基于在线学习神经网络的状态依赖型故障预测

提出外部激励故障和内部激励故障的概念,研究非线性系统状态依赖型故障的预测问题.将非线性系统的故障模型描述成外部激励与内部激励相耦合的一般非线性函数形式,函数的结构未知.通过反向传播(BP)神经网络在线学习故障函数模型实时逼近故障模型,提出基于在线神经网络的状态依赖型故障的预测算法.该算法能够实时地检测故障,对系统状态和故障进行迭代估计和预测.利用系统状态的预测值实时预测了系统的失效时间.故障模型的一般化拓展充分体现了系统状态对故障的影响,增强了算法的实用性.仿真结果验证了该方法的有效性.

Fault prediction for state-dependent fault based on online learning neural network

The concept of external and internal stimulated faults was proposed, and the fault prediction problem of nonlinear system with state dependent fault was investigated. The fault model of nonlinear system was formulated as a general nonlinear function with external and internal stimulated faults coupled together, and the structure of the function was unknown. The back propogation (BP) neural network was applied to learn the fault function online in order to approximate the fault model, and an online neural network based state dependent fault prediction algorithm was given. The algorithm can detect the fault online, and iteratively estimate and predict the fault and the system state. The failure time of the system was predicted online by using the predicted value of the system state. The generalization of fault model showed the effect of the system state on the fault, which made the algorithm more applicable. Simulation results demonstrated the effectiveness of the method.