Multi-site damage precise localization via the random vibration functional model based method: Formulation & concept validation

• Published in: Mechanical systems and signal processing Vol. 160; p. 107880
• Main Authors: Sakaris, C.S., Sakellariou, J.S., Fassois, S.D.
• Format: Journal Article
• Language: English
• Published: Berlin Elsevier Ltd 01.11.2021; Elsevier BV
• Subjects: Damage localization; Data-based methods; Decision making; Functional model based method; Functional models; Localization; Multi-site damage; Multiple site damage; Performance degradation; Random vibration; Sensitivity analysis; Statistical time series methods; Vibration analysis; Vibration based SHM; Multi-site damage; iid; ARX; FM; Functional model based method; DSD; NLS; OLS; SHM; Data-based methods; Functional models; GA; SM; BIC; VFP-ARX; MSD; MM; TSD; SSD; Statistical time series methods; Damage localization; FMBM; SQP; FP-ARX; FRF; Vibration based SHM
• DOI: 10.1016/j.ymssp.2021.107880
• ISSN: 0888-3270

•Random vibration data based Multi-Site Damage (MSD) characterization and precise localization.•Two versions of a Functional Model Based Method are postulated.•Precise localization is demonstrated with very few sensors and training signal records.•The versions are experimentally validated and assessed based on a laboratory structure.•The effects of further reduction in the number of training signal records is studied.. The postulation and experimental validation of a random vibration data based Functional Model Based Method (FMBM) for multi-site damage localization, involving the determination of the number of damage locations and interval estimation of their coordinates, in presented. The method is based on Functional Models (FMs), along with proper statistical decision making and estimation techniques, and is capable of working with a limited number of random vibration signals under normal operating conditions. Two distinct versions of the method are introduced: A Forward or Multiple Model version which is based on multiple Functional Models, and a Backward or Single Model version based on a single counterpart. Both versions are experimentally validated and compared via hundreds of single- and double-site damage scenarios on the wing of a lab-scale aircraft skeleton structure, while a sensitivity analysis examining performance degradation for reduced numbers of training vibration signals is also undertaken. The results indicate high effectiveness in both determining the number of damage sites and estimating their precise coordinates along with their associated uncertainty.