涡扇发动机气路传感器故障诊断

为了实现对某涡扇发动机传感器故障的在线诊断,提出并设计了1种基于在线贯序极端学习机的故障诊断算法。其核心思想是在定位某传感器故障后,在线建立针对该故障传感器"预学习"的信号重构算法,解决多故障混叠问题。在线信号重构算法以泛化能力指标为判定条件,利用选择策略对算法网络权值进行选择性更新,提高了故障诊断系统的实时性。以某型涡扇发动机为对象开展了传感器故障诊断与重构仿真,结果表明:该算法能够对发动机单、双传感器故障进行准确地诊断与信号重构,且具有良好的实时性。

Fault Diagnosis of Gas Path Sonsor for Turbofan Engine

In order to diagnose malfunctioning turbofan engines'' sensor, a corresponding fault diagnosis system was designed with the Online Sequential Extreme Learning (OS-ELM) algorithm. The core idea is that after finding some malfunction sensor, a predictive learning mechanismis applied to construct fault detection and isolation for the sensor. The fault diagnosis for multiple-sensor failures can be effectively solved by this mechanism. Meanwhile, the output layer weight vector of the algorithm net is updated selectively based on generalization capability, the method could significantly improve the really-time of fault diagnosis system. Simulations on a turbofan engine show that the diagnosis method of sensor faults could detect and isolate faults of single-sensor and double-sensor failures, which also prove the validity and feasibility of the algorithm.