As this processis expanded to include evaluations using RANS,automation will again be key, especially in rela-tion to the grid generation process. This work iscurrently ongoing.With proper automation, itbecomes possible to provide parametric infor-mation about changes to the global definition ofthe hull form that would help to guide much ofthe early stage design comparison studies andin the analysis of alternatives design stage. Theoptimization process could, for example, followa set-based design approach by providingresistance information for a series of length andbeam changes, or side hull clearance and staggerin the case of multihulls, which would stillsatisfy the overall design synthesis process.With the ongoing development of this technol-ogy, it is our hope and intent that the use ofhydrodynamic evaluation and optimizationtools within the CREATE IHDE design environ-ment will aid current and future ship designers.The capability from this effort has the potentialto significantly impact directly the issues that areof concern for current and future acquisitionprograms for US Navy ships.ComputationalToolsThis section follows some discussion of thecomputational tools being examined as part ofthis effort. In particular, these tools are eithercurrently implemented, or are slated forinclusion in the IHDE.Total ship drag (TSD) is a robust fast resistanceprediction tool appropriate for early stage designdeveloped by NSWCCD (Metcalf et al. 2004).The total drag of a ship as calculated by TSD ismade up of the following components: wave-making resistance, frictional resistance, formresistance, transom drag, and other drag. Eachresistance component is estimated in a way thatis faithful to the physics of the problem. Thewave-making resistance is computed usingslender ship theory (Noblesse 1983). The fric-tional resistance is estimated using the ITTCfriction line. Form resistance is approximatedfromseries 58 data. Transomdrag is pided intotwo components—a base drag component that ismodeled based on empirical data from subsonicbullet tests, and a hydrostatic component thataccounts for the missing hydrostatic pressure ona dry transom. Finally, an additional componentof drag is modeled that accounts for other dragsources such as spray. This component is empir-ically based on series 64 data and other formswith spray formation. All these components ofdrag respond to changes in the hull form andmake TSD a tool that is appropriate for use withan optimization code.TSD was used in two different modes for thisstudy. These are determined by a user-specifiedparameter (kext), which sets the relative impor-tance of speed versus accuracy. In the fast mode(kext5 1), it computes Noblesse’s zeroth-order slender-ship approximation to the far fieldwave resistance where the source strengthapplied on a panel depends only on thex-component (flow direction) of the normal tothe panel. In the slow mode (kext50), thezeroth-order flow is computed at each panel onthe hull. A local correction to the normal flowthrough the panel is then applied to the sourcestrength at each panel before computing thewave resistance. This correction can be appliediteratively, but it is much more sensitive to pan-elization and is not guaranteed to converge.CFDShip-Iowa is a general-purpose research,unsteady Reynolds-averaged Navier–Stokes(URANS) CFD code developed at University ofIowa (UI) over the past 10 years for support ofstudent theses and research projects at UI, aswell as transition to Navy laboratories, industry,and other universities. CFDShip-Iowa solvesRANS equations using curvilinear overset grids.A combination of finite difference and finitevolume methods is used to solve the equations.Second-, third-, and fourth-order upwind-biaseddiscretizations can be selected for the convectionterms, and the second-order central method is used for diffusion terms. The pressure–velocitycoupling is achieved using either a projectionalgorithm (faster) or a pressure implicit withsplitting of operators method that is slower,though more robust. The resulting pressurematrix is solved with preconditioned Krylov-space type solvers using the PETSc package fromArgonne National Laboratory. Boundary condi-tions are set using the graphical user interface inthe GRIDGEN software from Pointwise Inc.Implemented RANS turbulence closures includeone-equation, two-equation, and an anisotropicexplicit algebraic Reynolds stress model.A surface-capturing method using the level-setapproach is used to model the free surface. Inthis method, a distance function is transportedwith the flow both in air and water, the interfacebeing defined by the zero-contour (level set)of this function. This approach allows for thecalculation of motions with large amplitudes,breaking waves, and splashing.One potential optimization framework that iscurrently being investigated is the SHAPE code,developed by SAIC (Kuhn et al. 2007). TheSHAPE code determines changes to a baselinehull shape that produce improvements to someuser-defined metric and are bounded by a set oflocal and generic constraints that are also pre-scribed by the user. The optimized hull shape isdetermined by examining how perturbations tothe baseline hull shape change the evaluation ofthe objective function. The optimization routineis completely separate from the objective func-tion evaluations. In this way, it is possible toutilize a variety of different analysis tools, in-cluding more computationally intensive tools, toperform the evaluations and build up a databasethat reflects the derivatives of the objectivefunction for each of the perturbed hull shapes.The optimization routine itself, which uses lin-ear programming, can then be done very quicklyusing the pregenerated database of derivatives.This also has the advantage of allowing the userto perform a variety of different design studies ina very short time; for example, changing thedesign constraints and assessing a new optimumdesign based on those constraints. Examples ofthis approach will be demonstrated for localizedbow dome shape optimization studies.One of the key elements necessary for integra-tion with a design environment is automation.To that end, a semi-automated process has beenimplemented for determining the hull shape per-turbations and evaluating the objective functionfor each perturbed shape using TSD. As thisprocess is expanded to include evaluations usingRANS, automation will again be key, especiallyin relation to the grid generation process. Thiswork is currently ongoing.Planned Implementation 船体初步设计中的船型优化英文文献和中文翻译(2):http://www.youerw.com/fanyi/lunwen_63696.html