On the Use of Risk Measures in Digital Twins to Identify Weaknesses in Structures
Please login to view abstract download link
The increasing maturity of sensor technology and advanced numerical method for structural analysis make digital representation of structures (Digital-Twins) achievable and highly promising. Given that all materials exposed to the environment and/or undergoing loads eventually age and fail, the task of trying to detect and localize weaknesses in structures is important in many fields. Given measurements from sensors and a set of standard forces, an optimization based approach for system identification was introduced in our previous work. Our current works looks into the impact of uncertainty in our problem formulation. We allow the load and measurements to be random variables. The conditional-value-at-risk (CVaR) is minimized subject to the elasticity equations as constraints. CVaR is a risk measure that leads to minimization of rare and low probability events which the standard expectation cannot. An adjoint based approach is developed with quadrature in the random variables. We study the impact of the Gauss quadrature order chosen and the dimensionality of the random variables. We study multiple sources of uncertainty; the aleatoric ones, such as intrisic measurement errors in the sensors and epistemic ones, such as loads not properly accounted for (e.g. wind distributions in a large structure). Results are presented in the context of a plate, a large structure with trusses similar to those used in solar arrays or cranes, and a footbridge.
