摘要
表面热流的快速准确评估对于高超声速飞行器的安全飞行至关重要。由于传统热流评估方法的计算成本与时间开销过大,因此基于数据挖掘的热流快速预测方法因其高效性正在成为新的趋势。然而,当前基于数据驱动的方法只是将飞行外部环境信息和目标点的二维坐标信息作为预测模型的输入,而忽略了影响热流的另一个决定性因素,即飞行器表面的复杂几何信息,从而导致热流预测性能不佳。针对此问题,在已有的飞行外部环境信息的基础上,将三维飞行器外形离散化为点云,再计算与邻域点云间的坐标协方差矩阵,然后通过该协方差矩阵的特征值来构建飞行器表面的几何特征。最后,将原始的特征信息与构造的表面几何特征送入LightGBM模型中。实验结果表明,所提方法的性能优于未做几何结构特征的性能,且与CFD计算结果的相对误差仅为2%左右。
Rapid and accurate assessment of surface heat flow is essential for the safe flight of hypersonic vehicles.Due to the excessive computational cost and time overhead of traditional heat flow assessment methods,the rapid prediction method of heat flow based on data mining is becoming a new trend because of its high efficiency.However,the current data-driven methods only take the flight external environment information and the two-dimensional coordinate information of the target point as the input of the prediction model,while ignoring another decisive factor affecting the heat flow,that is,the complex geometric information of the aircraft surface,which leads to the heat flow poor prediction performance.In order to solve this problem,based on the existing flight external environment information,the shape of the three-dimensional aircraft is discretized into point clouds,and then the coordinate covariance matrix with the neighboring point clouds is calculated,and then the eigenvalues of the covariance matrix are used to construct the geometry of the aircraft surface.Finally,the original feature information and the constructed surface geometry are fed into the LightGBM model.The experimental results show that the performance of the proposed method is better than the performance without geometric structure features,and the relative error with the CFD calculation results is only about 2%.
作者
李林峰
龙吟
曾磊
李强
LI Linfeng;LONG Yin;ZENG Lei;LI Qiang(School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010;Computational Aerodynamics Institute,China Aerodynamics Research and Development Center,Mianyang 621000)
出处
《计算机与数字工程》
2025年第3期747-754,共8页
Computer & Digital Engineering
基金
国家自然科学基金青年科学基金项目(编号:12002361)资助。
