基于随机森林的风洞马赫数预测模型

发布时间：2018-08-21 12:37

【摘要】：在风洞试验中,马赫数的稳定性和快速性对风洞流场品质有着重要影响。为了实现马赫数的精确控制,必须对马赫数进行快速、准确的预测。风洞试验积累了大量数据,大数据集包含了更多的有益信息,为实现马赫数的精确预测提拱了可能性,但也增加了建模的复杂度。通常高度复杂的模型会加重其在实际使用时的计算负担。针对大数据集问题,本文将随机森林方法应用于风洞马赫数建模。随机森林是一种集成模型建模方法,它从3方面降低模型的复杂度:产生多个样本子集,减少了子模型的训练样本个数;具有并行集成结构,子模型可在不同的CPU上运行,提高了运行速度;以简单学习算法回归树作为基学习机,降低了子模型的复杂度。试验证明基于随机森林的马赫数预测模型能够有效利用试验积累的大数据,满足工程上预测速度及精度的要求。
[Abstract]:In wind tunnel test, the stability and rapidity of Mach number have an important effect on the quality of wind tunnel flow field. In order to control Mach numbers accurately, Mach numbers must be predicted quickly and accurately. Wind tunnel experiments have accumulated a large amount of data and the big data set contains more useful information, which raises the possibility of accurate prediction of Mach numbers, but also increases the complexity of modeling. Usually, highly complex models increase the computational burden when they are actually used. For the big data set problem, this paper applies the stochastic forest method to wind tunnel Mach number modeling. Stochastic forest is an integrated model modeling method, which reduces the complexity of the model in three aspects: generating multiple subsets of samples, reducing the number of training samples of the sub-model, having a parallel integration structure, and the sub-model can be run on different CPU. The simple learning algorithm regression tree is used as the basic learning machine to reduce the complexity of the submodel. The experimental results show that the Mach number prediction model based on random forest can effectively utilize the accumulated big data and meet the requirements of engineering prediction speed and precision.
【作者单位】：东北大学信息科学与工程学院;中国空气动力研究与发展中心高速空气动力研究所;
【基金】：国家自然科学基金(61473073,61333006)~~
【分类号】：V211.74

【参考文献】