A reliability-based approach to determine the minimum detectable damage for statistical damage detection. (1st June 2021)

• Record Type: Journal Article
• Authors: Mendler, Alexander; Döhler, Michael; Ventura, Carlos E.
• Is Part Of: Mechanical systems and signal processing. Volume 154(2021)
• Publication Date: 2021-06-01
• Subjects: Damage detectability -- Ambient vibrations -- Sensitivity -- Fisher information -- Probability of detection -- Measurement duration; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2020.107561
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: The minimum detectable change in structural parameters is determined for damage detection. The analysis requires vibration data from the reference state and a finite element model. Detectability is defined through a reliability concept for monitoring using statistical tests. Detectable damages explicitly depend on damage sensitivity, noise, data length and reliability. The predicted damage detectability is confirmed on experimental data of a beam with added masses. Abstract: This paper derives a formula to determine the minimum detectable damage based on ambient vibration data. It is a key element to analyze which damage scenarios can be detected before a monitoring system is installed. For the analysis, vibration data from the reference structure as well as a finite element model are required. Minimum detectability is defined by adopting a code-based reliability concept that considers the probability of detection and the probability of false alarms. The results demonstrate that the minimum detectable damage depends on three elements: the uncertainty of the damage-sensitive feature (which decreases with increasing measurement duration), its sensitivity towards model-based design parameters, and the reliability requirements regarding the damage diagnosis results. The theory is developed for the stochastic subspace-based damage detection method but can be applied to any damage-sensitive feature provided its sensitivities and statistical properties can be characterized. For proof of concept, the minimum detectable change in stiffness and mass of a pin-supported beam are analyzed in a numerical and experimental study, respectively. The predictions with the developed approach appear to be accurate and robust to noise effects for both simulated and real data.