Dealing with Significant Noise Levels in Vibration‐based Bridge Health Monitoring? A novel ARMA+Noise algorithm in the Frisch Scheme Context

Abstract

AbstractSystem Identification (SysId) refers to an ensemble of methodologies, as the ones built on statistical autoregressive models, which are among the most effective tools for spectral analysis and, by extension, for vibration‐based assessment. However, the application of standard SysId strategies might be hampered by the non‐negligible levels of noise dominating in harsh environments (or those intrinsic to electronic devices). This is especially true in bridge‐related applications, where faint modal components at low frequency are very common. To this end, the ARMA+Noise algorithm is proposed in this work, which is built on a novel frequency‐domain adaptation of the Autogressive with Moving Average (ARMA) model in the Frisch scheme context: the technique is superior in that it can ensure the best trade‐off between the frequency resolution and the hidden signal noise to be identified. A dedicated workflow has been developed, that extracts the ARMA model parameters by combining the advantages of the AR+Noise identification method with the Graupe's algorithm. The validity of the proposed technique has been tested on the Z24 bridge dataset, showing that the ARMA+Noise solution can properly identify the first four modes of the structure, even when the signal‐to‐noise ratio is low.