Learning methods for structural damage detection via entropy‐based sensors selection

Abstract

In this article the problem of data‐driven structural damage detection is considered exploiting historical data collected from a structure. First, a novel technique based on Kalman filtering and on a combination of regression trees theory from machine learning and auto‐regressive system identification from control theory is derived to build switching models that can be used to detect structural damages. A technique is also proposed leveraging principal component analysis together with the poly‐exponential approach to create nonlinear models to be used for structural damage detection. Finally, a novel sensors selection algorithm based on the notions of entropy and information gain from information theory is developed to reduce the number of sensors without affecting or even improving, as it happens in our experimental setup, the model accuracy. The presented techniques are validated on three independent experimental datasets, showing that the proposed algorithms outperform previous and classical approaches, improving the prediction accuracy and the damage detection sensitivity while reducing the number of sensors.

Robust adaptive Kalman filter for structural performance assessment

“…Let xj,k be the estimate of the Jerk at time k, that is, x j,k , given Y k = {y 0 , … y k }. Then, based on the Yule-Walker algorithm, 𝛼 k and 𝜎 2 k in the process model (1a) can be calculated in a recursive way as follows:…”

“…Structural health monitoring (SHM) is an ongoing process involving continuous observation and analysis of engineering structures, for example, steel structures, over time. 1,2 Recent severe earthquakes, such as the 2019 Peru earthquake and the 2021 Haiti earthquake, have emphasized its importance, in which the structural performance assessment is a critical stage in SHM for steel structure maintenance and risk management. Over the last two decades, in order to assess the steel structural health and detect its potential structural damage or degradation, various SHM approaches have been proposed, including model-based, data-based, and performance-based approaches.…”

“…Dynamic state estimation algorithms, such as Kalman filtering (KF) and KF-like filtering, have been extensively used for the tracking problem. 2,23 However, the so-called measurement-origin uncertainty and target motion uncertainty might arise in the aforementioned approaches since both the dynamic model and noise parameters are estimated from measurements. 24,25 Therefore, some robust versions based on the KF or KF-like algorithms have been developed to guarantee some robustness to modelling uncertainties.…”

“…Structural health monitoring (SHM) is an ongoing process involving continuous observation and analysis of engineering structures, for example, steel structures, over time 1,2 . Recent severe earthquakes, such as the 2019 Peru earthquake and the 2021 Haiti earthquake, have emphasized its importance, in which the structural performance assessment is a critical stage in SHM for steel structure maintenance and risk management.…”

“…In addition, let us turn to design the tracking algorithms. Dynamic state estimation algorithms, such as Kalman filtering (KF) and KF‐like filtering, have been extensively used for the tracking problem 2,23 . However, the so‐called measurement‐origin uncertainty and target motion uncertainty might arise in the aforementioned approaches since both the dynamic model and noise parameters are estimated from measurements 24,25 .…”