Non-Gaussian Response Extrema And Fatigue Of Marine Structures

Compared with onshore structures.not only the service environment of marine structures is harsher but also the environmental loads are more complex.In order to ensure the safety and reliability of marine structures during their service life,accurate extreme value prediction and fatigue damage assessment are necessarily needed at the design stage.By contrast to the time-domain analysis method,the frequency-domain analysis method based on power spectrum is faster and can meet the requirements of rapid prediction of many operating conditions in practical engineering.However,in lieu of inevitable non-linear problems in research and design of ocean engineering,as well as the wide-banded characteristics and the non-Gaussian statistical characteristics of structural responses,the developed frequency-domain analysis methods are still more or less inadequate.Thereby,this thesis has carried out the following work based on time-frequency domain analysis.When frequecny-domain or probabilistic methods are adopted to deal with non-Gaussian extrema prediction and/or fatigue damage assessment,establishing an explicit formula for the non-Gaussian stochastic process within the probability framework is inevitable.Therefore,this thesis first discusses the accuracy of formulae of Hermite transformation models for nonlinear systems.A hybrid combination scheme of ordinary central moments(C-moments)and linear moments(L-moments)is proposed to construct Hermite models up to quartic order for handling strongly non-Gaussian cases.A lognormal function is chosen as the original nonlinear system for validating the performance of various Hermite models.Solutions based on analytical moments and sampled moments are both investigated.The comparative studies involve the conventional Gumbel method and the recent averaged conditional exceedance rate(ACER)method.The results show that the proposed hybrid Hermite models render better accuracy and higher robustness in the extrema prediction.A novel spectral method for fatigue damage assessment of bimodal and trimodal Gaussian process is proposed in this paper.This method differs from many conventional probabilistic methods as it is developed based on the spectral discretization scheme.When the response spectrum is discretized into a large number of infinitesimal frequency bands,the fatigue damage with respective to each band can be computed individually.Meanwhile,a coupling coefficient? is introduced to account for the cross-mode coupling between any two frequency bands.It follows that the fatigue damages due to the discretized frequency bands and their couplings are assembled to obtain the total fatigue damage.Comprehensive case studies involving several dominant spectral methods reveal that the proposed cross-mode coupling method is not only physically more reasonable,but also more accurate and robust.For time-domain simulation,a homogeneous reproduction method for strongly non-Gaussian stochastic processes based on quartic and quintic Hermite models is proposed.The relationships of autocorrelation function between the underlying Gaussian process and the target non-Gaussian process defined by these high-order Hermite models is derived.Therefore.the underlying Gaussian power spectrum can be computed from the target non-Gaussian power spectrum directly,while in other methods iteratively updating are required.The non-Gaussian processes simulated by this method has similar statistical characteristics(statistical moments and the same power spectrum)to those of the target process.In case studies,the Morison drag force,total wave load on a jack-up platform and the tower stress of a floating turbine are all simulated using the proposed method.It is shown that the method can effectively generate time series samples of strongly non-Gaussian processes for extrema prediction and fatigue damage assessment.By using this time-domain simulation method,the fatigue damage induced by wide-banded non-Gaussian process is studied.In terms of the non-Gaussian fatigue correction factor,the effects of bandwidth and non-Gaussianity on fatigue damage are studied separately.The drawbacks of every existed spectral method in wide-banded non-Gaussian fatigue damage assessment are pointed out,when the method hiresa Hermite transformation.Practical bimodal spectra in ocean engineering are used to judge the performance of existed spectral methods.The results show that compared with these methods,the scheme of combining the proposed bimodal method in this dissertation with non-Gaussian fatigue correction factor can more effectively and accurately calculate the fatigue damage of moderately and weakly non-Gaussian processes which are bimodal in power spectra.