在实现e^(N)方法时,需要搜索流场中的不稳定波,并大量求解当地边界层的稳定性问题,因此为高效求解当地边界层的不稳定波参数,提出了一种基于神经网络的线性稳定性分析方法(neural network-based linear stability analysis,NNLSA)。采...在实现e^(N)方法时,需要搜索流场中的不稳定波,并大量求解当地边界层的稳定性问题,因此为高效求解当地边界层的不稳定波参数,提出了一种基于神经网络的线性稳定性分析方法(neural network-based linear stability analysis,NNLSA)。采用卷积神经网络给出最不稳定波频率ω、展向波数β、流向波数αr和增长率σmax的初值对,再通过迭代法计算失稳扰动波的实际空间失稳波数和增长率。使用平板数据集训练神经网络模型,并利用平板和尖锥算例对NNLSA方法的准确性和计算效率进行验证。结果表明:神经网络部分对不稳定波参数的预测结果与线性稳定性理论的计算结果吻合较好;LSA部分可根据神经网络提供的预测值,通过迭代法找到最不稳定波;NN-LSA方法的求解效率较高,求解时间比全局搜索方法约低20~50倍,大大减小了人为因素在计算过程中的影响。本文提出的NN-LSA方法可以实现自动分析边界层流动的线性稳定性,具有一定的应用潜力。展开更多
The reliability analysis, based on the reliability index method, of two dimensional slopes is generalized by taking Sarma′s acceleration as the performance function. That is to say, a general expression of the perfo...The reliability analysis, based on the reliability index method, of two dimensional slopes is generalized by taking Sarma′s acceleration as the performance function. That is to say, a general expression of the performance function is given under various kinds of slice methods, even under various shapes of slice partition, beyond the traditional vertical slice method. A simple example shows explicitly the relationship of four commonly used slice methods in the slope reliability analysis. It is also found that the results of the reliability analysis are basically consistent with those of the stability analysis based on Sarma′s method.展开更多
文摘在实现e^(N)方法时,需要搜索流场中的不稳定波,并大量求解当地边界层的稳定性问题,因此为高效求解当地边界层的不稳定波参数,提出了一种基于神经网络的线性稳定性分析方法(neural network-based linear stability analysis,NNLSA)。采用卷积神经网络给出最不稳定波频率ω、展向波数β、流向波数αr和增长率σmax的初值对,再通过迭代法计算失稳扰动波的实际空间失稳波数和增长率。使用平板数据集训练神经网络模型,并利用平板和尖锥算例对NNLSA方法的准确性和计算效率进行验证。结果表明:神经网络部分对不稳定波参数的预测结果与线性稳定性理论的计算结果吻合较好;LSA部分可根据神经网络提供的预测值,通过迭代法找到最不稳定波;NN-LSA方法的求解效率较高,求解时间比全局搜索方法约低20~50倍,大大减小了人为因素在计算过程中的影响。本文提出的NN-LSA方法可以实现自动分析边界层流动的线性稳定性,具有一定的应用潜力。
文摘The reliability analysis, based on the reliability index method, of two dimensional slopes is generalized by taking Sarma′s acceleration as the performance function. That is to say, a general expression of the performance function is given under various kinds of slice methods, even under various shapes of slice partition, beyond the traditional vertical slice method. A simple example shows explicitly the relationship of four commonly used slice methods in the slope reliability analysis. It is also found that the results of the reliability analysis are basically consistent with those of the stability analysis based on Sarma′s method.
