摘要
本文采用浸入边界法模拟了全自由度柔性细丝在涡街中被动变形产生的自适应运动,研究了不同柔性分布方式的细丝在不同涡街中的运动性能,并探讨了单根细丝个体与涡系之间的关系.此外,文章采用长短期记忆(LSTM)神经网络方法建立预测模型.该模型利用细丝上有限测点的训练数据来获得流场的物理参数特征,进而对细丝和扑翼的性能参数的数值变化进行预测,得到了细丝后25%运动周期内的性能参数。该方法可有效简化数值模拟过程.结果表明,预测模型对性能参数预测结果的平均绝对百分比误差稳定在0.09以下,该方法减少了总体实验和计算工作量的15%.进一步的研究表明,扑翼尾迹中的全自由度细丝共产生三类运动模式。仅当整根细丝被包裹在涡带中时,前缘和后缘产生的逆流向压力差可以推动细丝稳定逆来流运动,且细丝刚度的非均匀分布方式和细丝后缘更大的摆幅都有助于其推力的增加.
Inmmersed boundary method is adopted to simulate the adaptive motion generated by the passive deformation of a full degree freedom of flexible filament in the vortex street,the motion performances of filaments with different flexible distribution methods are investigated in different vortex streets,and the relationship between individual filaments and the vortex systems is explored.Moreover,the long short-term memory(LSTM)method is utilized to establish a prediction model.The model uses the training data from the limited measurement points on the filament to obtain the characteristics of the physical parameters of the flows.It predicts the values of the performance parameters of the thin filament and the flapping wing in the future.The performance parameters during the 25%motion period after the filament were obtained by prediction.This method effectively simplifies the numerical simulation process.The results show that the prediction model predicts performance parameters with a stable mean absolute percentage error lower than 0.09.The use of the predictive model reduces the overall experimental and computational effort by 15%.Further investigation indicates three types of motion patterns of the filament in the flapping-wing wakes.When the entire filament is wrapped only in the vortex band,the leading and trailing edges produce a counter-current pressure difference that can propel the filament against steady movement.The non-uniform distribution of filament stiffness and the oscillation at the end of the filament both contribute to the increase of thrust.
作者
孙唯真
何国毅
金飞宇
王琦
于沨
Weizhen Sun;Guoyi He;Feiyu Jin;Qi Wang;Feng Yu(School of Aircraf Engineering,Nanchang Hangkong University,Nanchang 330063,China)
基金
supported by the National Natural Science Foundation of China(Grant Nos.12362026 and 11862017).
关键词
摆幅
运动周期
浸入边界法
参数预测
预测模型
计算工作量
物理参数
涡街
Immersed boundary method
Full degrees of freedom
Flexible filament
Motion performance
LSTM neural network parameter prediction