A multitask SHM algorithm to identify damage with random severity and location in IPE beams using EMI technique

• Published in: Structures (Oxford) Vol. 70; p. 107659
• Main Authors: Zamanian, Mehrab, Sepehry, Naserodin, Zahrai, Seyed Mehdi
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2024
• Subjects: Damage identification; Electromechanical impedance; Machine learning; Multitask learning; Structural health monitoring; Multitask learning; Electromechanical impedance; Damage identification; Structural health monitoring; Machine learning
• ISSN: 2352-0124; 2352-0124
• DOI: 10.1016/j.istruc.2024.107659

Existing electromechanical impedance (EMI) damage identification algorithms often face challenges in terms of generalizability. This paper presents a robust algorithm that can simultaneously estimate the region and severity of damage with random damage scenarios across a surface and any severity, rather than being limited to specific points and severities. The host structure was an I-beam. Simulated damage was introduced as a subtle added mass to evaluate the algorithm's effectiveness in early-stage damage identification. Initially, various EMI tests with different damage specifications were conducted, and validated through numerical simulation. Damage-sensitive features were extracted and were input into three ML models: support vector machine, random forest, and multilayer perceptron. An ensemble learning approach was employed to combine the individual predictions from these models. The algorithm achieved classification accuracies of 97.3 % and 94.4 % on the validation and test sets, respectively, for identifying damaged regions. The algorithm also quantifies damage severity, achieving R-squared values of 92 % and 88 % on the validation and test sets, respectively.