Non-probabilistic Informed Structural Health Assessment with Virtual Modelling Technique

Abstract

AbstractIn real-life engineering, non-probabilistic structural information is very common in many and varied disciplines. This class of information is characterized by incompleteness and imprecision, such as interval, fuzzy sets, etc. Non-probabilistic structural information can be reflected in the structural performance and cause it to fluctuate within a specific range, instead of being deterministic. Thus, without appropriate consideration of non-probabilistic information, serious or even disastrous accidents may occur. Therefore, fully estimating the structural health status using non-probabilistic information, especially detecting the lower and upper bounds of the concerned structural response, is extremely significant in uncertainty-sensitive fields. To conquer this challenge, a virtual modeling technique underpinning a structural health assessment framework is introduced. The twin extended support vector regression (T-X-SVR) approach is embedded for virtual model construction. Continuous, differentiable expression of the established virtual model allows the optimal solutions for each interval analysis to be easily achieved. Information update is another inherent feature, which enables structural health assessment to be implemented with updated conditions without rebuilding the virtual model. To demonstrate the applicability of the proposed virtual modeling technique underpinned structural health assessment framework, the non-probabilistic informed elastoplastic nonlocal damage analysis was investigated for engineering structures.