Optimization of Damage Features Contaminated by Nonstationary Colored Noise Algorithms using Johansen Cointegration

Abstract

For structural health monitoring (SHM), researchers have primarily investigated the effects of signals contaminated with stationary white noise, with very few studies examining pollution caused by highly correlated nonstationary colored noise, such as Brown noise. To address this issue, this paper proposes an optimization-based damage detection technique for composite structures exposed to nonstationary colored noise using condensed frequency response functions (CFRF) as damage-sensitive features (DSF). Two different signal-to-noise ratios (SNR) of Brownian motions, i.e., 20 and 10, are used in the investigation to contaminate CFRFs. Contamination produces nonstationary patterns, making it difficult to detect damage with vibration-driven methods. In this study, we propose a new goal function based on Johansen cointegration, an econometric concept. The proposed objective function converts nonstationary CFRF signals into stationary representations, subsequently fed into optimization-based model updating algorithms. The Reptile Search Algorithm (RSA) is employed to update unknown structural damage indices based on the constructed objective function. The new method is validated on a finite element (FE) model simulating composite laminates with different ply orientations. By comparing the proposed method to a damage detection approach in the literature, the superiority of the proposed method is demonstrated.