摘要: A three-dimensional numerical study of the turbulent convective heat transfer of the cryogenic methane flowing inside a square engine cooling channel under supercritical pressures was systematically conducted. Numerical results indicate that increasing the fluid pressure results in enhanced heat transfer of the cryogenic methane under supercritical pressures. At the pseudo-critical temperature under a corresponding supercritical pressure, drastic property variations cause heat transfer deterioration and sharp wall temperature increase at a high wall heat flux of 7MW/m2. A modified Jackson and Hall heat transfer equation, which can be used for supercritical heat transfer calculations of the cryogenic methane, has been successfully established in this paper.
摘要: A hybrid optimization algorithm for the time-domain identification of multivariable,state space model for aero-engine was presented in this paper.The optimization procedure runs particle swarm optimization (PSO) and least squares optimization (LSO) "in series".PSO starts from an initial population and searches for the optimum solution by updating generations.However,it can sometimes run into a suboptimal solution.Then LSO can start from the suboptimal solution of PSO,and get an optimum solution by conjugate gradient algorithm.The algorithm is suitable for the high-order multivariable system which has many parameters to be estimated in wide ranges.Hybrid optimization algorithm is applied to estimate the parameters of a 4-input 4-output state variable model (SVM) for aero-engine.The simulation results demonstrate the effectiveness of the proposed algorithm.
摘要: 分析了无人机用电控活塞发动机试验特点以及试验中存在的难点,针对电控发动机高海拔标定试验中进气歧管压力(manifold air pressure,简称MAP)传感器数据的传统线性插值方法不能完全表述电控发动机非线性特性的缺陷,提出采用BP(back propagation)神经网络模型的解决方案.为避免目前应用神经网络方法中所存在的不足,通过采用原始数据分组方法进行网络训练误差的实时反馈和控制,较好地解决了神经网络训练过程中容易陷入"局部最优"和"过拟合"状态,并对BP神经网络预测结果给予了详细研究,训练误差和预测误差分析结果表明了该方法的可行性和计算结果的可信性.