Abstract:  If the sensors of gas turbine fail during operations, and the measurement deviation would seriously affect the accuracy of gas path diagnosis results. In order to solve the above problems, a gas path and sensor coupling fault diagnosis model for gas turbines was proposed. In this model, the measurement variations caused by the changes of working conditions were used instead of the conventional measurements, to reduce the interferences of sensor bias faults. Then, the unscented Kalman filter was adopted to estimate the gas path health parameters by tracking the residual errors between the measured variation and predicted variation caused by different working conditions. Each filter in the sensor diagnosis systems was updated, and the sensor faults were isolated according to the weighted square sum of the residual sequence of the filters. Finally, a simulation example of a typical three-shaft gas turbine was carried out, which shows that the proposed model may reduce the coupling effects between gas path faults and sensor faults, and may realize accurate diagnosis of gas path fault and sensor fault. 

Key words: multi-condition operation, gas path diagnosis system, sensor fault, coupling fault, unscented Kalman filter

摘要： 燃气轮机传感器如在运行期间发生故障，产生的测量偏差会严重影响气路诊断结果的准确性。针对以上问题，提出一种燃气轮机气路与传感器耦合故障诊断模型，该模型采用工况变化引起的测量变化量替代常规的测量量，以减少传感器偏置故障的干扰。在此基础上，应用无迹卡尔曼滤波器跟踪测量变化量与不同工况引起的非线性模型预测变化量之间的残差估计气路健康参数，并以此更新传感器诊断系统中各个滤波器，根据滤波器的残差序列加权平方和进行传感器故障隔离。最后，对某典型三轴式燃气轮机进行仿真实例研究，结果表明该模型可减少气路故障与传感器故障的耦合效应，实现气路与传感器故障的准确诊断。

关键词: 多工况, 气路诊断系统, 传感器故障, 耦合故障, 无迹卡尔曼滤波器