船体初步设计中的船型优化英文文献和中文翻译(3)
The eventual goal of this effort is to be able toimplement a hull form optimization strategywithin the CREATE IHDE. The current plan isfor this process to include a suite of differentfidelity tools to arrive more efficiently at anoptimum solution. The envisioned processwould include using fast, robust potential flowsolution methods to sweep the design space andcreate a response surface of the influence ofgeometry changes on the objective function (e.g.,total resistance). To these results would be addeda series of nonlinear resistance evaluations,which could include predictions made by, forinstance, an RANS tool. These newly addedpredictions would then be used to modify theresponse surface for use in the optimizationprocedure. It is the hope that this provides aprocess by which the designer can make aninformed decision about the planned hull form;and using different fidelity tools provides a fastertime to solution.Validation E¡ortsCurrent optimization efforts have focused on theJoint High Speed Sealift (JHSS) hull form, a verylarge (970 ft) high-speed ship concept operatingat a transit speed of at least 36 knots. This par-ticular concept was chosen because it providesinformation related to a conventional propellerconcept as compared with a waterjet propulsionconcept, and also includes experimental data forfour different bow variants. This provides forvalidation efforts and optimization efforts to be assessed for different propulsion configurationsand for detailed feature shape optimization (e.g.,bow shaping).The work detailed in this paper focuses on thebaseline shafts & struts configuration for theJHSS hull concept.
Denoted DTMBModel5653, it was tested with four different bowshapes, including a stem bow and three differentbow bulb profiles (Cusanelli 2006). A photo ofthe model is given in Figure 1. The top viewshows the entire model configured with thegooseneck bow, and with the rudders and pro-peller shafts & struts included. The lower leftview shows three of the four bow shape variants(gooseneck bulb, baseline bulb, and ellipticalbulb from left to right) and the lower right viewagain shows the gooseneck bulb in a closer view.In order to support the use of varying tool sets aspart of this optimization framework, first vali-dation studies have been performed using bothCFDShip-Iowa and TSD. These were done forthe JHSS shafts & struts Model 5653. Resultsfor the predicted total resistance coefficient aregiven in Figure 2 for the baseline bulb configura-tion. TSD was run in multiple modes to examinethe effects.The TSD results run using the fast mode indicategood agreement over the lower Froude numberrange but then deviate moderately for Froudenumbers greater than about 0.3. But overallthese results seem reasonable for fast, early stagedesign studies that examine gross changes. Inaddition, TSD does not incorporate changes tothe model attitude due to dynamic sinkage andtrimas it was executed in this case. TSD was alsorun using the slower, more accurate method.Here the corrected velocity field is iterated overto improve the accuracy. As shown in the figure,this produces a considerably more accurateresult when compared with the experiments overthe majority of the speed range. This is the modethat will be used to perform optimization studiesshown later. There are some spurious resultsaround Fr50.25 and one point in particular atFr50.318. These are currently being investi-gated. The CFDShip-Iowa results also showgood agreement and provide a further increasein accuracy due to a more realistic representa-tion of the physics, as expected. Here theCFDShip-Iowa predictions are within 6.5%across the speed range, whereas the TSD predic-tions are within about 23% (fast mode) and15% (slow mode, excepting the one spuriouspoint at Fr50.318, excepting the other spuriouspoints this drops to 6.5%). The disparity,however, occurs with regard to the total time tosolution, where CFDShip-Iowa required severalhours as compared with only a few minutes forTSD using the slow mode and seconds for the fast mode for each speed. This is the primarydriver for proposing a multifidelity solutionstrategy when dealing with resistance predic-tions and shape optimization for early stagedesign. In the interim, the current efforts usedonly TSD to evaluate the objective function. Thisprovides for a quick solution time, and the vali-dation exercise indicates sufficient accuracy inTSD to predict the trends as a function of speed.Also, as the optimization process examines thechange in the total resistance, then as long as thetool is used consistently it is believed that it canprovide an improved design, but the magnitudeof the predicted resistance will reflect theuncertainty of the prediction tool.The previous example at model scale providessome confidence in the predictions; however, ingeneral the influences of ship characteristics on s.eship-scale performance are desired in the shipdesign process. Figure 3 shows a comparison ofthe predicted total resistance coefficient at shipscale. Here the 1957 ITTC friction line wasassumed, and the TSD predictions were repeatedfor the appropriate Reynolds numbers tocorrespond to ship scale for the same geometry.By comparing Figure 3 with Figure 2, you cansee the decrease in the total resistance coefficientby moving to the ship scale, as expected. Also,the trends in the TSD predictions, whencompared with the model scale predictions, arevery similar.Hull FormOptimizationIn this section, the results of some of the prelim-inary optimization studies will be presented. Theprocess for performing the optimization includesdefining the baseline geometry, determining theobjective function and design constraints, per-forming the assessment of the objective functionfor all of the basis pairs, and finally determiningthe hull shape that minimizes the objectivefunction. For all of the examples given in thefollowing sections, only the fast mode was usedfor the TSD predictions. This was done to try togauge how effectively the fast, efficient methodcould be used for design optimization problems.Full Ship Optimization forJHSS ConceptDesignThe optimization process was tested using theJHSS conventional baseline shaft & strut hullform concept. The physical model tests includedvariations in the model draft as well, to accountfor changes in the ship displacement (light,design, and heavy). For the purposes of thiseffort, only the design displacement was consid-ered. One case that was examined was if theinitial geometry consisted of the baseline bulbgeometry that was tested. The objective functionwas the total resistance. An example optimiza-tion was performed for a single speed,corresponding to Froude number of 0.29, and inorder to save computational time, the fast modewas used for the TSD evaluations. For this initialevaluation, the optimization was allowed toperturb the entire hull shape. The comparison of55 The design constraints that were used for this ex-ample were that the optimizer allowed no changein the total displacement of the ship and nochange in the longitudinal center of buoyancy(LCB). These were used just as constraints thatmight be typical of a design problem. The numberof basis functions used was 7 in the longitudinal direction and 5 in the transversedirection, yielding a total of 35 degrees offreedom. In order to determine the predictedimprovement in the optimized geometry, the finalsolution from the optimizer is then reevaluatedusing the same solver used to evaluate theobjective function. By comparing the optimizedhull form with the baseline hull form, thetotal resistance was reduced by approximately6.4%. JHSS BowShape Optimization(InitialGeometry5Baseline Bulb)In this case, the optimization procedure waslimited to only focus on the bow section. Theinitial intent of performing this study was tocompare an optimization process for determin-ing the best bow shape to what wasexperimentally observed from the several bowvariants that were tested with physical models.The single objective function optimization wasperformed for three separate speeds (20, 30, and40 knots or 10.3, 15.5, and 20.6m/s) corre-sponding to Froude numbers of 0.193, 0.290,and 0.386. This would provide an optimum forseveral speeds around the design speed of36 knots.The application of the basis pairs to determinethe hull shape perturbations then was limited tothe bow section.