Turbine pump health state detection method based on fusion trend sensitivity index

Abstract

The invention discloses a turbine pump health state detection method based on a fusion trend sensitivity index, belongs to the technical field of turbine pump vibration detection, solves the technical problem of inaccurate health state characterization, and comprises the following steps: obtaining an acceleration signal and a displacement signal of a turbine pump in a hot test run process of an engine; processing the acceleration signal, and calculating to obtain an acceleration signal feature; adopting a two-dimensional holographic spectrum algorithm to extract displacement signal features; and the acceleration signal features and the displacement signal features are fused, a fusion trend sensitivity index is obtained, and the health state of the turbine pump of the liquid rocket engine is obtained. The method is used for judging the health state of the turbine pump of the engine and managing the operation of the engine.

Description

Technical Field

[0001] This invention belongs to the field of turbine pump vibration detection technology, specifically relating to a method for detecting the health status of turbine pumps based on a fusion trend sensitivity index. Background Technology

[0002] The turbopump is the "heart" of a liquid rocket engine. Its main function is to pressurize the cryogenic liquid propellant (fuel and oxidizer) in the propellant tank and deliver it to the gas generator and thrust chamber. Its health is crucial to the reliable operation of the liquid rocket engine. Because turbopumps operate in extremely harsh and complex environments with high pressure, high speed, large temperature differences, and strong vibrations, they are prone to failures such as turbopump blade breakage, blade crown detachment, and bearing damage during rocket engine start-up, shutdown, and operation. Statistics show that over 60% of global space launch mission failures are caused by rocket engine malfunctions (reaching 80% in China), and turbopump failures account for approximately 50% of all rocket engine system failures. Therefore, the safe and reliable operation of the engine turbopump system is a crucial foundation for the successful execution of rocket launch missions. Developing turbopump and liquid rocket engine operational safety assurance technologies is of great significance for ensuring the smooth implementation of the space power strategy and safeguarding core national security interests.

[0003] Traditional hot-fire test data analysis often employs peak-to-peak value (PVP) and RMS analysis techniques. By comparing historical hot-fire test data to a database, data within the historical envelope is considered normal. However, these traditional PVP and RMS analyses are insufficient for representing turbopump health status information. Furthermore, the analysis methods for acceleration and displacement signals are independent, and there is no unified quantitative evaluation standard for their results. The analysis methods ignore the correlation between the two, easily leading to the loss of crucial turbopump health status information. The complex and inconsistent analysis results can cause uncertainty in turbopump operation and management decisions. When engine health deteriorates or even early failures occur, PVP and RMS analyses fail to detect abnormalities in key components, hindering health status identification and potentially causing serious engine explosions that threaten the lives of on-site technicians and national property. With the increasing intensity of space launch missions, the workload of test data analysis is growing. The independent analysis and diagnosis of information carriers like acceleration and displacement signals, which jointly represent turbopump health status from different perspectives, leads to a waste of human resources and struggles to meet the ever-increasing demands of test data analysis. Therefore, it is urgent to conduct research on the fusion analysis method of vibration acceleration and displacement signals of key components such as liquid rocket engine turbopumps, so as to achieve the organic combination of engine turbopump acceleration and displacement data analysis under interpretable premise, thereby improving the effectiveness of engine turbopump speed variation data mining, providing a more accurate theoretical basis for judging the health status of engine turbopumps, and meeting the needs of the increasing number and scale of ground hot test data analysis tasks. Summary of the Invention

[0004] To overcome the shortcomings of traditional peak-to-peak value and RMS analysis as global quantitative indicators, which cannot accurately depict the evolution of the health status of each component of the turbopump during long-term hot commissioning and are difficult to characterize the health status of key components of the turbopump, this invention proposes a turbopump health status detection method based on a fusion trend sensitivity index.

[0005] The technical solution adopted by this invention to solve its technical problem is: A method for detecting the health status of a turbopump based on a fusion trend sensitivity index includes the following steps: Step 1: Acquire the turbopump acceleration and displacement signals. An acceleration sensor is installed between the oxidizer pump and the turbine of the engine turbopump. The acceleration sensor is used to measure the vibration signal in the three directions of shaft diameter tangentiality. Two displacement sensors are installed on the shaft section between the fuel pump and the oxidizer pump, arranged in mutually perpendicular directions.

[0006] During engine hot-run testing, acceleration and displacement signals of the engine turbopump were collected.

[0007] Step 2, Acceleration signal processing A high-precision time-frequency domain analysis method is used to process the acceleration signal to obtain the time spectrum; using the time spectrum, the variation trend characteristics of the frequency band energy are calculated, i.e., the acceleration signal characteristics.

[0008] Step 3, Displacement Signal Processing A two-dimensional holographic spectrum algorithm is used to calculate the holographic spectrum (Holospectrum) of the displacement signal. The variation trend characteristics of the holographic spectrum frequency ellipse, i.e., the displacement signal characteristics, are extracted from the Holospectrum.

[0009] Step 4, Signal Feature Fusion By fusing acceleration signal features and displacement signal features, a fusion trend sensitivity index is obtained to determine the health status of the liquid rocket engine turbopump.

[0010] In the above-described method for detecting the health status of a turbopump, the acceleration sensor is a vibration acceleration sensor.

[0011] The aforementioned method for detecting the health status of a turbopump refers to a liquid rocket engine turbopump.

[0012] The above-mentioned method for detecting the health status of a turbopump, wherein step 2 further includes: The short-time Fourier transform (STFT) of the time-domain signal x(t) yields the following time-frequency spectrum: (1); In equation (1), F x ω represents the spectrum of the original time-domain signal x(t) after short-time Fourier transform, which contains the amplitude and phase information of the signal in time and frequency; g is a tightly supported window function that can move with time t, and different window functions have different spectral characteristics; ω is the frequency of the harmonic signal.

[0013] The synchronous compressed wavelet transform (WSST) algorithm is employed. WSST uses a compression operator to compress the dispersed time-frequency coefficients in the STFT spectrum to the instantaneous frequency, thus obtaining a new time-frequency spectrum. The transformation formula is as follows: (2); (3); In equations (2) and (3), The time spectrum is obtained by compressing the wavelet coefficient energy along the instantaneous frequency direction; F x The spectrum of the original time-domain signal x(t) after short-time Fourier transform; The Dirac function is used to concentrate energy at an instantaneous frequency. superior; The frequency is instantaneous, and the spectrum is more accurate at this frequency.

[0014] For each component, the WSST calculation is performed on the acceleration signal to obtain the time spectrum. Using the time spectrum and based on the characteristics of turbopump failure, key frequency components are selected. A fixed bandwidth is set around these key frequency components, and the trend characteristics of frequency band energy variation within this bandwidth are extracted. The frequency band energy calculation process is as follows: For an acceleration signal obtained from a single measurement, the signal is first preprocessed, including data splicing, time period truncation, and mean removal. Secondly, a Fourier transform is performed on the preprocessed signal to obtain the frequency domain signal. Thirdly, the frequency component of interest is selected, denoted as f, with a bandwidth of [value missing]. In the frequency range There are n spectral lines in the memory, and the amplitude of the i-th spectral line is denoted as a. i The formula for calculating the band energy at that frequency is as follows: (4); In equation (4), E nf Frequency range Bandwidth energy, E nf The larger the value, the more concentrated the signal energy is within that frequency band; f is the frequency of interest in time-frequency analysis; n is the frequency range. The number of spectral lines in the spectrum within; a i Let be the amplitude of the i-th spectral line.

[0015] Finally, the same method was used to calculate the frequency band energy for different time periods, and the trend of frequency band energy change over time was obtained.

[0016] The above-mentioned method for detecting the health status of a turbopump, wherein step 3 further includes: Based on two mutually perpendicular displacement signals, the amplitude spectrum and phase spectrum of the displacement signals are calculated respectively; a two-dimensional holographic spectrum algorithm is used to merge the amplitude spectrum and phase spectrum of the two displacement signals to obtain the holographic spectrum of the displacement vibration signal and draw an ellipse; the variation trend characteristics of the holographic spectrum frequency-transition ellipse are extracted from the holographic spectrum.

[0017] First, simultaneous collection , Displacement signals in two mutually perpendicular directions, the two mutually perpendicular directions are , ,right , The displacement signals in two directions are preprocessed to obtain a finite discrete data sequence x. t y t ; Secondly, for a finite discrete data sequence x t yt Performing a finite discrete Fourier transform yields the corresponding spectral sequence X. k Y m ; Next, the spectral sequence is expressed as the sum of the real and imaginary parts, the corresponding amplitude and phase are listed, the synthesis equation of the i-th spectral line is calculated, and the trajectory diagram of the synthesis equation is arranged in order of the magnitude of the spectral line frequency to form a two-dimensional holographic spectrum.

[0018] Finally, feature extraction is performed on the two-dimensional holographic spectrum to obtain the variation trends of the major and minor semi-axis, and the variation trend graph is plotted, which is the displacement signal feature.

[0019] The above-mentioned method for detecting the health status of a turbopump, wherein step 3 further includes: The equation for the synthesis of the i-th spectral line is calculated as follows: (5); In equation (5), A i Let B be the amplitude of the i-th spectral line in the x-direction. i Let be the amplitude of the i-th spectral line in the y-direction. Let be the phase of the i-th spectral line in the x-direction; The phase of the i-th spectral line in the y-direction.

[0020] For the i-th spectral line at the same frequency of two displacement signals, the semi-major axis of the ellipse... short half shaft The calculation equation is as follows: (6); In equation (6), Let be the length of the major semi-axis of the holographic ellipse of the i-th spectral line; A is the length of the minor semi-axis of the holographic spectrum ellipse; i Let B be the amplitude of the i-th spectral line in the x-direction. i Let be the amplitude of the i-th spectral line in the y-direction; Let be the phase of the i-th spectral line in the x-direction; The phase of the i-th spectral line in the y-direction.

[0021] Similarly, by selecting different times to calculate the spectral lines of the two displacement signals, the values ​​of the short and long semi-axis at different time lengths are obtained, and thus the trend of the change of the short and long semi-axis of the spectral lines with time is obtained.

[0022] The above-mentioned method for detecting the health status of a turbopump, step 4, further includes: The frequency band energy amplitudes of each frequency component extracted from the acceleration signal are normalized; the average of the major and minor semi-axes of the frequency transfer ellipse extracted from the holographic spectrum is used as the weight to achieve uniformity of the magnitudes of different components.

[0023] The fusion calculation formula is as follows: (7); In the formula, , The result is a fusion of acceleration and displacement signals, which includes information such as the frequency band amplitude and energy of the acceleration signal, and the frequency transition ellipse of the holographic spectrum of the displacement signal. , These are the amplitudes of the major and minor semi-axes of the frequency-transition ellipse extracted from the holographic spectrum. , They are respectively , The mean; For numerical normalization, ; Frequency components extracted from acceleration signals The frequency band energy amplitude centered on the frequency band.

[0024] Based on fusion results , Construct the FDS index and define FDS as the fusion result. , The weighted sum of the local mean and local standard deviation is shown in the following formula.

[0025] Based on fusion results , Construct a fusion trend sensitivity index, FDS, and define FDS as the fusion result. , The weighted sum of the local mean and local standard deviation is shown in the following formula.

[0026] (8); In equation (8), FDS is the proposed fusion trend sensitivity index, which describes the health status of each component of the turbopump. , The result of fusing acceleration and displacement signals. The standard deviation is denoted as .

[0027] The beneficial effects of this invention are: Traditional peak-to-peak value and RMS analyses, as global quantitative indicators, cannot accurately depict the evolution of the health status of various turbopump components during long-term hot-running tests, nor can they characterize the health status of key turbopump components. Furthermore, previous methods for analyzing acceleration and displacement signals were independent, lacking a unified quantitative standard. These methods often ignored the correlation between the two signals, leading to the loss of crucial information about turbopump health status. The resulting complex and inconsistent analysis can also cause uncertainty in turbopump operation and management decisions. This paper proposes a turbopump health status detection method based on a fusion trend sensitivity index. The proposed FDS index organically combines the analysis of engine turbopump acceleration and displacement data, thereby improving the effectiveness of turbopump speed variation data mining and providing a more reliable and accurate theoretical basis and technical support for turbopump health status assessment and engine operation and management decisions. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the measuring point positions of the accelerometer and displacement sensor in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the turbine pump health status detection process according to Embodiment 1 of the present invention; Figure 3 This is a trend diagram of frequency band energy change over time obtained by feature extraction of the time spectrum according to an embodiment of the present invention; Figure 4 This is a graph showing the trend of the semi-major and semi-minor axes of an ellipse over time, obtained by feature extraction from a holographic spectrum according to an embodiment of the present invention. Figure 4 (a) shows the trend of the amplitude of the semi-major axis changing over time. Figure 4 (b) shows the trend of the amplitude of the minor semi-axis over time; Figure 5 This is an embodiment of the present invention, which uses a fusion analysis of the vibration signal of an oxidizer pump during a certain test run to obtain FDS curves of different harmonic frequencies over time; among them, the nine graphs, from a to i, are the FDS curves of the main turbine speed at the 1st, 3rd, 6th, 9th, 12th, 18th, 21st, 24th and 30th harmonic frequencies as the trend of FDS curves over time.

[0029] The attached figures are labeled as follows: 1. Fuel pump, 2. Displacement sensor, 3. Displacement sensor mounting section, 4. Oxidant pump, 5. Accelerometer, 6. Turbine. Detailed Implementation

[0030] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0031] Example 1 A method for detecting the health status of a turbopump based on a fusion trend sensitivity index, the detection process is as follows: Figure 2As shown, by organically combining the analysis of engine turbopump acceleration and displacement data, the effectiveness of turbopump speed variation data mining is improved, providing a more reliable and accurate theoretical basis and technical support for turbopump health status assessment and engine pre-delivery status evaluation. Specifically, the steps include the following: Step 1: Acquire turbopump acceleration and displacement signals during the hot-fire test of the liquid rocket engine.

[0032] As an optional method, acquiring turbopump acceleration and displacement signals during the hot-fire test of the liquid rocket engine includes: Vibration acceleration sensors and displacement sensors are selected and installed at key locations on various components of the engine turbopump. For example, for the oxidizer pump, the installation locations of the acceleration and displacement sensor measuring points are as follows: Figure 1 The accelerometer is installed near the bearing section of the oxidizer pump to measure the vibration signals in the three directions of shaft diameter and tangential direction; the displacement sensor is installed on the shaft section between the oxidizer pump and the fuel pump, with two sensors arranged in mutually perpendicular directions.

[0033] The vibration acceleration sensor and displacement sensor were used to measure and record acceleration and displacement signals throughout the engine hot test.

[0034] A hot-fire test system was constructed, and signal acquisition equipment was used to record the signals from the acceleration and displacement sensors. The signal acquisition equipment was turned on before the test began, and vibration acceleration and displacement signals were continuously acquired and stored in real time during the test until the hot-fire test ended, thereby obtaining the vibration acceleration and displacement signals of the engine during the hot-fire test.

[0035] Step 2: The acceleration signal is processed using a high-precision time-frequency domain analysis method to obtain the time spectrum, and then the trend characteristics of frequency band energy variation are calculated from the time spectrum.

[0036] As an optional approach, the acceleration signal is processed using a high-precision time-frequency domain analysis method to obtain a time-frequency spectrum, and then the frequency band energy variation trend characteristics are calculated from the time-frequency spectrum, including: For example, the high-precision time-frequency domain analysis method employs the Synchronous Compressed Wavelet Transform (WSST) algorithm. The WSST algorithm rearranges the linear time-frequency transform result based on the original linear time-frequency spectrum, compressing time-frequency coefficients with the same instantaneous frequency to their corresponding instantaneous frequency positions, thus obtaining a more accurate and clearer time-frequency spectrum. For example, firstly, a Short-Time Fourier Transform (STFT) is performed on the time-domain signal x(t), yielding its time-frequency spectrum as follows: (1); In equation (1), F xω represents the spectrum of the original time-domain signal x(t) after short-time Fourier transform, which contains the amplitude and phase information of the signal in time and frequency; g is a tightly supported window function that can be shifted with time t, and different window functions have different spectral characteristics; ω is the frequency of the harmonic signal. In the time spectrum after STFT, the peak of the time-frequency coefficients occurs at the harmonic signal frequency ω0, while the time-frequency coefficients distributed on both sides of ω0 decrease as the distance from ω0 increases. Although the time-frequency coefficients of the signal are distributed near the instantaneous frequency after STFT transformation, energy leakage in the window function causes ambiguity in the time spectrum, reducing its resolution. WSST employs a compression operator to compress the dispersed time-frequency coefficients in the STFT spectrum to the instantaneous frequency, obtaining a new time spectrum. The transformation formula is as follows: (2); (3); In equations (2) and (3), The time spectrum is obtained by compressing the wavelet coefficient energy along the instantaneous frequency direction; F x The spectrum of the original time-domain signal x(t) after short-time Fourier transform; The Dirac function is used to concentrate energy at an instantaneous frequency. superior; The frequency is instantaneous, and the spectrum is more accurate at this frequency.

[0037] For each component, after obtaining the time-frequency spectrum by performing WSST calculation on the acceleration signal, key frequency components are selected based on the characteristics of turbopump faults. A fixed bandwidth is set around these key frequency components, and the frequency band energy variation trend within this bandwidth is extracted. Specifically, the frequency band energy calculation method is as follows: For an acceleration signal obtained from a single measurement, firstly, the signal is preprocessed, including data splicing, time period truncation, and mean removal. Secondly, a Fourier transform is performed on the preprocessed signal to obtain the frequency domain signal. Thirdly, the frequency component of interest is selected, denoted as f, with a bandwidth of [value missing]. In the frequency range There are n spectral lines in the memory, and the amplitude of the i-th spectral line is denoted as a. i The formula for calculating the band energy at that frequency is as follows: (4); Finally, by calculating the frequency band energy for each time period using the same method, the trend of frequency band energy variation over time can be obtained. For example, for the oxygen pre-compression pump acceleration signal during an engine hot-run test, the trend of frequency band energy variation over time at the first octave is as follows: Figure 3 .

[0038] Step 3: Calculate the Holospectrum of the displacement signal using a two-dimensional holographic spectrum algorithm, and then extract the variation trend features of the holographic spectrum frequency ellipse from the Holospectrum.

[0039] As an optional approach, the Holospectrum of the displacement signal is calculated using a two-dimensional holographic spectrum algorithm, and then the trend characteristics of the holographic spectrum frequency transfer ellipse are extracted from the Holospectrum, including: 1) Simultaneously collect data from the oxygen pre-compression pump. , Two displacement signals in mutually perpendicular directions are preprocessed to obtain a finite discrete data sequence x. t y t ; 2) The corresponding spectral sequence X is obtained from the finite discrete Fourier transform formula. k Y m ; 3) The spectral sequence is a complex number, which can be written as the sum of its real and imaginary parts: (5); The corresponding amplitudes and phases are as follows: (6); (7); The trigonometric equation for the i-th spectral line at the same frequency of two vibration signals, with time t as a parameter, is as follows: (8); Adding the above equations together and eliminating the parameter t, we obtain the synthesis equation for the i-th spectral line as follows: (9); The trajectory obtained from the equation can be circular, elliptical, or straight, depending on the magnitude of the amplitude and phase difference between the two mutually perpendicular vibration signals involved in the synthesis. Arranging the trajectory diagrams obtained from the above equation in order of spectral line frequency constitutes a two-dimensional holographic spectrum.

[0040] Feature extraction is performed on the two-dimensional holographic spectrum to obtain parameters such as the variation trends of the major and minor semi-axis, and the corresponding graphs are plotted. Specifically, for the i-th spectral line at the same frequency of two vibration signals, the major semi-axis of the ellipse... short half shaft The calculation equation is as follows: (10); Similarly, by calculating the spectral lines of the two signals at different times, the values ​​of the short and long semi-axes at different times can be obtained, thus revealing the trend of the spectral lines' short and long semi-axes changing over time. For example, for two radial vibration displacement signals at the end of an oxygen pump shaft obtained during a hot-run test, a two-dimensional holographic spectrum was calculated, and the trend of the short and long semi-axes changing over time was extracted as follows: Figure 4 , Figure 4 (a) shows the trend of the amplitude of the semi-major axis changing over time. Figure 4 (b) shows the trend of the amplitude of the short semi-axis over time.

[0041] Step 4: Fuse the acceleration signal features and displacement signal features to obtain the fusion trend sensitivity index, and realize the health status detection of the turbopump.

[0042] As an optional approach, the fusion of acceleration signal features and displacement signal features to obtain a fusion trend sensitivity index for detecting the health status of the turbopump includes: The frequency band energy amplitude of each frequency component extracted from the acceleration signal in step 2 Normalization is performed, and then the major and minor semi-axes of the frequency conversion ellipse extracted from the holographic spectrum in step 3 are used. , The mean is used as the weight to achieve a unified magnitude for different components. The calculation equation for the fusion result is as follows: (11); In the formula , —Fusion results; , —Amplitudes of the major and minor semi-axes of the frequency-transition ellipse extracted from the holographic spectrum; , —— , Mean; —Numerical normalization, ; —Frequency components extracted from acceleration signals The frequency band energy amplitude centered on the frequency band.

[0043] Based on the above fusion results , Construct the FDS index and define FDS as the fusion result. , The weighted sum of the local mean and local standard deviation is shown in the following formula.

[0044] (12); In the formula —Standard deviation.

[0045] For example, for a certain type of liquid rocket engine, vibration acceleration and vibration displacement sensor monitoring data related to the turbopump component during hot-fire testing were selected. The vibration acceleration signal mainly includes the oxygen pump vibration acceleration signal, and the vibration displacement signal mainly includes two radial vibration displacement signals at the oxygen pump shaft end. The energy variation trends of nine main frequency components (1n, 3n, 6n, 9n, 12n, 18n, 21n, 24n, and 30n) of the oxygen pump were extracted and fused for analysis according to the aforementioned fusion trend sensitivity index (FDS) construction method. The analysis results are as follows: Figure 5 The nine figures, from a to i, represent the FDS curves of the main turbine speed at the 1st, 3rd, 6th, 9th, 12th, 18th, 21st, 24th, and 30th harmonics over time. It can be seen that, except for the 1n harmonic component, the amplitude of the other frequency components gradually increases at some point between 40s and 45s, eventually exceeding the empirical threshold between 45s and 75s. The FDS curve of the 1n harmonic component undergoes a significant state change at approximately 50s. Based on the post-test disassembly and fault analysis results, the oxygen pump experienced a rubbing-induced fire fault around 67.69s, which is consistent with the identification results of the proposed method, verifying the accuracy and effectiveness of the method. Traditional steady-state RMS analysis and displacement signal analysis methods cannot effectively detect the above-mentioned abnormal process.

Claims

1. A method for detecting the health status of a turbopump based on a fusion trend sensitivity index, characterized in that, Includes the following steps: Step 1: Obtain the turbopump acceleration and displacement signals: An acceleration sensor is installed between the oxidizer pump and the turbine of the engine turbopump. The acceleration sensor is used to measure the vibration signal in the three directions of shaft diameter tangentially. Displacement sensors are installed on the shaft section between the fuel pump and the oxidizer pump, with two sensors arranged in mutually perpendicular directions. During engine hot-run testing, acceleration and displacement signals of the engine turbopump were collected. Step 2, Acceleration signal processing: A high-precision time-frequency domain analysis method is used to process the acceleration signal to obtain the time spectrum; using the time spectrum, the variation trend characteristics of the frequency band energy, i.e., the acceleration signal characteristics, are calculated. Step 3, Displacement signal processing: A two-dimensional holographic spectrum algorithm is used to calculate the holographic spectrum Holospectrum of the displacement signal, and the variation trend features of the holographic spectrum frequency transfer ellipse, i.e., the displacement signal features, are extracted from the Holospectrum. Step 4, Signal Feature Fusion: By fusing acceleration signal features and displacement signal features, a fusion trend sensitivity index is obtained to determine the health status of the liquid rocket engine turbopump.

2. The method for detecting the health status of a turbopump according to claim 1, characterized in that, The acceleration sensor is a vibration acceleration sensor.

3. The method for detecting the health status of a turbopump according to claim 1, characterized in that, The turbopump is a liquid rocket engine turbopump.

4. The method for detecting the health status of a turbopump according to claim 1, characterized in that, Step 2 further includes: The short-time Fourier transform (STFT) of the time-domain signal x(t) yields the following time-frequency spectrum: (1)； In equation (1), F x ω represents the spectrum of the original time-domain signal x(t) after short-time Fourier transform, which contains the amplitude and phase information of the signal in time and frequency; g is a tightly supported window function that can be shifted with time t, and different window functions have different spectral characteristics; ω is the frequency of the harmonic signal. The synchronous compressed wavelet transform (WSST) algorithm is employed. WSST uses a compression operator to compress the dispersed time-frequency coefficients in the STFT spectrum to the instantaneous frequency, thus obtaining a new time-frequency spectrum. The transformation formula is as follows: (2)； (3)； In equations (2) and (3), The time spectrum is obtained by compressing the wavelet coefficient energy along the instantaneous frequency direction; F x The spectrum of the original time-domain signal x(t) after short-time Fourier transform; The Dirac function is used to concentrate energy at an instantaneous frequency. superior; The frequency is instantaneous, and the spectrum is more accurate at this frequency; For each component, the WSST calculation is performed on the acceleration signal to obtain the time spectrum. Using the time spectrum, based on the characteristics of turbopump failure, key frequency components are selected, and a fixed bandwidth is set near the key frequency components. The frequency band energy variation trend characteristics within this bandwidth are extracted. The frequency band energy calculation process is as follows: For an acceleration signal obtained from a single measurement, the signal is first preprocessed, including data splicing, time period truncation, and mean removal. Secondly, a Fourier transform is performed on the preprocessed signal to obtain the frequency domain signal. Thirdly, the frequency component of interest is selected, denoted as f, with a bandwidth of [value missing]. In the frequency range There are n spectral lines in the memory, and the amplitude of the i-th spectral line is denoted as a. i The formula for calculating the band energy at that frequency is as follows: (4)； In equation (4), E nf Frequency range Bandwidth energy, E nf The larger the value, the more concentrated the signal energy is within that frequency band; f is the frequency of interest in time-frequency analysis; n is the frequency range. The number of spectral lines in the spectrum within; a i Let be the amplitude of the i-th spectral line; Finally, the same method was used to calculate the frequency band energy for different time periods, and the trend of frequency band energy change over time was obtained.

5. The method for detecting the health status of a turbopump according to claim 1, characterized in that, Step 3 further includes: Based on two mutually perpendicular displacement signals, the amplitude spectrum and phase spectrum of the displacement signals are calculated respectively; using a two-dimensional holographic spectrum algorithm, the amplitude spectrum and phase spectrum of the two displacement signals are merged to obtain the holographic spectrum of the displacement vibration signal and an ellipse is drawn; the variation trend characteristics of the holographic spectrum frequency-transition ellipse are extracted from the holographic spectrum. First, simultaneous collection , Displacement signals in two mutually perpendicular directions, the two mutually perpendicular directions are , ,right , The displacement signals in two directions are preprocessed to obtain a finite discrete data sequence x. t y t ; Secondly, for a finite discrete data sequence x t y t Performing a finite discrete Fourier transform yields the corresponding spectral sequence X. k Y m ; Next, the spectral sequence is expressed as the sum of the real and imaginary parts, the corresponding amplitude and phase are listed, the synthesis equation of the i-th spectral line is calculated, and the trajectory diagram of the synthesis equation is arranged in order of the magnitude of the spectral line frequency to form a two-dimensional holographic spectrum. Finally, feature extraction is performed on the two-dimensional holographic spectrum to obtain the variation trends of the major and minor semi-axis, and the variation trend graph is plotted, which is the displacement signal feature.

6. The method for detecting the health status of a turbopump according to claim 5, characterized in that, Step 3 further includes: The equation for the synthesis of the i-th spectral line is calculated as follows: (5)； In equation (5), A i Let B be the amplitude of the i-th spectral line in the x-direction. i Let be the amplitude of the i-th spectral line in the y-direction. Let be the phase of the i-th spectral line in the x-direction; The phase of the i-th spectral line in the y-direction; For the i-th spectral line at the same frequency of two displacement signals, the semi-major axis of the ellipse... short half shaft The calculation equation is as follows: (6)； In equation (6), Let be the length of the semi-major axis of the holographic ellipse of the i-th spectral line; A is the length of the minor semi-axis of the holographic spectrum ellipse; i Let B be the amplitude of the i-th spectral line in the x-direction. i Let be the amplitude of the i-th spectral line in the y-direction; Let be the phase of the i-th spectral line in the x-direction; The phase of the i-th spectral line in the y-direction; Similarly, by selecting different times to calculate the spectral lines of the two displacement signals, the values ​​of the short and long semi-axis at different time lengths are obtained, and thus the trend of the change of the short and long semi-axis of the spectral lines with time is obtained.

7. The method for detecting the health status of a turbopump according to claim 1, characterized in that, Step 4 further includes: The frequency band energy amplitudes of each frequency component extracted from the acceleration signal are normalized; the average of the major and minor semi-axes of the frequency conversion ellipse extracted from the holographic spectrum is used as the weight to achieve uniformity of the magnitude of different components. The fusion calculation formula is as follows: (7)； In equation (7), , The result is a fusion of acceleration and displacement signals, which includes information such as the frequency band amplitude and energy of the acceleration signal, and the frequency transition ellipse of the holographic spectrum of the displacement signal. , These are the amplitudes of the major and minor semi-axes of the frequency-transition ellipse extracted from the holographic spectrum. , They are respectively , The mean; For numerical normalization, ; Frequency components extracted from acceleration signals The frequency band energy amplitude centered on; Based on fusion results , Construct a fusion trend sensitivity index FDS, and define FDS as the fusion result. , The weighted average of the local mean and local standard deviation is shown in the following formula; (8) In equation (8), FDS is the proposed fusion trend sensitivity index, which describes the health status of each component of the turbopump. , The result of fusing acceleration and displacement signals, The standard deviation is denoted as .

Images