Abstract Cavitation detection is important in ensuring the reliability of fluid machinery, such as pumps. Vibration signal analysis is widely accepted as an effective tool in condition monitoring and fault diagnosis due to its ability to obtain substantial information and convenience of sensor arrangement. However, cavitation characterization based on vibration measurement is challenging because of the complicated underlying mechanism and low signal-to-noise ratio (SNR) of actual data. This study proposes a carrier wave extraction method for cavitation characterization by combining time synchronous average and time-frequency analysis (TATF) based on amplitude-modulated (AM) signal theory. The proposed method can reasonably measure cavitation severity by distinguishing time-frequency characteristics between different cavitation states. Compared with traditional vibration/acoustic signal monitoring or intelligent diagnostic techniques, cavitation detection based on TATF has the advantages of accurate classification and outstanding physical significance. First, cavitation state division criterion based on energy indicator is proposed. Its superiority is verified via comparison with the traditional criterion of hydraulic head. Second, the vibration signal model of pumps is established as an AM signal model, and the modulation mechanism is elaborated. Extraction of carrier wave components caused by cavitation is regarded as the critical issue in cavitation characterization. Then, TATF is described detailedly and its effectiveness is validated by simulation signals and actual data. Finally, the intelligent classification results of cavitation state by deep convolutional neural network (DCNN) demonstrate the superiority of TATF over short-time Fourier transform (STFT) and wavelet transform (WT).

中文翻译：

---

一种基于时间同步平均和时频分析的空化特征载波提取方法

摘要 汽蚀检测对于保证泵等流体机械的可靠性具有重要意义。振动信号分析因其获取大量信息的能力和传感器布置的便利性而被广泛接受为状态监测和故障诊断的有效工具。然而，由于实际数据的复杂底层机制和低信噪比（SNR），基于振动测量的空化表征具有挑战性。本研究提出了一种基于调幅(AM)信号理论的时间同步平均和时频分析(TATF)相结合的空化特征载波提取方法。该方法可以通过区分不同空化状态之间的时频特性来合理地测量空化严重程度。与传统的振动/声学信号监测或智能诊断技术相比，基于TATF的气穴检测具有分类准确、物理意义突出等优点。首先，提出了基于能量指标的空化状态划分准则。其优越性通过与传统的水头判据的比较得到验证。其次，建立了泵的振动信号模型作为AM信号模型，阐述了调制机制。空化引起的载波分量的提取被认为是空化表征中的关键问题。然后，详细描述了TATF，并通过仿真信号和实际数据验证了其有效性。最后，