In this research, two novel online fault detection algorithms are proposed for sensor faults in aircraft. Radial basis function neural network (RBFNN) is used as a fault detection technique, for weight updating parameters adaptive learning rates are used instead of fixed learning rates, two different adaptive learning rate strategies are proposed based on Lyapunov functions which are compared to Extended Kalman Filter (EKF) algorithm. Boeing 747–100/200 aircraft is used for testing and validation of these algorithms. All algorithms have the ability to detect various types of faults such as simultaneous, intermittent, abrupt and incipient with high preciseness and accuracy occur in aircraft sensors. The capability of sensors fault detection of all algorithms are compared, it is proved that all algorithms have the ability to detect faults but Lyapunov function theory II based algorithm is more efficient and having a fast response in faults detection as compared to Lyapunov function theory I and EKF based algorithms. It is also proved that the Lyapunov function theory II based algorithm is more effective in reducing the computational time and computation load.

在这项研究中，针对飞机传感器故障提出了两种新颖的在线故障检测算法。径向基函数神经网络（RBFNN）被用作故障检测技术，为权重更新参数使用自适应学习率代替固定学习率，基于Lyapunov函数提出了两种不同的自适应学习率策略，并将它们与扩展卡尔曼滤波器进行了比较（EKF）算法。波音747–100 / 200飞机用于测试和验证这些算法。所有算法都具有检测各种类型的故障的能力，例如飞机传感器中发生的同时，间歇，突然和初期的故障，具有很高的精度和准确性。比较了所有算法的传感器故障检测能力，事实证明，所有算法都具有检测故障的能力，但是与基于Lyapunov函数理论I和EKF的算法相比，基于Lyapunov函数理论II的算法效率更高，并且在故障检测中具有快速响应。也证明了基于李雅普诺夫函数论Ⅱ的算法在减少计算时间和减少计算量方面更为有效。