Wind-induced fragility of a monopole structure via Artificial Neural Network based surrogate analysis

Abstract

Although brute-force Monte Carlo (BFMC) sampling has been commonly used in performance-based wind engineering (PBWE) due to its robustness and efficiency, the computing cost required by the considerable amounts of repetitious, stochastic realizations for situations that involve large structures and a multitude of uncertainties is still prohibitive. To alleviate the computational burden, this communication proposes a surrogate modeling approach, based on artificial neural networks (ANNs), to evaluate the probabilistic integral required to quantify wind-induced fragilities. The resultant ANN models form a computationally viable alternative between input and output variables, based on a database of observations obtained through BFMC stochastic simulations. A preliminary study is conducted to examine the fragility of a slender, monopole tower structure under the excitation of multidirectional, mixed-climate wind loads. The ANN-powered surrogate results show adequate accuracy while drastically reducing the computing time to less than 1% of the cost of BFMC approach. It is promising to incorporate the surrogate ANN modeling in a PBWE framework.