提出了一种应用神经网络预测电磁干扰的方法.针对遗传算法总体搜索能力较强但容易陷入局部最优,而模拟退火算法具有较强的局部搜索能力,又能避免搜索陷入局部最优解的特点,将模拟退火算法与遗传算法相结合,优化多层前馈(BP, Back Propagation)神经网络,获取最优的权值和阈值,并采用模拟退火的思想确定隐含层神经元的个数,进而建立基于神经网络的电磁干扰预测模型.以双平行导线间的电磁干扰问题为实例,明确干扰要素,建立训练样本和测试样本,对比期望输出和预测输出之间的误差,结果表明该方法可以准确有效地进行电磁干扰预测. 相似文献
Based on modified Leishman-Beddoes (L-B) state space model at low Mach number (lower than 0.3), the airfoil aeroelastic system is presented in this paper. The main modifications for L-B model include a new dynamic stall criterion and revisions of normal force and pitching moment coefficient. The bifurcation diagrams, the limit cycle oscillation (LCO) phase plane plots and the time domain response figures are applied to investigating the stall flutter bifurcation behavior of airfoil aeroelastic systems with symmetry or asymmetry. It is shown that the symmetric periodical oscillation happens after subcritical bifurcation caused by dynamic stall, and the asymmetric periodical oscillation, which is caused by the interaction of dynamic stall and static divergence, only happens in the airfoil aeroelastic system with asymmetry. Validations of the modified L-B model and the airfoil aeroelastic system are presented with the experimental airload data of NACA0012 and OA207 and experimental stall flutter data of NACA0012 respectively. Results demonstrate that the airfoil aeroelastic system presented in this paper is effective and accurate, which can be applied to the investigation of airfoil stall flutter at low Mach number. 相似文献