A grey Bayesian inference framework for structural damage assessment. Issue 3 (18th November 2021)

• Record Type: Journal Article
• Authors: Fang, Sheng‐en; Chen, Shan
• Is Part Of: Structural control and health monitoring. Volume 29:Issue 3(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-11-18
• Subjects: damage assessment -- evolutionary particle pool -- grey Bayesian inference -- grey population Monte Carlo sampler -- particle interchange mechanism; Structural engineering -- Periodicals; Structural control (Engineering) -- Periodicals; Automatic data collection systems -- Periodicals; Detectors -- Periodicals; 624.17
• DOI: 10.1002/stc.2889
• Languages: English
• ISSNs: 1545-2255

Abstract

Summary: Real‐life civil structures often stay in a grey situation due to lack of knowledge and measurements. In a poor information condition, classic Bayesian inference is practically difficult to perform owing to intractable likelihood functions. Under such circumstance, a structure should be defined as a grey system. Thereby, a grey Bayesian inference framework has been proposed by incorporating the grey theory with approximate Bayesian computation for damage assessment purposes. Structural parameters are represented by interval grey variables, whose grey kernels and degrees of greyness are used to define probability distributions for sampling. A grey population Monte Carlo sampler has also been developed by introducing interval grey numbers and an evolutionary particle pool into the particle filtering process. A novel particle interchange mechanism is simultaneously proposed for effective particle interflow. Meanwhile, stochastic response surfaces are used to correlate structural parameters with responses for fast response computation. Also, likelihood calculation is replaced with a distance measure between simulated and measured samples. Lastly, the proposed inference strategy has been verified against both the numerical and experimental beams having the different damage scenarios. The damage locations and severities were successfully identified by the changes in the grey intervals and kernels.