Abstract:This paper presents a two-step vibration-based strategy for damage identification of framed structures using ensemble bagged trees known as a well-known supervised machine learning (ML) paradigm in conjunction with evolutionary optimization algorithms. The proposed model incorporates the actual response, wavelet coefficients, and wavelet energy to extract damage-sensitive features from the time-domain of the measured and simulated signals. Unlike available studies in this scope, the key objective of this research is to identify damage with a localization precision down to a single structural member, rather than limiting the evaluation to the group of elements. In order to increase the training performance in contributing to extremely large datasets with numerous class labels, the proposed strategy involves the artificial generation of features. Additionally, a modified genetic algorithm is proposed for fast damage localization. It is deduced that the damage locations are confidently detected within a fast computational time. Subsequently, damage identification is followed by the application of evolutionary optimization algorithms. For comparison purpose, the employment of the water cycle optimization algorithm (WCA) is comparatively investigated with the other three state-of-the-art optimizers, i.e., particle swarm optimization (PSO), imperialist competitive algorithm (ICA), and differential evolution algorithm (DE). The numerical and experimental validation studies evidence satisfactorily reliable identification results with no false detection in dealing with multiple damage scenarios in large-scale and real-world applications. It is concluded that developing the most damage-sensitive features and using the proposed data fusion strategy lead to informative features with a reasonably small size and significantly improve the ML performance.

Structural Damage Identification Using Gravitational Search Algorithm

Structural damage identification using static test data and changes in frequencies

Output-Only Damage Identification Using Enhanced Structural Characteristic Deflection Shapes and Adaptive Gapped Smoothing Method

Damage identification based on improved GSO algorithm

Robust Structural Damage Detection and Localization Based on Joint Approximate Diagonalization Technique in Frequency Domain

Structural Damage Identification Based on Modified Cuckoo Search Algorithm

Structural Damage Detection Using Changes in Natural Frequencies: Theory and Applications

Substructure Damage Identification Using Damage Tracking Technique

Structural Damage Identification Based On The Anti-Resonant Frequency And Least Square Algorithm

Structural damage detection based on residual force vector and imperialist competitive algorithm

Structural damage identification using an optimization technique based on generalized flexibility matrix

Structural Damage Detections Based on a General Vibration Model Identification Approach

Structural damage identification by support vector machine and particle swarm algorithm

Time-Domain Structural Damage Identification Using Ensemble Bagged Trees and Evolutionary Optimization Algorithms

Vibration-Based Structural Damage Identification under Varying Temperature Effects

STUDY ON STRUCRTUAL DAMAGE IDENTIFICATION BASED ON FPE PROCEDURE AND GENETIC ALGORITHM

Structural damage detection based on cross-correlation function with data fusion of various dynamic measurements

Identification of Incipient Damage Using High-Frequency Vibrational Responses

Improved Damage Detection for Beam-type Structures Using a Uniform Load Surface

Structural Damage Localization and Quantification Based on Additional Virtual Masses and Bayesian Theory

Structural damage identification using system dynamic properties