在汽车气动外形优化设计中,往往需要大量的高精度CFD数据作为支撑。然而,高精度CFD数据获取难度大、成本高。为了缓解汽车气动优化设计中气动特性评估精度和效率之间的矛盾,根据迁移学习与数据融合的思想,提出了一种基于多精度深度神经...在汽车气动外形优化设计中,往往需要大量的高精度CFD数据作为支撑。然而,高精度CFD数据获取难度大、成本高。为了缓解汽车气动优化设计中气动特性评估精度和效率之间的矛盾,根据迁移学习与数据融合的思想,提出了一种基于多精度深度神经网络(multi-fidelity deep neural network, MFDNN)的汽车外形优化设计方法,以减少优化设计中所需的高精度数据个数,从而有效提升优化速度、降低优化成本。将所发展的优化方法应用于快背式MIRA标准模型减阻优化设计中,优化结果表明,该方法能够充分融合不同精度数据所蕴含的知识,加速气动外形优化进程,提升优化效率。以收敛用时作为评价指标,在取得相近或更优优化结果的前提下,基于多精度神经网络的优化框架的收敛速度是基于单精度神经网络的离线优化框架的5.85倍,是基于单精度神经网络的在线优化框架的2.81倍。展开更多
The high aerodynamic noise induced by automotive air conditioning systems has important effects on the ride comfort, and the centrifugal fan is the largest noise source in these systems. It is very important to reduce...The high aerodynamic noise induced by automotive air conditioning systems has important effects on the ride comfort, and the centrifugal fan is the largest noise source in these systems. It is very important to reduce the aerodynamic noise generated by the centrifugal fan. The flow field and the sound field on the whole centrifugal fan configuration have been carried out using the computational fluid dynamics. Simulation results show that the sound pressure level near the outlet of the centrifugal fan is too high. Based on the relationship between flow characteristics and the aerodynamic noise, four parameters of the centrifugal fan, i.e., impeller blade＇s outlet angle 0, volute tongue＇s gap t, collector inclination angle fl, and rotating speed n, were selected as design variables and optimized using response surface methodology. While keeping the function of flow rate unchanged, the peak noise level is reduced by 8 dB or 10.8%. The noise level is satisfactorily reduced.展开更多
文摘在汽车气动外形优化设计中,往往需要大量的高精度CFD数据作为支撑。然而,高精度CFD数据获取难度大、成本高。为了缓解汽车气动优化设计中气动特性评估精度和效率之间的矛盾,根据迁移学习与数据融合的思想,提出了一种基于多精度深度神经网络(multi-fidelity deep neural network, MFDNN)的汽车外形优化设计方法,以减少优化设计中所需的高精度数据个数,从而有效提升优化速度、降低优化成本。将所发展的优化方法应用于快背式MIRA标准模型减阻优化设计中,优化结果表明,该方法能够充分融合不同精度数据所蕴含的知识,加速气动外形优化进程,提升优化效率。以收敛用时作为评价指标,在取得相近或更优优化结果的前提下,基于多精度神经网络的优化框架的收敛速度是基于单精度神经网络的离线优化框架的5.85倍,是基于单精度神经网络的在线优化框架的2.81倍。
基金国家自然科学基金资助项目(51305477)重庆市重点产业共性关键技术创新专项(cstc2015zdcy-ztzx60011)+2 种基金The Collaborative Research Project of the Institute of Fluid ScienceTohoku UniversityJapan
基金Project(50975083) supported by the National Natural Science Foundation of ChinaProject(61075001) supported by China State Key Laboratory of Advanced Design and Manufacturing for Vehicle BodyProject(201-IV-068) supported by the Fundamental Research Funds for the Central Universities,China
文摘The high aerodynamic noise induced by automotive air conditioning systems has important effects on the ride comfort, and the centrifugal fan is the largest noise source in these systems. It is very important to reduce the aerodynamic noise generated by the centrifugal fan. The flow field and the sound field on the whole centrifugal fan configuration have been carried out using the computational fluid dynamics. Simulation results show that the sound pressure level near the outlet of the centrifugal fan is too high. Based on the relationship between flow characteristics and the aerodynamic noise, four parameters of the centrifugal fan, i.e., impeller blade＇s outlet angle 0, volute tongue＇s gap t, collector inclination angle fl, and rotating speed n, were selected as design variables and optimized using response surface methodology. While keeping the function of flow rate unchanged, the peak noise level is reduced by 8 dB or 10.8%. The noise level is satisfactorily reduced.
