Uncertainty Handling in Structural Damage Detection via Non-Probabilistic Meta-Models and Interval Mathematics, a Data-Analytics Approach

Recent advancements in sensor technology have resulted the collection of massive amounts measured data from structures that are being monitored. However, these include inherent measurement errors often cause assessment quantitative damage to be ill-conditioned. Attempts incorporate a probabilistic method into model provided promising solutions this problem by considering uncertainties as random variables, mostly modeled with Gaussian probability distribution. success methods is limited due lack adequate information required obtain an unbiased distribution uncertainties. Moreover, surrogate models involve complicated and expensive computations, especially when generating output data. In study, non-probabilistic based on wavelet weighted least squares support vector machine (WWLS-SVM) proposed address uncertainty vibration-based detection. The input for WWLS-SVM consists selected packet decomposition (WPD) features structural response signals, Young’s modulus elements. This calculates changes lower upper boundaries interval analysis method. Considering parameters, used predict interval-bound output. approach applied detect simulated four-story benchmark structure IASC-ASCE SHM group. results show performance superior direct finite element uncertainty-based detection requires less computational effort.