基于自适应多尺度线性调频小波分解的水轮机非平稳振动信号分析

针对在连续参数空间内优化求解线性调频小波(Chirplet)时频参数存在计算量大的问题，构造了离散的多尺度Chirplet字典，以适应信号复杂多变的局部结构，降低计算量。针对自适应Chirplet谱图中的幅值畸变缺陷，应用归一化的最优Chirplet原子的Wigner-Ville分布识别了信号成分，将其作为时频窗口对信号的Wigner-Ville分布进行时频滤波处理，得到的改进的自适应Chirplet谱图不仅具有最佳的时频分辨率，不含交叉干扰项，而且能量分布的幅值变化平滑连续，不存在幅值畸变缺陷。根据时频分析结果，识别了机组起动过程中主轴摆度信号的频率成分及其幅值的时变情况，验证了自适应多尺度Gauss Chirplet分解以及改进的自适应Gauss Chirplet谱图在分析旋转机械转速变化过程中转子非平稳振动信号方面的有效性。

Time-frequency Analysis of Hydroturbine Nonstationary Vibration Signal Based on Adaptive Multi-scale Chirplet Decomposition

A multi-scale Gaussian chirplet dictionary was constructed to match the local structure of complicated signals, and to avoid the large computational complexity inherent with the optimization of chirplet parameters in an infinite space. Wigner-Ville distribution of normalized optimal chirplets is effective in identifying signal components. In order to avoid the amplitude distortion in adaptive chirplet spectrogram, it was employed as a time frequency window to filter the time frequency distribution, thus resulting in an improved adaptive chirplet spectrogram. It possesses the best time frequency resolution, and immunity of cross term interferences and amplitude distortions. By means of time frequency analysis, the frequency components and their time evolutions of a hydroturbine main shaft vibration signal during a start-up process are identified, demonstrating the effectiveness of adaptive chirplet decomposition in analyzing the nonstationary rotor vibration during speed varying processes.