神经网络算法及其在GPS导航中的应用

GPS导航解算中常采用最小二乘算法。由于导航解算中存在观测异常（粗差），采用的模型精度很难满足用户需求，而高动态用户需求的精度却不断提高，为此文中提出一种利用神经网络的非线性逼近能力，在顾及观测信息权阵的条件下，利用Hopfieid神经网络进行GPS导航解算的新算法。该算法具有很好的抗差性和自适应性，能较好地抑制观测粗差对导航解的影响。计算结果证明了该算法的有效性。     

The automatic control of boundary-layer transition — Experiments and computation

In recent years there has been an increasing interest in the control of boundary-layer transition through the use of wall suction. In the current work suction is provided through one or more suction panels situated close to the leading edge of a plate. Experiments show that boundary-layer pressure fluctuation measurements can be used to identify the position of transition. Transition can be maintained at a desired location with minimum power consumption by employing an automatic adaptive feedback control loop which regulates the suction flow rates of two independent suction panels. This can be expressed as a constrained optimization problem. To allow the suction flow rates to be updated, a modified least mean squares algorithm is used within the control loop. Experimental measurements show that the control algorithm allows fast and stable convergence towards the optimum suction distribution for a double suction panel configuration. Numerical simulations have also been performed. The two-dimensional boundary layer was calculated allowing the viscous boundary layer to interact with the inviscid outer flow. Following linear stability theory the spatial growth rates are calculated by solving an Orr-Sommerfeld type eigenvalue problem, with the streamwise location of transition predicted via thee ^N -method. Applying the same optimization strategy as in the experiments, good qualitative agreement between computations and experiments was found. The optimization algorithm has been applied to computer models where the relation between suction flow rates and transition location is described by an empirical analytical function. This shows that the controller can in principle be applied to systems with more than two suction panels.Nomenclature b transition location with zero suction - d desired transition location - e(k) error signal - k iteration index - p rms pressure - p _ref reference rms pressure - r sum of the reference pressure - u streamwise velocity - u _e external velocity - inviscid external velocity - A wave amplitude - F( ) cost function - I identity matrix - N maximum amplification factor - P projection matrix - R Reynolds number - Re Reynolds number based on the boundary-layer thickness - R matrix of weights - Tu turbulence level - vector of suction flow rates - v normal velocity - v _wall suction velocity at the surface - x streamwise coordinates - x _m microphone location - x _T(k) measured transition location - y normal coordinate - y(k) sum of the measured pressures - w(k) noise - plate length - r +i _i - free stream velocity - * displacement thickness - gradient vector - Lagrange multiplier - controller gain - disturbance stream function - disturbance amplitude - wave frequency = complex wave number     

Reduced‐order modelling of an adaptive mesh ocean model

A novel proper orthogonal decomposition (POD) model has been developed for use with an advanced unstructured mesh finite‐element ocean model, the Imperial College Ocean Model (ICOM, described in detail below), which includes many recent developments in ocean modelling and numerical analysis. The advantages of the POD model developed here over existing POD approaches are the ability:

• 1. To increase accuracy when representing geostrophic balance (the balance between the Coriolis terms and the pressure gradient). This is achieved through the use of two sets of geostrophic basis functions where each one is calculated by basis functions for velocities u and v.
• 2. To speed up the POD simulation. To achieve this a new numerical technique is introduced, whereby a time‐dependent matrix in the discretized equation is rapidly constructed from a series of time‐independent matrices. This development imparts considerable efficiency gains over the often‐used alternative of calculating each finite element over the computational domain at each time level.
• 3. To use dynamically adaptive meshes in the above POD model.

Copyright © 2008 John Wiley & Sons, Ltd.     

h-ADAPTIVITY ANALYSIS BASED ON MULTIPLE SCALE REPRODUCING KERNEL PARTICLE METHOD

IntroductionAs one of the meshfree methods, multiple scale reproducing kernel particle method(RKPM)[1,2]bears, besides the common features of all meshfree methods, a uniqueinherent feature of multiresolution analysis by which the structural response can b…     

高速移动荷载下黏弹性半空间体的动力响应

分别以移动荷载和黏弹性半空间体模拟运动列车荷载和地基,分析了地基在运动列车作用下的动力响应.首先采用Green函数法求解黏弹性半空间体在各种移动荷载模式作用下的动力响应的解析解,包括恒常和简谐移动点源、线源和面源荷载.然后采用IFFT算法和自适应数值积分算法计算解析解中的二维积分,得到了包括低音速、跨音速和超音速移动荷载作用下位移的数值结果.最后分析了速度对位移的分布和最大值的影响,发现当速度大于Rayleigh波速时,位移发生显著变化.     

采用自适应直角网格计算三维增升装置绕流

针对三维增升装置绕流，对存在剪刀叉的不连续外形，基于自适应直角网格，提出并介绍了分区和面搭接技术，采用变长宽比网格，进行了直角网格生成和流场Euler方程数值计算. 根据几何外形的特点，在直角网格生成过程中，以外形不连续面作为分区边界，对初始``根'网格实施分区处理，降低了整个网格的生成难度. 通过基于外形的自适应网格加密，详细描述了剪刀叉外形和缝道，提高了网格质量. 在分区边界面上，基于面搭接技术，构造重叠面积切割算法，实现边界两侧网格间的流场信息传递，保证流场计算中的通量守恒. 采用中心有限体积方法，结合双时间推进算法，完成了两段机翼、带增升襟翼翼身组合体绕流流场的Euler方程数值模拟，对计算结果与实验数据进行了对比，验证了所提方法、算法的合理性和实用性.     

无刷直流力矩电机系统的自适应扰动力矩补偿研究

陀螺漂移测试转台无刷直流力矩电机系统中存在波动力矩和负载力矩振动，这严重地影响了转台速率平稳度。为提高转台速率状态位置跟踪精度，设计了一种自适应补偿方法。该方法包含一个参自适应律和等效PID控制律，它利用前馈补偿原理，来估计电机中未知参数以及波动力矩和负载力矩参数并给与补偿。该自适应补偿方法保证了闭环系统全局稳定性和对期望位置信号的渐进跟踪。仿真结果证明：该方法有效地提高了转台速率状态跟踪精度。     

作动器参数对结构性能的影响和调整受力状态的实施策略模拟

基于规划法研究了作动器各参数对自适应超静定桁架性能的影响。该规划法能够使结构工作状态更合理化。这一方法能够改善自适应结构的工作状态因而提高承载能力。以此为基础，研究了具有作动器的组合杆的有关参数诸如变形上限、刚度和承载力上限对于控制效果的影响。从而为作动器的合理选择提供了理论根据。对于调整作动器变形以控制结构强度的实施策略进行了模拟和分析。该策略包括3部分：1．模拟和分析了“加载、调控、卸载”重复循环直至调控的完成；2．不同作动器按比例同时放大变形直到调控实现；3．多次加载、调控同一次加载、调控结构的等价性。研究表明本文提出的使结构工作状态合理化的调控过程是安全、可靠和灵活的。     

自适应结构多工况下强度控制的研究

对自适应超静定桁架结构的强度控制问题进行了研究，把模型从单一工况推广到了多工况，实现了理论的完整性，定义了结构工作状态系数， 分析了作动器的联入对结构强度的影响，利用超静定桁架的耦合特性和作动器的调节功能，把强度控制问题转化为数学规划问题，方法简单有效。     

INS/ESGM组合导航系统的多模型自适应滤波技术研究

研究了一种随机系统多模型自适应估计的方法，并将此方法应用于INS／ESGM组合导航系统。通过数字仿真将单一模型的INS／ESGM组合导航系统与多模型的INS／ESGM组合导航系统的性能进行了详细比较，说明将多模型自适应估计理论应用于INS／ESGM组合导航系统，可增强系统对环境的适用性，大大提高组合导航系统的精度。