一种新的极值搜索算法及其在航空发动机燃烧主动控制中的应用

 针对航空发动机存在的不稳定燃烧控制问题,推导出当发动机不稳定燃烧时,燃烧室内的振荡压力与平均燃空比值之间的函数模型,此函数模型构建了压力振荡变化的幅值与平均燃空比值之间的极值关系。提出了一种新颖的退火递归神经网络极值搜索算法,可以自适应地搜寻出极值模型中最优的平均燃空比值,实现对燃烧室内压力振荡变化的最小幅值控制,并有效地抑制了燃烧室的不稳定燃烧现象。通过仿真对比,验证了方法的有效性。     

复杂背景下红外弱小运动目标检测的新方法

提出了一种用于检测不同类型复杂背景下红外弱小运动目标的新方法。该方法能够根据图像信息自动选择背景预测算子;同时,针对不同类型复杂背景中目标和背景特性的差异,提出了“局部小目标可辨识度”的概念,并定义了一种有效的方法将其量化。在此基础上,采用蒙特卡罗实验方法构造了一种新的阈值函数,实现了单帧目标的检测,然后采用移动加权管道滤波提取目标的运动轨迹。实验结果表明,该方法对不同类型复杂背景的红外弱小运动目标具有很好的检测性能。     

航空发动机数控系统多故障识别

提出了一种基于模糊逻辑的航空发动机数控系统多故障的识别方法.首先采用Takagi-Sugeno(T-S)模糊模型描述一种涡喷发动机数控系统的非线性模型,然后应用全解耦奇偶方程进行多故障的检测和隔离,并结合卡尔曼滤波器识别出传感器和执行器故障参数.仿真结果表明,针对传感器和执行器同时发生故障的情况,此方法能有效检测故障,并能准确识别出故障参数.      

复杂结构的非线性检测

由于非线性系统的复杂性,现有的各种非线性检测方法都有局限性。从频域、幅域和时域三方面综合考察复杂结构的非线性表现,采用六种方法进行非线性检测,给出一个综合描述系统非线性程度的指标－非线性因子。鉴于非线性系统表现出的非线性强弱与系统所受的激励有直接的关系,将非线性因子定义为与激励的幅值特性和频率特性相关的系数指标。对几种典型非线性系统的仿真模型进行了检测,以验证非线性因子的有效性。     

GRACE gravity field modeling with an investigation on correlation between nuisance parameters and gravity field coefficients

The GRACE (Gravity Recovery And Climate Experiment) monthly gravity models have been independently produced and published by several research institutions, such as Center for Space Research (CSR), GeoForschungsZentrum (GFZ), Jet Propulsion Laboratory (JPL), Centre National d’Etudes Spatiales (CNES) and Delft Institute of Earth Observation and Space Systems (DEOS). According to their processing standards, above institutions use the traditional variational approach except that the DEOS exploits the acceleration approach. The background force models employed are rather similar. The produced gravity field models generally agree with one another in the spatial pattern. However, there are some discrepancies in the gravity signal amplitude between solutions produced by different institutions. In particular, 10%–30% signal amplitude differences in some river basins can be observed. In this paper, we implemented a variant of the traditional variational approach and computed two sets of monthly gravity field solutions using the data from January 2005 to December 2006. The input data are K-band range-rates (KBRR) and kinematic orbits of GRACE satellites. The main difference in the production of our two types of models is how to deal with nuisance parameters. This type of parameters is necessary to absorb low-frequency errors in the data, which are mainly the aliasing and instrument errors. One way is to remove the nuisance parameters before estimating the geopotential coefficients, called NPARB approach in the paper. The other way is to estimate the nuisance parameters and geopotential coefficients simultaneously, called NPESS approach. These two types of solutions mainly differ in geopotential coefficients from degree 2 to 5. This can be explained by the fact that the nuisance parameters and the gravity field coefficients are highly correlated, particularly at low degrees. We compare these solutions with the official and published ones by means of spectral analysis. It is found that our solutions are, in general, consistent with others in the spatial pattern. The water storage variations of the Amazon, Chari and Ganges river basins have also been computed. The variations computed with the NPARB approach are closer to those produced by JPL and DEOS solutions, while the variations produced with the NPESS approach are in good agreement with those produced by the CSR and GFZ solutions. A simulation study is implemented with considering realistic noise and low-frequency error. The two approaches are used to recover the true model. The NPESS solution appears closer to the true one. Therefore we are inclined to estimate the nuisance parameters simultaneously with the geopential coefficients.     

航空永磁发电机的火灾保护

介绍一种应用于航空永磁发电机绕组短路保护的脱开机构，并为此脱开机构设计了测试电路，通过对脱开时间的测试，证明了这种脱开机构是有效的，最后对测试误差进行了分析。     

多层设计参数的叶型气动优化

针对叶片型面的设计问题,提出一种基于遗传优化算法的多层参数化方法.这种方法类似多层网格法原理,利用Bézier曲线的递推算法进行各层之间的设计变量转化,使得优化迭代过程中,下层群体中得以保存上层的优秀个体.根据遗传算法固有的并行特性构建了局域网并行优化平台,并对基本遗传算法进行了改进,从而大大缩短优化时间、提高优化效率.最后设计了曲线逼近和叶型优化的算例,结果显示多层参数化方法能明显加速收敛,在个体数较少时,效果更为明显.     

基于光纤智能夹层和模糊RBF神经网络的飞行器载荷识别

为了更好地保障航空飞行器的安全,提高飞行器的可靠性,提出了一种通过性能参数稳定的光纤智能夹层采集数据,并且结合模糊RBF神经网络对机翼盒段载荷进行识别实验的方法.该方法融合了模糊理论和神经网络各自的优点,通过改进的模糊C均值聚类(FCM)聚类算法删除冗余的规则以进行规则的优化,能自适应地从学习样本数据中提取相应信息,实时地进行载荷辨识.从仿真结果可以看出:该网络模型具有学习时间较短、学习速率较快和精度较高等优点.     

径向圆跳动误差的最小二乘评定数学模型和微机数据处理

用传统的测量方法只能得到被测零件的径向圆跳动误差的近似值。为了得到该项误差的准确值，需要研究新的测量方法。建立了径向圆跳动误差最小二乘评定数学模型，并编制了高级语言数据处理程序。在万能工具显微镜上获得了采样数据，给出了计算结果。     

带有附加扰动的系统族的鲁棒镇定

考虑线性不确定系统的鲁棒镇定问题,系统的不确定性由参数摄动和附加扰动组成。系统的参数摄动采用凸边界系统族的方法描述,给出了该系统可鲁棒δ－镇定的一个充分条件,基于这个结果将不确定系统的鲁棒δ－镇定问题转化为一个优化问题,并且进一步证明了该优化问题是凸的,因此其任何一个局部最优点也是全局最优点。算例说明了该方法的有效性。