Abstract: Sensors are usually used to get the parameters for performance estimate of gas turbines in power plants. The measured values are immediately affected by sensors states. It is hard to detect periodic fault, which is one of common faults of sensors, since the periodic signal has small magnitude and can be easily covered by noise. In this paper, wavelet singular entropy (WSE) is used for periodic fault detection of sensors. Definition and calculation methods of the WSE are introduced. Through the simulation, the advantage of WSE in periodic fault detection is found in comparison with the traditional wavelet transform. It is proved that the WSE is independent of fault amplitude. The method is verified by experimental data from combustor’s outlet temperature sensors.

Key words: vibration and wave, gas turbine power plants, sensor, wavelet singular entropy, wavelet transform, singular value

摘要： 燃气轮机电厂实际运行中，利用传感器得到的数据来进行燃机的状态评估。而传感器的状态直接影响到各个热参数的观测值。常见传感器故障中的周期故障由于故障幅值较小，易被噪声掩盖而无法检测到。应用小波奇异熵(Wavelet Singular Entropy，WSE)构建传感器周期故障检测的方案。介绍小波奇异熵的定义和计算方法，并通过仿真的方法，比较小波奇异熵相对小波变换在周期故障检测方面的优势，并证明小波奇异熵与故障幅值无关。以实验室的小型燃机的燃烧室出口温度传感器为实例进行验证，证明该方法对燃机传感器的周期故障检测有较好的工程应用价值。

关键词: 振动与波, 燃气轮机电厂, 传感器, 小波奇异熵, 小波变换, 奇异值