基于神经网络的飞控系统故障诊断

飞控系统的故障诊断与维护,可提高其可靠性和可维护性,降低寿命周期费用,有效地降低成本,提高飞机的生存性,减轻驾驶员的负担。首先采用VC与Matlab混合编程技术建立了飞行作动器故障仿真软件,然后通过两种不同的神经网络辨识方法(基于小波包分解的BP网络验证、基于径向基函数的概率神经网络),对卡死和损伤故障的辨识效果进行对比分析,仿真结果显示概率神经网络更适合于大步长采样下的网络辨识,且有较好的泛化能力。所得检测结果符合预期要求,并在所设计软件中得以清晰表达。

Fault-Diagnosis System of Flight Control System Based on Neural Network

Fault diagnosis and maintenance of flight control system could improve the reliability,reduce the fees and cost of life circle,enhance the capability of survival and lessen the burden of aviator.The technology of VC and Matlab is combined to design the simulation software of flight actuator faults.And then,two kinds of neural network methods(back propagation neural network based on wavelet technique,probability neural network based on radio basis function) are used to differentiate the faults of block and impairment,and their results of differential effectiveness are compared and analyzed.It is indicated that probability neural network holds the effectively extensive ability,more suitable for the condition that the fault data is simulated based on large-step sample.The final result fits the expected requirement well and is displayed clearly in the designed software.