A major design driver for marine systems is lifetime performance. How a vessel responds in harsh environments has stark consequences for safety and operability, necessitating the consideration of lifetime performance analysis during the design stage. However, extreme events associated with marine dynamic systems may not be caused by the most extreme ocean environments, like the largest wave. Some severe vessel responses may be due to simultaneous combinations of potentially correlated, non-Gaussian loading, which may be excited by any number of wave profiles.Different analytical methods based on extrapolation or solving for threshold exceedances can examine certain aspects of this problem: extreme system responses, combined loading, and long exposures to harsh excitation. But these methods, in general, do not retain the wave profiles which lead to extreme responses. These waves profiles can drive high-fidelity codes, like Computational Fluid Dynamics or Finite Element Analysis time-domain simulations, to give pressure and loading distributions. Such analyses can give an overall account of a system during lifetime events and refine estimates of system performance due to lifetime loading.The Design Loads Generator (DLG) was developed to construct wave profiles that lead to a distribution of linear extreme responses, given an operational profile and exposure period. However, there are some limitations when applying the DLG method to non-linear problems. Some marine systems may experience extreme responses due to varying combinations of non-linear loading. If those loads are strongly correlated, or have an unknown correlation, it is unclear how the capability of the DLG, which considers a single linear load, can be utilized. It may also be desirable to consider lifetime system performance, and not performance conditioned on a specific excitation input, as is estimated by the DLG framework.This dissertation addresses those concerns by expanding the DLG method to what is called the non-linear Design Loads Generator (NL-DLG) process. Given a complex system, operational profile, and exposure, the NL-DLG process uses the DLG capability to determine an ensemble of excitation inputs which lead to lifetime extreme events. Unlike the DLG, which is developed for a single response, the NL-DLG process considers the interaction of multiple stochastic processes which excite the system. These processes, which may be non-Gaussian, are examined so that the resulting ensemble of excitation inputs are demonstrably exhaustive in generating possible defined responses. Short excitation inputs are constructed that estimate the same distribution of responses as would full Monte Carlo Simulations (MCS). Instead of conducting the necessary large number of full-exposure MCS for converged statistics of joint responses, the ensemble of short excitation inputs assembled by the NL-DLG process approximates that same distribution. Various examples are given in this dissertation where comparisons between MCS and NL-DLG extreme value probabilities validate the method.For a complex system with a threshold of allowable responses, the ensemble of NL-DLG generated inputs can estimate an exceedance probability, given the exposure and operational profile. This threshold may be multi-dimensional and a non-linear function of multiple loads. The NL-DLG process examines complex system responses due to combined loading, and maintains links back to the excitation environment, without the computational cost associated with brute-force MCS. These capabilities give deeper insights into system responses, and aid in the design of safer, better operating systems.
【 预 览 】
附件列表
Files
Size
Format
View
Extreme Design Events due to Combined, Non-Gaussian Loading