelerometers capture blended vibrations from multiple sources. Sudden damage changes a structure’s vibration characteristics, but these changes are not easily identified. Independent Component Analysis (ICA) offers a mathematical framework to separate mixed signals into statistically independent components, facilitating the detection of abnormal responses. Structural damage often appears as a change in stiffness, typically analyzed in the frequency domain. However, conventional tools such as the Fourier Transform (FT) and Fast Fourier Transform (FFT) cannot capture short-lived transients associated with cracking or spalling. Wavelet transforms overcome this limitation by providing simultaneous time–frequency resolution, making them effective for identifying sudden changes in vibration characteristics. This study combines wavelet-based preprocessing with FastICA to detect system-level damage using data from the Building Nonstructural Components and Systems (BNCS) test program at NEES@UCSD. Although the BNCS program primarily investigated nonstructural systems, the recorded acceleration responses reflect the integrated behavior of both structural and nonstructural elements. Because both contribute to mass and stiffness, the method identifies overall system damage without distinguishing between the two. A six-story reinforced concrete building was constructed, equipped with 24 accelerometers, and subjected to a series of earthquake simulations in both base-isolated and fixed-base configurations. The results indicate that sudden damage appears as spiky independent components, and the damage locations predicted by the proposed method align with observed damages reported in the BNCS study, demonstrating the promise of this approach for damage detection.