Agent仿真中具有先验知识的混合学习算法与混合结构模型

强化学习是一种有效的机器学习方法,是无监督学习,通过不断地和环境交互得到外部环境评价信号,选择合适的动作。Q学习是一种典型的强化学习,其学习效率较低,尤其是当状态空间和决策空间较大时。为提高Q学习学习效率和收敛速度,采用具有先验知识的Q学习算法,利用模糊综合决策方法处理专家经验和环境信息得到Q学习的先验知识,对Q学习的初始状态进行优化;针对Agent个体学习与群体学习各自的不足,提出了采用混合学习算法,将个体学习与群体学习有效结合起来,提高了Agent的个体性能及系统整体的智能水平;同时为满足复杂适应性需求,采用Agent混合结构模型,在该模型中构造了基于知识的协调控制器,通过它来协调慎思式过程和反应式过程。     

Ion dynamics associated with Alfven wave in the near-Earth magnetotail: Two-dimensional global hybrid simulation

Ion dynamics in the near-Earth magnetotail region is examined during periods of fast Earthward flow with a two-dimensional (2-D) global-scale hybrid simulation. The simulation shows that shear Alfven waves are generated at x ∼ −10R_E, where the strong earthward flow is arrested by the dipole field, and propagate along field lines from the equator to both southern and northern polar ionosphere. Non-gyrotropic ion velocity distributions occur where the large-amplitude Alfven waves are dominant. The simulation indicates that the Alfven waves are generated by interaction of the fast earthward flow with the stationary near-Earth plasma. Beam ions are found to be pitch-angle scattered and trapped in the wave field, leading to the non-gyrotropic ion distributions in the high-latitude plasma sheet boundary. In addition, significant particle heating and acceleration are found to occur behind the dipolarization front due to the effect of wave turbulence.     

SOLUTION OF 3-D TURBULENCE NAVIER-STOKES EQUATIONS USING HYBRID GRIDS

INTRODUCTIONWith the development of computer,consid-erable process has been made in the developmentof numerical methods for the solution of Navier-Stokes equations.However,most of these me-thods are impractical for complicated flows in adesign environment.The primary reason is thatwe can′t generate a suitable mesh easily for acomplicated configuration.Although there aremany techniques,such as structured grid or un-structured grid,which have been successfullyused for numerical simulation…     

SYNTHETIC STRESS LIFE TESTING OF MECHATRONIC PRODUCTS UNDER VARIABLE STRESS SPECTRUM

Mechatronic products usually endure the variable stress spectrum when they operate in certain operational condition and environmental condition, which obey the Weibull distribution. In accordance with the features of mechatronic product, this paper analyzes the failure mode, its corresponding sensitive stress and the design principles of life testing profiles. Based on the above analyses, this paper presents a synthetic stress life testing method based on the hybrid Weibull distribution and its statistical method under variable stress spectrum to evaluate the reliability and life indices of mechatronic products. Because the mechatronic products have many characteristics such as high price, long life and small testing samples, the synthetic stress life testing method under variable load spectrum can simulate the real various spectra, decrease the life testing time and reduce the testing samples. So it is effective to carry out the life testing to mechatronic products. The application results of hydraulic pumps indicate that this method can easily handle the experimental data under variable amplitude spectrum, obtain the high precision parameters point estimation and confidence interval estimation and reduce the testing cost greatly.     

仪表着陆系统设备混合监控网络构建

针对传统仪表着陆系统设备"点-点"遥控监控方式易中断,影响运行安全这一问题,提出了一种基于光纤与线缆的"多点-环网"混合监控方案。光纤与线缆互为补充,互为备份,并且当某路监控链路故障时,可自动选择其它监控路径,有效提高仪表着陆系统设备监控效率,增加监控冗余度,为我国民用机场导航监控提供一种新思路。该方案已经成功在重庆江北机场运行。     

粒子群混合算法在变导向器面积计算中的应用

求解非线性方程组经典方法具有严格的局部收敛性;粒子群等进化算法解决了全局收敛的问题,但计算效率偏低,存在最优解不稳定的问题。结合经典Newton-Raphson法的超线性收敛速度和粒子群算法全局收敛能力的粒子群混合算法具备2类算法的优点。在迭代初期采用粒子群算法获得的近似全局解作为Newton-Raphson算法的初始值,以确定高精度的解。利用粒子群混合算法在发动机变导向器面积的大偏离计算中获得了较好的收敛效果,解决了常规Newton-Raphson法不收敛的问题。     

A New Single-blade Based Hybrid CFD Method for Hovering and Forward-flight Rotor Computation

A hybrid Euler/full potential/Lagrangian wake method,based on single-blade simulation,for predicting unsteady aerodynamic flow around helicopter rotors in hover and forward flight has been developed.In this method,an Euler solver is used to model the near wake evolution and transonic flow phenomena in the vicinity of the blade,and a full potential equation(FPE) is used to model the isentropic potential flow region far away from the rotor,while the wake effects of other blades and the far wake are incorporated into the flow solution as an induced inflow distribution using a Lagrangian based wake analysis.To further reduce the execution time,the computational fluid dynamics(CFD) solution and rotor wake analysis(including induced velocity up-date) are conducted parallelly,and a load balancing strategy is employed to account for the information exchange between two solvers.By the developed method,several hover and forward-flight cases on Caradonna-Tung and Helishape 7A rotors are per-formed.Good agreements of the loadings on blade surface with available measured data demonstrate the validation of the method.Also,the CPU time required for different computation runs is compared in the paper,and the results show that the pre-sent hybrid method is superior to conventional CFD method in time cost,and will be more efficient with the number of blades increasing.     

利用飞参数据实现飞机机体结构寿命的动态监控

首先,通过对某型直升机重心法向过载飞参数据的研究得到直升机在实际飞行中的重心过载谱;然后,发展了一种基于飞参数据的名义应力法并对某机群中的每架直升机进行了相对疲劳损伤的计算,计算结果比较符合实际情况;最后,对由相对损伤计算得到的寿命进行了可靠性分析,分析表明机体结构的寿命管理是个动态的过程,利用飞参数据实现从定时维修到...     

燃料电池无人机混合电源动态平衡能量管理策略

飞机电推进的动力系统趋于混合能源形式的发展方向，不同类型的源具有不同的特性，混合能源协调工作的方式可以提高动力系统的性能。本文所研究的飞机电推进系统的能源形式为燃料电池和锂电池所做成的混合能源。针对无人机动力系统工况的特殊性，本文在基于规则的能量管理策略研究基础上，提出了一种基于燃料电池氢气消耗的动态平衡能量管理策略，使燃料电池和辅助电源的能量消耗处于相对平衡的状态，避免了其中一种电源能量先耗尽的情况，可以满足多种工况的变化，提高了混合电源的能量利用率和稳定性，保证了无人机动力系统的可靠性。通过仿真分析结果证明了可行性，最后设计了能量管理系统的硬件并进行了实验验证，通过对实验结果计算分析验证了该能量管理策略的可行性。     

马尔科夫尼可夫规则机理的讨论

不对称烯烃与不对称试剂进行亲电加成反应时,遵守马尔科夫尼可夫规则,本文从生成碳正离子和环状高价正离子的两种反应途径对其反应机理加以讨论,并指出决定两种途径的主要因素.