A Reduced-order Time-Frequency Transforming Method for Non-Stationary Vibration Analysis
پذیرفته شده برای ارائه شفاهی
کد مقاله : 1062-ISAV2024 (R1)
نویسندگان
1Harbin Institute of Technology
2Principal Acoustic Consultant, Hawkins & Associates
چکیده
This paper introduces a novel time-frequency signal analysis method for analysing non-stationary signals called the Reduced-order Time-Frequency Transform (RTFT). Spectral analysis using the Fourier Transform is effective for stationary time series where signal char-acteristics remain constant over time. However, for non-stationary time series such as modu-lated signals, the spectral content varies with time, thus rendering the time-averaged ampli-tude spectrum derived from the Fourier Transform insufficient for tracking changes in signal magnitude, frequency, or phase. The RTFT technique offers the capabilities of traditional time-frequency transformations by employing Pearson's Correlation Coefficient to selectively reduce the data volume in the joint time-frequency domain. This method emphasizes highly correlated frequencies and phases leading to a more efficient data representation without sig-nificant loss of accuracy. The RTFT is validated through comparative analysis with estab-lished methods, including the Short-Time Fourier Transform (STFT), Hilbert-Huang Trans-form (HHT), Fourier Synchrosqueezed Transform (FSST), and Wavelet Synchrosqueezed Transform (WSST). A non-stationary synthesized and real-world vibration-based condition monitoring signal is analysed using both RTFT and the traditional methods to demonstrate the superiority of the RTFT in reducing data volume while maintaining accuracy.
کلیدواژه ها
Title
A Reduced-order Time-Frequency Transforming Method for Non-Stationary Vibration Analysis
Authors
JAVAD ISAVAND, Jihong Yan, Andrew Peplow
Abstract
This paper introduces a novel time-frequency signal analysis method for analyzing non-stationary signals called the Reduced-order Time-Frequency Transform (RTFT). Spectral analysis using the Fourier Transform is effective for stationary time series where signal characteristics remain constant over time. However, for non-stationary time series such as modulated signals, the spectral content varies with time, thus rendering the time-averaged amplitude spectrum derived from the Fourier Transform insufficient for tracking changes in signal magnitude, frequency, or phase. The RTFT technique offers the capabilities of traditional time-frequency transformations by employing Pearson's Correlation Coefficient to selectively reduce the data volume in the joint time-frequency domain. This method emphasizes highly correlated frequencies and phases leading to a more efficient data representation without significant loss of accuracy. The RTFT is validated through comparative analysis with established methods, including the Short-Time Fourier Transform (STFT), Hilbert-Huang Transform (HHT), Fourier Synchrosqueezed Transform (FSST), and Wavelet Synchrosqueezed Transform (WSST). A non-stationary synthesized and real-world vibration-based condition monitoring signal is analyzed using both RTFT and the traditional methods to demonstrate the superiority of the RTFT in reducing data volume while maintaining accuracy.
Keywords
Time-Frequency Analysis, Condition Monitoring, Data Volume Reduction