Ship Uncertainty-based Design Optimization Based On Mixed Uncertainty Modeling

At present,most ship design optimization models are based on deterministic optimization theory.However,in the actual design process,ship optimal case will be affected by many uncertain factors,which may lead to a bad result with some hidden dangers in practical application or be unable to meet the designers’ demands for ship products.Therefore,as the key part of the ship preliminary design,the ship design optimization needs to fully consider various uncertainties in the early design stage to ensure that the optimal case can adapt to the fluctuation of uncertain parameters,that is,to conduct the research on ship uncertainty-based design optimization.Currently,ship uncertainty-based design optimization still falls behind.First,designers do not know enough about the mechanism of the uncertainty,blindly classify all uncertain factors as aleatory uncertainties and adopt the probabilistic method to model them,leading to bad preliminary identification of the uncertainties;Second,for uncertainty quantification,analysis and propagation considering the aleatory uncertainty,most designers still use the traditional time-consuming Monte Carlo method,which is obviously unfeasible for ship uncertainty-based design optimization which requires massive simulations and uncertainty analysis;Besides,the current research on ship uncertainty-based design optimization almost all focuses on the optimization considering aleatory uncertainty,ignoring modeling and optimization under the epistemic uncertainty,caused by the lack of designer’s knowledge,data and incomplete information.Even though the related researches in this field have made promising progress,there are still obvious shortcomings;In addition,the current uncertainty-based design optimization usually uses a single-source uncertainty quantification method which does not consider the synthesis and combination of uncertainty information from different sources,thus after consideration of epistemic uncertainty,it is a major difficulty to conduct the uncertainty analysis of these two types of the uncertainty in a unified framework.In summary,in ship uncertainty-based design optimization,the traditional method is far from sufficient to improve the robustness and reliability of ship system performance economically.New theories and methods that can deal with this optimization problem are urgently needed.Under the circumstance of new design background and requirements,this dissertation has conducted and completed the following researches to solve the above problems:1.Research on uncertain factors identification for ship design optimizationAccording to the collected ship data,environmental data and expert data,the parameter estimation and goodness-of-fit methods are combined to analyze aleatory and epistemic uncertainties that affect the design optimization objectives/constraints and complete the uncertainty identification towards ship design optimization,uncertain factors are rationally classified into three different types of uncertainty,namely,strong statistical variables,sparse variables and interval variables.2.Research on the uncertainty quantification and propagation method under the influence of multi-source aleatory uncertaintyConsidering the influence of aleatory uncertainty,multi-dimensional polynomial chaos expansion method is used instead of the traditional Monte Carlo simulation method to conduct the uncertainty analysis and propagation.Based on this,dealing with the problem of the uncertainty propagation and analysis in aleatory uncertainty using polynomial chaos expansion method,is that the number of samples increases sharply with increasing variables when solving the polynomial coefficients of the PCE method;also,PCE method cannot directly obtain the failure probability of constraints;the efficiency of uncertainty analysis is low when multi-dimensional uncertain factors coexist.By solving the above problems,an efficient uncertainty analysis method based on the polynomial chaos expansion method is proposed to significantly improve the efficiency and quality of ship uncertainty-based design optimization.3.Uncertainty modeling under the influence of epistemic uncertaintyAfter the uncertainty identification,mixed distribution types of the sparse variables are identified by Akaike Information Criterion method and corresponding weights are given;the interval variables are modeled using evidence theory,which are expressed by focal elements and their corresponding BPA values.4.Research on the unified uncertainty analysis and propagation method under aleatory and epistemic uncertaintiesUnder the influence of aleatory and epistemic uncertainties,a unified quantification and propagation method considering mixed uncertainty is first proposed based on the evidence theory.Although this method is clear and easy to implement,the high computational cost is still a problem that hinders its application in engineering.Therefore,another more efficient method is proposed based on the polynomial chaos expansions method to complete the uncertainty analysis quantification and propagation under the influence of aleatory and epistemic uncertainties.5.Application of uncertainty-based design optimization method in ship design optimizationBased on the above researches,the uncertainty optimization design is applied to seakeeping optimization of typical ship types,and finally a robust and reliable optimal solution is obtained efficiently.