基于变保真度模型的AUV流体动力参数预测
发布时间：2019-05-10 07:22
【摘要】：变保真度模型(Variable-fidelity modeling,VFM)作为一种有效的代理模型方法被广泛用于工程设计,其中低保真度模型(Low-fidelity modeling,LF)用于获取目标模型的整体趋势,高保真度模型(High-fidelity modeling,HF)用于校正低保真度模型的准确度,二者通过桥函数(Bridge function,BF)融合,完成对目标模型的近似。将VFM用于预测自主水下航行器(Autonomous underwater vehicle,AUV)在外界工况(速度和攻角)下的流体动力参数(阻力系数、升力系数、力矩系数),其中高低保真度模型使用径向基函数构建,桥函数使用克里金函数构建,从遗传算法交叉算子产生的候选样本集中选择变保真度模型与高保真度模型误差最大的样本进行更新,最后建立变保真度代理模型。结果显示VFM能够准确描述流体动力参数在设计空间的变化;在相同的计算条件和时间下,VFM比高保真度模型精度高;随着高保真度样本的增加,模型的精度稳步提升。
[Abstract]:Variable fidelity model (Variable-fidelity modeling,VFM), as an effective proxy model method, is widely used in engineering design, in which the low fidelity model (Low-fidelity modeling,LF) is used to obtain the overall trend of the target model. The high fidelity model (High-fidelity modeling,HF) is used to correct the accuracy of the low fidelity model. The two models are approximated to the target model by bridge function (Bridge function,BF) fusion. VFM is used to predict the hydrodynamic parameters (resistance coefficient, lift coefficient, torque coefficient) of autonomous underwater vehicle (Autonomous underwater vehicle,AUV under external conditions (velocity and angle of attack). The high and low fidelity model is constructed by radial basis function. The bridge function is constructed by Kriging function. The variable fidelity model and the sample with the largest error between the variable fidelity model and the high fidelity model are selected from the candidate sample set generated by the genetic algorithm cross operator to update. Finally, the variable fidelity proxy model is established. The results show that VFM can accurately describe the change of hydrodynamic parameters in design space; under the same calculation conditions and time, the accuracy of VFM is higher than that of high fidelity model; with the increase of high fidelity samples, the accuracy of the model increases steadily.
【作者单位】： 西北工业大学航海学院;
【基金】：国家自然科学基金资助项目(51375389)
【分类号】：O35;U661.1
，
本文编号：2473450
[Abstract]:Variable fidelity model (Variable-fidelity modeling,VFM), as an effective proxy model method, is widely used in engineering design, in which the low fidelity model (Low-fidelity modeling,LF) is used to obtain the overall trend of the target model. The high fidelity model (High-fidelity modeling,HF) is used to correct the accuracy of the low fidelity model. The two models are approximated to the target model by bridge function (Bridge function,BF) fusion. VFM is used to predict the hydrodynamic parameters (resistance coefficient, lift coefficient, torque coefficient) of autonomous underwater vehicle (Autonomous underwater vehicle,AUV under external conditions (velocity and angle of attack). The high and low fidelity model is constructed by radial basis function. The bridge function is constructed by Kriging function. The variable fidelity model and the sample with the largest error between the variable fidelity model and the high fidelity model are selected from the candidate sample set generated by the genetic algorithm cross operator to update. Finally, the variable fidelity proxy model is established. The results show that VFM can accurately describe the change of hydrodynamic parameters in design space; under the same calculation conditions and time, the accuracy of VFM is higher than that of high fidelity model; with the increase of high fidelity samples, the accuracy of the model increases steadily.
【作者单位】： 西北工业大学航海学院;
【基金】：国家自然科学基金资助项目(51375389)
【分类号】：O35;U661.1
，
本文编号：2473450
本文链接：https://www.wllwen.com/kejilunwen/lxlw/2473450.html
