A fast Monte-Carlo method to predict failure probability of offshore wind turbine caused by stochastic variations in soil. (1st March 2021)

• Record Type: Journal Article
• Authors: Oh, Ki-Yong; Nam, Woochul
• Is Part Of: Ocean engineering. Volume 223(2021)
• Publication Date: 2021-03-01
• Subjects: Monte-carlo method -- Offshore wind turbine -- Soil modulus spatial variation -- Random field -- Natural frequency -- Failure probability; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2021.108635
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: As spatial seabed soil variations are stochastic and considerable, it is important to predict the changes in the dynamic behavior of offshore structures for safety. Particularly, if the natural frequency of a structure is significantly changed by the soil variations, resonance can cause structural failure; hereafter, this probability is referred to as failure probability. Although several models have been proposed, they require a long computational time to predict failure probabilities. To overcome this limitation, this work proposes a fast method for estimating the failure probability. First, since a very large computation memory is required to construct a finite element model for stochastic soil, block-wise matrix calculations and a dynamic memory allocation technique were adopted. Second, conventional models require a large number of soil samples to calculate the failure probability, which results in heavy computations. A newly developed fast Monte-Carlo (FMC) method is 5.8 times faster than the conventional method with high accuracy (99.41%). This method was applied to offshore wind turbines and successfully predicted various structural characteristics. A noteworthy prediction is that slender and long foundation is more robust to stochastic soil variations than thick and short foundation. The FMC method can be used for preliminary design of offshore structures. Highlights: Failure probability of offshore wind turbine due to spatial variation in soil and resonance is calculated. Block-wise matrix calculation and dynamic memory allocation are used to generate 3D stochastic soil domains. A fast Monte-Carlo method significantly reduces computation time to estimate the failure probability. Model predicted that a long and slender substructure is more reliable than a thick and short one.