Fast unsupervised learning methods for structural health monitoring with large vibration data from dense sensor networks

Abstract

Data-driven damage localization is an important step of vibration-based structural health monitoring. Statistical pattern recognition based on the prominent steps of feature extraction and statistical decision-making provides an effective and efficient framework for structural health monitoring. However, these steps may become time-consuming or complex when there are large volumes of vibration measurements acquired by dense sensor networks. To deal with this issue, this study proposes fast unsupervised learning methods for feature extraction through autoregressive modeling and damage localization through a new distance measure called Kullback–Leibler divergence with empirical probability measure. The feature extraction approach consists of an iterative algorithm for order selection and parameter estimation aiming to extract residuals in the training phase and another iterative process aiming to extract residuals only in the monitoring phase. The key feature of the proposed approach is the use of correlated residual samples of the autoregressive model as a new time series at each iteration, rather than handling the measured vibration response of the structure. This is shown to highly reduce the computational burden of order selection and feature extraction; moreover, it effectively provides low-order autoregressive models with uncorrelated residuals. The Kullback–Leibler divergence with empirical probability measure method exploits a segmentation technique to subdivide random data into independent sets and provides a distance metric based on the theory of empirical probability measure with no need to explicitly compute the actual probability distributions at the training and monitoring stages. Numerical and experimental benchmarks are then used to assess accuracy and performance of the proposed methods and compare them with some state-of-the-art approaches. Results show that the proposed approaches are successful in feature extraction and damage localization, with a reduced computational burden.

Data-Driven Structural Health Monitoring Using Feature Fusion and Hybrid Deep Learning

“…Thus, one adopts the Grid search technique to test all possible combinations of hyperparameters for the identification of the optimal architecture. Specifically, k varies in the range [5,50], L k in [10,100], and N h in [3,30]. It is noted that other parameters are set by common values found in literature and fixed throughout the whole training process.…”

“…It is shown that the method is able to identify all investigated damage levels in an unsupervised learning fashion, which is significantly useful in real application because of the scarcity of data related to different damage scenarios. Entezami et al [3] developed a fast unsupervised approach for SHM having the capability of dealing with large vibration data based on the AR models for feature extraction and Kullback-Leibler distance for classification. The applicability of the method was supported through both numerical simulation and experimental datasets, with emphasis on its computational efficiency and highly accurate damage localization.…”

An Unsupervised Learning Approach for Early Damage Detection by Time Series Analysis and Deep Neural Network to Deal with Output-Only (Big) Data

“…The central core of all these methods relies upon statistical pattern recognition, and comprises feature extraction and feature classification. The former step is a signal processing strategy, which aims at extracting meaningful information (here called damage-sensitive features) from raw measured data (e.g., acceleration time histories), while the latter is a machine learning algorithm for analyzing and classifying the extracted features for early damage detection, localization and quantification [4][5][6][7]. Time series modeling is one of the powerful feature extraction methods, which is intended to fit a parametric representation (model) to raw measured data [8,9].…”

A Time Series Autoencoder for Load Identification via Dimensionality Reduction of Sensor Recordings

“…Synthetic features like spectral peak frequencies, usually exploited when the acquired data are shaped as Time Series (TS), are extracted to solve engineering tasks, like load identification and Structural Health Monitoring (SHM) [1,2]. Deep Learning (DL) allows extracting features from the data according to the required task, avoiding any preliminar feature design [3][4][5][6]. Among DL techniques, AutoEncoders (AEs) are special type of Neural Networks (NN) able to obtain a reduced data representation [7], also called latent representation, without specifying the task the reduced data representation must be used for.…”