基于RBFNN自适应混合学习算法的航天测控系统任务可靠性分配

为解决执行航天测控任务的各设备存在复杂的时空关联、可视与信息关联等动态约束关系,使得航天测控系统任务可靠性分配建模和分析极其困难,同时模型求解效率低的问题,提出了自适应混合学习算法的径向基神经网络建模方法.算法通过训练样本相关性矩阵的主成分分析确定网络隐含层初始节点数;在此基础上,利用梯度信息衰减因子改进了迭代过程中网络参数的梯度信息计算方式,避免了学习过程早熟的不足,且加快了迭代收敛速度.最后,通过采集航天测控系统输入-输出数据,将自适应混合学习算法应用于参数训练,并给出了具体实现步骤.通过算例仿真,表明算法在解决航天测控系统任务可靠性分配问题时具有较高泛化能力和分配结果稳定等优点. 

Mission reliability allocation of spaceflight TT&C system based on RBFNN adaptive hybrid learning algorithm

Due to the existence of a complex spatial and temporal correlation,visualization and information related dynamic constraints to each equipment when they executed TT&C(telemetry tracking & commanding) mission,it was extremely difficult to model and analyze,and the solving efficiency was low.Therefore a radial basis function neural network (RBFNN) modeling method with adaptive hybrid learning(AHL) algorithm was proposed.Principal component analysis were used to determine the initial number of hidden units.Intelligence optimization algorithm combined with decaying gradient descent information was used to train parameters of RBFNN which were improved to accelerate convergence and guarantee the result to converge at the global optimum with high probability.The algorithm was employed in the offline training of model parameters by sampling the input/output data of the system,and the realization details were also provided.By simulation,it shows that the AHL has higher generalization power and can reach more optimal results in solving the mission reliability allocation of spaceflight TT&C system.