Research On Reliability Optimization Of Ship Structure Based On Dynamic Approximate Model

Ships need to bear uncertain loads such as cargo and people,as well as random loads from external wind and waves during the voyage.Meanwhile,the ship's carrying capacity is uncertain due to the uncertainty of material characteristics and structural dimensions during the actual construction of the ship.In recent years,due to insufficient ship's load carrying capacity,many ships have suffered sea damage accidents that caused partial failure,damage or even overall breakage in dangerous sea conditions.Therefore,it is necessary to carry out reliability analysis on the ship structure to ensure adequate safety.At present,ships are developing towards large scale,which leads to more steel consumption,energy consumption and pollution discharge.The adoption of lightweight technology can help save energy and reduce emissions,and improve the economic performance of ships.At present,the lightweight in the field of ship mainly focuses on the deterministic optimization,while the reliability optimization has great advantages because it helps to improve both economy and reliability.Approximate models are ususlly used to replace time-consuming finite element simulation in reliability analysis and reliability optimization design of engineering structures.The traditional static approximate model uses one-shot sampling to construct the approximate model.However,the ship structure is a complex large-scale system,and the reliability analysis and optimization design of the ship face problems such as multiple parameters,strong nonlinearity,high dimensions,and multi-peak,so it is difficult to construct an approximate model with high fitting accuracy by only one-shot sampling.Moreover,ship engineering issue often cannot provide a large number of sample points due to excessive calculation costs,so it is necessery to constructe a high-precision approximate model with small-scale sample points.This paper constructs dynamic approximate models with high accuracy and high computational efficiency for ship structural reliability analysis and reliability optimization design,respectively,so as to achieve a high degree of fitting to the finite element model with as few sample points as possible.Aiming at the problem of ship structure failure due to insufficient ultimate bearing capacity,an active learning reliability method combining Kriging and Monte Carlo simulation(AK-MCS)has been improved in this paper.This paper proposes the secondary selection criteria for sample points by combining the advantages of two active learning functions,and applies the k-fold cross-validation method to form a sequential sampling strategy based on active learning.Based on this strategy,a dynamic Kriging model is constructed,and the MCS method is used to calculate reliability.The feasibility and efficiency of the method are verified by two mathematical models of three-span beam and nonlinear oscillator.This method is used to analyze the reliability of the ultimate strength of the bottom stiffened plate under combined load.This method guarantees accuracy and effectively reduces calculation costs.Therefore,it is verified that the improved AK-MCS method has good applicability and high efficiency in the study of ship structure reliability.By embedding the dynamic Kriging approximate model based on the interest subdomain and multi-island genetic algorithm into the Sequential Optimization and Reliability Assessment(SORA)method,this paper proposes a reliability optimization design method based on the interest subdomain.This method proposes the concept of interest subdomain based on the SORA method,determines the range of interest subdomains,and formulates adaptive spatial reduction rules,and then proposes an adaptive spatial reduction sequence sampling strategy based on interest subdomains,thereby constructing a dynamic Kriging approximate model that locally highly fits the subdomain of interest with as few sample points as possible.The paper proposes a probability constraint feasibility check method to reduce unnecessary reliability assessment processes.A mathematical example is used to verify the calculation efficiency and accuracy of the reliability optimization strategy of the interest subdomain.The method is applied to the reliability optimization design of ship cabin structure,and the results show the efficiency and feasibility of the method.