Non-rotating-speed adaptive data processing method and system for unstable signals

Abstract

The invention belongs to the technical field of digital signal processing, particularly relates to a non-rotating-speed self-adaptive data processing method and system for unstable signals, and belongs to a data-driven self-adaptive processing strategy. Firstly, a time-frequency distribution spectrogram of a signal is calculated through short-time Fourier transform, a time-varying frequency window is designed, and the number of points in the window is determined according to the frequency domain width and the sampling frequency of the signal. Thirdly, calculating the maximum energy in each frequency window, dynamically adjusting the center frequency of the window, extracting a frequency ridge line in the optimal time-varying frequency window, and obtaining the instantaneous frequency distribution of the original signal; and finally, constructing a secondary sampling time sequence according to the instantaneous frequency distribution, and resampling the original signal by using cubic spline interpolation to obtain a new steady-state signal. The problem of signal processing under the condition of no rotating speed is solved, the precision and robustness of unsteady state signal processing are remarkably improved, and the method is suitable for a scene in which rotating speed data and vibration or sound data are difficult to accurately synchronize in a wireless monitoring system.

Description

Technical Field

[0001] The invention belongs to the technical field of digital signal processing, and in particular relates to a non-speed adaptive data processing method and system for unstable signals. Background Art

[0002] At present, the fault diagnosis of rotating equipment is mainly based on the analysis and processing of stable signals, such as spectrum analysis, frequency doubling analysis, sideband analysis, frequency band energy, etc. This type of analysis method cannot fully utilize the information contained in the non-stationary signals generated during the equipment's rise, fall, speed change, etc. In addition, we often encounter a large number of non-stationary signals when diagnosing mechanical faults. This is usually because the rotating equipment will experience load fluctuations, power supply fluctuations, speed fluctuations, etc. during operation, resulting in the acquisition of non-stationary sound or vibration signals. If Fourier analysis is directly performed on the basis of non-stationary signals to obtain a spectrum, severe frequency ambiguity will occur, resulting in the fault characteristics being completely concealed and unable to be effectively diagnosed.

[0003] Many experts and scholars have proposed a variety of solutions and technical paths to this kind of problem. The most mature method currently used is to install a speed sensor on the equipment, and perform order sampling on the original sound or vibration signal by encrypting the collected speed information (H. Hotmansport Order Analysis System CN107192550B). On the hardware side, the speed data is fused with the vibration or sound signal to directly achieve equal-angle order sampling of periodic motion equipment. Another method is to collect the speed information of the equipment through the speed sensor, draw the angle change curve according to the speed, divide the angle change curve into equal angles, determine the time series of equal-angle resampling, and implement interpolation resampling from the software (Zhao Hongshan A wind turbine bearing fault diagnosis method under variable speed CN105784366A).

[0004] Another strategy is to obtain the key phase or speed information of the equipment by deeply analyzing the original signal without installing the speed sensor. Jiang Zhansi et al. proposed a bearing fault diagnosis method based on comprehensive analysis of order and envelope spectrum (Jiang Zhansi's method for fault diagnosis of rolling bearings in rotating machinery based on order spectrum and envelope spectrum CN109520738B), but the method did not clearly point out the calculation steps of its order spectrum (COT calculation order tracking algorithm), and the accuracy and divergence problems in the process of calculating order analysis have been solved. There is also a method of extracting the peak sequence in the time domain signal to approximate the phase information of the signal and further realize the order analysis of the non-steady-state signal (Zeng Zhijie's method for order analysis of wind power gear vibration signals using time series peak search CN109063387B). This method obtains the peak time series of the narrowband signal through bandpass filtering. When the frequency components of the original signal are complex and there are multi-frequency components in a single narrow band, this method will not be able to achieve effective order analysis. Due to the uncertainty of time domain characteristics, Chen Jian adopted a time-frequency ridge extraction method based on fault feature interpretation to obtain the time-frequency distribution of the original signal, and further completed the order resampling analysis (Chen Jian A time-frequency ridge extraction method and device for fault feature interpretation CN 111797789A). Although this method can ensure the convergence of the algorithm, it extracts multiple ridges in the global spectrum of the time-frequency distribution and uses these ridges for order analysis at the same time. It cannot guarantee that the optimal solution for steady-state signal recovery can be obtained. Moreover, when all ridges are extracted, there may be some invalid ridges, which will cause misdiagnosis in the final fault analysis.

[0005] Among the existing methods, there are two technical paths: one is to obtain the instantaneous speed information of the equipment and perform order analysis by adding a speed sensor through external acquisition hardware; the other method is to obtain the phase or instantaneous frequency of the original signal from the time domain waveform or time-frequency distribution of the signal through data-driven method, and perform secondary sampling in software to obtain an approximate steady-state signal. However, the existing methods have one or more of the following defects and deficiencies: 1. When installing a speed sensor on the equipment, unlike vibration and sound sensors, most speed sensors use the principle of distance change. Therefore, a relatively static fixed point is required during installation, and the probe needs to point directly to the rotating part, which brings great difficulties to on-site construction and installation; the commonly used photoelectric speed sensors have certain requirements for the environment. For sites with harsh environments, the accumulation of dust may cause collection failure, and another type of Hall speed sensor requires bonding or welding of metal bonds on the rotating parts, which is often difficult to achieve on site.

[0006] 2. On the other hand, in the existing equipment detection system, wireless monitoring has gradually replaced the original wired monitoring system due to its own advantages and is widely used. The wireless monitoring system is generally composed of wireless sensors and wireless acquisition gateways. The sensors are usually powered by batteries, and absolute time is used to achieve synchronization between each measurement point. This synchronization method has low accuracy and is difficult to achieve in hardware for order analysis that requires high-precision synchronous acquisition of speed and vibration signals.

[0007] 3. Currently, there are two main types of technical solutions for speedless order analysis. The first is to extract the phase information of the original signal through the waveform through precise analysis of the time domain signal. However, this type of method requires the original signal to have a high signal-to-noise ratio or relatively obvious time domain periodic characteristics, such as periodic impacts, periodic pulse signals, etc. However, the vibration or sound signals in engineering applications basically contain a lot of background noise, and the time domain characteristics are not obvious. In this case, this type of method is often difficult to implement.

[0008] 4. Another method is based on the time-frequency analysis of the original signal, such as short-time Fourier transform, wavelet transform, Gabor transform, etc. This type of method generally requires filtering the original signal before performing time-frequency analysis. Whether selecting the envelope signal of the low-frequency band or the resonance band for time-frequency analysis, it is necessary to involve the design of filter parameters, which will bring certain uncertainties to the results of frequency extraction. In addition, most of the current time-frequency ridge extraction processes require manual intervention, and the ridge extracted by the simple peak optimal extraction may come from different frequency components in the original signal, resulting in the failure of order analysis. Summary of the invention

[0009] The purpose of the present invention is to provide a speed-free adaptive data processing method for non-steady signals, so as to solve the problem in the prior art of how to adaptively process non-steady-state vibration or sound signals in a completely data-driven manner in the absence of speed data, extract instantaneous frequency characteristics and restore spectral line characteristics in the spectrum, thereby overcoming the frequency ambiguity problem caused by the lack of speed information in traditional methods.

[0010] The present invention achieves the above-mentioned purpose through the following technical solutions: In a first aspect, the present invention proposes a non-speed adaptive data processing method for an unstable signal, the method comprising: Receive the original non-steady-state signal of the device under test and calculate its time-frequency distribution spectrum; Designing a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension, wherein the number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal; Calculating the maximum energy in the current frequency window, and adjusting the center frequency of the next frequency window according to the maximum energy; Obtaining the maximum frequency energy sum of all time-varying frequency windows, and determining the optimal time-varying frequency window according to the maximum frequency energy sum; Extracting the frequency ridge within the optimal time-varying frequency window to obtain the instantaneous frequency distribution of the original non-steady-state signal; A secondary sampling time series is constructed according to the instantaneous frequency distribution, and the original non-steady-state signal is resampled to obtain a new steady-state signal of the device under test.

[0011] Furthermore, the original non-steady-state signal is any one of a vibration signal, a sound signal, a current signal, and a voltage signal.

[0012] Furthermore, the time-frequency distribution spectrum is obtained by short-time Fourier transform, and the calculation parameters of the short-time Fourier transform include window function, window length, overlap rate and sampling frequency, and the window function is a rectangular window.

[0013] Furthermore, designing a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension includes: Construct a time-varying frequency domain window function with a fixed window length and a center frequency that varies with time , The window length and center frequency satisfy the following conditions: ; In the above formula, is the discretized index of time domain and frequency domain, Indicates the starting center frequency is At the moment The frequency domain window function is For a fixed window length, for time The center frequency of the window, is the window length of the window function, is the frequency in the time-frequency distribution.

[0014] Furthermore, the method for determining the optimal time-varying frequency window includes: Calculate the maximum energy in the time-varying frequency window corresponding to each starting center frequency; Sum the maximum energy of all time-varying frequency windows to obtain the energy and corresponding to each starting center frequency , as follows: ; Select energy and The time-varying frequency window corresponding to the maximum starting center frequency is taken as the optimal time-varying frequency window, as shown in the following formula: ; In the above formula, is the sampling frequency of the original signal.

[0015] Further, constructing a secondary sampling time series according to the instantaneous frequency distribution includes: Determine the optimal instantaneous frequency of the original signal The characteristic expression is as follows: ; According to the instantaneous frequency Calculate the sampling interval at each moment as follows: ; In the above formula, and Respectively The instantaneous sampling frequency and sampling interval of the secondary sampling at the moment, is the number of sampling points, is a fixed sampling interval, is the average sampling frequency, is the sampling frequency of the original signal; Using the sampling interval to reverse the cubic spline to establish a secondary sampling time series , as follows: ; In the above formula, to Represents the regeneration of a uniform time series. A point in time, Indicates the corresponding time, is a fixed sampling interval.

[0016] Furthermore, resampling the original non-steady-state signal to obtain a new steady-state signal of the device under test includes: Get time series The signal amplitude corresponding to each moment in ; The original non-steady-state signal is interpolated by cubic spline and combined with the signal amplitude. Interpolate to get the new steady-state signal of the device under test , as follows: .

[0017] In a second aspect, the present invention proposes a non-speed adaptive data processing system for non-steady signals, which is applied to execute any of the above-mentioned non-speed adaptive data processing methods, and the system comprises: A signal acquisition module, used for receiving the original non-steady-state signal of the device under test; A time-frequency distribution calculation module, used for calculating the time-frequency distribution spectrum of the original non-steady-state signal by using short-time Fourier transform; A time-varying frequency window design module, used to design a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension, wherein the number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal; A maximum energy calculation module, used to calculate the maximum energy in the current frequency window, and adjust the center frequency of the next frequency window according to the maximum energy; An optimal time-varying frequency window determination module is used to obtain the maximum frequency energy sum of all time-varying frequency windows, and determine the optimal time-varying frequency window according to the maximum frequency energy sum; An instantaneous frequency extraction module, used to extract the frequency ridge within the optimal time-varying frequency window to obtain the instantaneous frequency distribution of the original non-steady-state signal; The resampling module is used to construct a secondary sampling time series according to the instantaneous frequency distribution, resample the original non-steady-state signal, and obtain a new steady-state signal of the device under test.

[0018] Furthermore, a design rule is preset in the time-varying frequency window design module, and the design rule is specifically that the center frequency of the time-varying frequency window varies with time, and the window length of the time-varying frequency window is fixed.

[0019] The beneficial effects of the present invention are: The present invention proposes a completely data-driven adaptive processing strategy for non-steady-state vibration or sound signals when there is no speed data. By calculating the time-frequency distribution spectrogram of the signal, extracting the optimal frequency ridge in the time-frequency graph, obtaining the instantaneous frequency characteristics of the original signal, and then performing secondary sampling to obtain a new steady-state signal. This method effectively solves the frequency ambiguity problem caused by the lack of speed information in traditional methods, significantly improves the accuracy and reliability of non-steady-state signal processing, and is particularly suitable for scenarios in wireless monitoring systems where speed data and vibration or sound data are difficult to accurately synchronize.

[0020] The present invention adopts a method of extracting frequency ridges in the time-frequency domain, which has stronger robustness and universality compared to the method of extracting phase only through time-domain signals. By constructing a time-varying frequency window and optimizing the window function based on the maximum time-frequency spectrum energy, the accuracy and convergence of frequency ridge extraction are ensured. This method not only avoids the ridge extraction error caused by improper frequency window selection in traditional methods, but also significantly improves the robustness and accuracy of non-steady-state signal processing. It is suitable for a variety of non-steady-state signals reflecting the status of equipment, such as vibration, sound, current and voltage signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic flow chart of a method for processing non-speed adaptive data of unstable signals provided in an embodiment of the present application; Figure 2A schematic diagram of another flow chart of the non-speed adaptive data processing method for unstable signals provided in an embodiment of the present application; Figure 3 The original waveform and spectrum of the signal in the specific implementation example of this application; Figure 4 It is the time-frequency distribution spectrum of the original signal in the specific implementation example of this application; Figure 5 A time-varying window center frequency corresponding to a certain starting center frequency in a specific implementation example of the present application and a maximum energy and distribution diagram in the frequency domain window; Figure 6 The optimal time-varying window function group and the local time-frequency distribution diagram within the window in the specific implementation example of this application; Figure 7 It is the instantaneous frequency ridge and the constructed new sampling time series in the specific implementation example of this application; Figure 8 The processed approximate steady-state signal waveform and spectrum in the specific implementation example of this application; Fig. 9 It is the time-frequency distribution spectrum of the approximately steady-state signal after processing in the specific implementation example of the present application. DETAILED DESCRIPTION

[0022] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0023] Example 1

[0024] like Figure 1-2 As shown, this embodiment proposes a speed-free adaptive data processing method for non-steady-state signals, where "speed-free" means that in the signal processing process, there is no need to rely on a speed sensor or other external devices to obtain the speed information of the rotating machinery; the method includes the following steps: receiving the original non-steady-state signal of the device to be tested, and calculating its time-frequency distribution spectrum using STFT (short-time Fourier transform); designing a time-varying frequency window in the time-frequency distribution spectrum in the frequency dimension, and the number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal; calculating the maximum energy in the current frequency window, and adjusting the center frequency of the next frequency window according to the maximum energy; obtaining the maximum frequency energy sum of all time-varying frequency windows, and determining the optimal time-varying frequency window according to the maximum frequency energy sum; extracting the frequency ridge in the optimal time-varying frequency window, and obtaining the instantaneous frequency distribution of the original non-steady-state signal; constructing a secondary sampling time series according to the instantaneous frequency distribution, and resampling the original non-steady-state signal to obtain a new steady-state signal of the device to be tested.

[0025] In a specific embodiment, the original non-steady-state signal is any one of a vibration signal, a sound signal, a current signal, and a voltage signal.

[0026] In a specific embodiment, the time-frequency distribution spectrum is obtained by short-time Fourier transform, and the calculation parameters of the short-time Fourier transform include a window function, a window length, an overlap rate and a sampling frequency, and the window function is a rectangular window.

[0027] In specific implementation, the time-frequency distribution of the original signal may be calculated by other time-frequency analysis methods such as Hilbert-Huang transform, wavelet transform, Gabor transform, etc.

[0028] More specifically, Represents the original input signal. According to the following short-time Fourier transform formula, the time-frequency distribution of the original signal is calculated: ; in Unwindowed function, When using this formula to calculate the time-frequency distribution of discrete signals, it is necessary to determine parameters including window function, window length, overlap rate, sampling frequency, etc.

[0029] In the above scheme, a time-varying frequency window is designed in the frequency dimension in the time-frequency distribution spectrum, including: constructing a time-varying frequency domain window function with a fixed window length and a center frequency that varies with time , The window length and center frequency satisfy the following conditions: ; In the above formula, is the discretized index of time domain and frequency domain, Indicates the starting center frequency is At the moment The frequency domain window function is For a fixed window length, for time The center frequency of the window, is the window length of the window function, is the frequency in the time-frequency distribution.

[0030] The method for determining the optimal time-varying frequency window includes: for each starting center frequency A corresponding set of time-varying window functions is calculated, and the maximum energy in the time-varying frequency window corresponding to each starting center frequency is calculated under each window function; the maximum energy of all time-varying frequency windows is summed to obtain the energy and corresponding to each starting center frequency , as follows: ; Finally, calculate all The maximum value in obtains the optimal time-varying window function group , select energy and The time-varying frequency window corresponding to the maximum starting center frequency is taken as the optimal time-varying frequency window, as shown in the following formula: ; In the above formula, is the sampling frequency of the original signal.

[0031] It should be emphasized that in the implementation process of the present invention, when constructing the time-varying frequency window, the maximum time-frequency spectrum energy in each group of time-varying windows is used as the optimization index of the window function. Specifically, for each starting center frequency The corresponding set of time-varying window functions is used to calculate the maximum energy in the frequency domain under each window function, and the maximum energy in all time-varying windows is summed to obtain the energy and corresponding to each starting center frequency. By comparing all The value of is selected, and the time-varying frequency window corresponding to the energy and the largest starting center frequency is selected as the optimal time-varying frequency window.

[0032] This optimization index ensures that the time-varying frequency window can adaptively focus on the most significant energy change area in the signal, thereby improving the accuracy and reliability of frequency ridge extraction. This method not only avoids the ridge extraction error caused by improper frequency window selection in traditional methods, but also significantly improves the robustness and accuracy of non-stationary signal processing.

[0033] In a specific embodiment, the frequency ridge within the above-mentioned optimal time-varying window is extracted to obtain the optimal instantaneous frequency characteristic expression of the original signal, and a secondary sampling time series is constructed according to the instantaneous frequency distribution, including: determining the optimal instantaneous frequency of the original signal The characteristic expression is as follows: ; In the implementation process of the present invention, the extraction of the time-frequency ridges and instantaneous frequency distribution of the original signal is realized by the optimal time-varying window function group, which ensures the validity and convergence of the extraction results. Specifically, the optimal time-varying frequency window is first determined according to the maximum energy and the time-frequency distribution spectrum, and then the frequency ridges are extracted within the optimal window to obtain the instantaneous frequency distribution of the original signal. Since the optimal time-varying window function group can adaptively focus on the most significant energy change area in the signal, the ridge extraction error caused by improper frequency window selection in the traditional method is avoided. Through this method, the instantaneous frequency characteristics of the original signal can be effectively extracted to ensure the accuracy of subsequent resampling and signal processing. This method not only improves the robustness of non-steady-state signal processing, but also significantly improves the convergence of frequency ridge extraction.

[0034] In a non-steady-state signal, since the signal frequency changes with time, if a fixed sampling interval is used The signal is sampled, so the number of sampling points in each cycle is will also change with time. In order to To restore the signal frequency to a steady state, it is necessary to take an unfixed time interval to ensure that there is a fixed number of sampling points in each cycle. , then the time-varying sampling frequency The relationship between the instantaneous frequency of the signal should be as follows: ; In the above formula, and Respectively The instantaneous sampling frequency and sampling interval of the secondary sampling at the moment, is the number of sampling points, is a fixed sampling interval, is the average sampling frequency, is the sampling frequency of the original signal; Constructing a quadratic sampling time series using sampling interval inverse cubic spline , as follows: ; In the above formula, to Represents the regeneration of a uniform time series. A point in time, Indicates the corresponding time, is a fixed sampling interval.

[0035] In a specific embodiment, the original non-steady-state signal is resampled to obtain a new steady-state signal of the device under test, including: obtaining a time series The signal amplitude corresponding to each moment in the original non-steady-state signal is calculated by using cubic spline interpolation and combining the signal amplitude. Interpolate to get the new steady-state signal of the device under test , as follows: .

[0036] Based on the same inventive concept, this embodiment further proposes a non-speed adaptive data processing system for unstable signals, which is applied to execute the non-speed adaptive data processing method as described above, and the system includes: A signal acquisition module, used for receiving the original non-steady-state signal of the device under test; A time-frequency distribution calculation module is used to calculate the time-frequency distribution spectrum of the original non-steady-state signal using short-time Fourier transform; A time-varying frequency window design module is used to design a time-varying frequency window in the frequency dimension in the time-frequency distribution spectrum. The number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal. The maximum energy calculation module is used to calculate the maximum energy in the current frequency window and adjust the center frequency of the next frequency window according to the maximum energy; An optimal time-varying frequency window determination module is used to obtain the maximum frequency energy sum of all time-varying frequency windows and determine the optimal time-varying frequency window according to the maximum frequency energy sum; The instantaneous frequency extraction module is used to extract the frequency ridge within the optimal time-varying frequency window and obtain the instantaneous frequency distribution of the original non-steady-state signal; The resampling module is used to construct a secondary sampling time series according to the instantaneous frequency distribution, resample the original non-steady-state signal, and obtain a new steady-state signal of the device under test.

[0037] In a specific embodiment, a design rule is preset in the time-varying frequency window design module, and the design rule is specifically that the center frequency of the time-varying frequency window varies with time, and the window length of the time-varying frequency window is fixed.

[0038] It should be noted here that each module in the above-mentioned speed-free adaptive data processing system corresponds to each step in implementing the above-mentioned speed-free adaptive data processing method, and the instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0039] In order to more clearly illustrate the present invention and its advantages, the method provided by the present invention will be further explained below in conjunction with specific examples and related partial figures.

[0040] 1. Install the acceleration sensor on the surface of the equipment to be monitored, sample at a certain sampling frequency, and obtain the vibration signal (The present invention is not limited to vibration signals in practical applications, but can also be non-steady-state process signals such as sound, current, voltage, etc. that can reflect the variable frequency operation state of the equipment.) The main acquisition parameters used in this example are: sampling frequency , number of sampling points .

[0041] the following Figure 3 The specific implementation steps of the present invention are further described by taking the original signal shown as an example, wherein the upper figure is the waveform of the original signal and the lower figure is the corresponding spectrum. It can be clearly seen from the spectrum that since the device is in a variable speed state during operation, the original frequency of the signal is broadened, making it difficult to accurately distinguish the original frequency components.

[0042] Figure 4This is the time-frequency distribution spectrum of the original non-steady-state signal. The bright colors in the figure represent the positions with high spectral energy, and the dark colors represent the positions with low spectral energy. It can be seen that there are several obvious frequency components between 1000Hz and 1400Hz, but these frequencies change with time, which indicates that the original signal has frequency fluctuations in this time period and is a non-steady-state signal. This phenomenon leads to Figure 2 Frequency broadening appears on the spectrum. The time-frequency analysis parameters used in this example are: rectangular window, window length 256, overlap rate 255, and analysis frequency 2000Hz.

[0043] 2. Time-frequency distribution of the original signal , take a fixed frequency domain window width , starting center frequency , randomly select the starting center frequency The corresponding time-varying window function frequency center sequence is as follows Figure 5 As shown above, it can be seen that the center frequency of the frequency window at different times has certain fluctuations. Further calculate the maximum energy in the frequency domain of each window, and sum up all the maximum energies to obtain like Figure 5 As shown below, it can be seen There will be a maximum value at a certain center frequency.

[0044] 3. According to the previous step The result is calculated to obtain the optimal time-varying window function group like Figure 6 As shown above, the local time-frequency distribution within the frequency window is as follows Figure 6 Down.

[0045] Extract the time-frequency distribution ridge within the window function group, and finally obtain the approximate instantaneous frequency distribution of the original signal ( Figure 7 superior).

[0046] 4. Get the instantaneous frequency Finally, the instantaneous frequency integral is used to calculate the total number of signal cycles in the actual sampling time interval. , and then use the inverse cubic spline to establish a new sampling time series .

[0047] The time difference between the new sampling time series and the original sampling time series is as follows: Figure 7 As shown below, it can be seen that at frequencies above When the subsampling interval becomes shorter, the sampling frequency increases, and the frequency is lower than The subsampling interval becomes longer and the sampling frequency becomes slower.

[0048] 5. Finally, use the resampled time series Combined with the cubic spline interpolation of the original signal, a new approximate steady-state signal is obtained like Figure 8 As shown in the figure above, the time domain waveform is shown, and the spectrum is shown below. From the spectrum, we can see that the broadening phenomenon of several original main frequency components has been well eliminated, and several frequency components are clearly visible. Fig. 9 It can also be seen from the time-frequency spectrum that the main frequency components no longer change with time, but basically become a straight line, indicating that the original non-steady-state signal has been processed into an approximate steady-state signal, which is convenient for further diagnosis and analysis.

[0049] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0050] In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0051] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A non-speed adaptive data processing method for unstable signals, characterized in that: The method comprises: Receive the original non-steady-state signal of the device under test and calculate its time-frequency distribution spectrum; Designing a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension, wherein the number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal; Calculating the maximum energy in the current frequency window, and adjusting the center frequency of the next frequency window according to the maximum energy; Obtaining the maximum frequency energy sum of all time-varying frequency windows, and determining the optimal time-varying frequency window according to the maximum frequency energy sum; Extracting the frequency ridge within the optimal time-varying frequency window to obtain the instantaneous frequency distribution of the original non-steady-state signal; A secondary sampling time series is constructed according to the instantaneous frequency distribution, and the original non-steady-state signal is resampled to obtain a new steady-state signal of the device under test.

2. The non-speed adaptive data processing method of a non-steady signal according to claim 1 is characterized in that: The original non-steady-state signal is any one of a vibration signal, a sound signal, a current signal, and a voltage signal.

3. The non-speed adaptive data processing method of a non-steady signal according to claim 1 is characterized in that: The time-frequency distribution spectrum is obtained by short-time Fourier transform, and the calculation parameters of the short-time Fourier transform include window function, window length, overlap rate and sampling frequency, and the window function is a rectangular window.

4. The method for non-speed adaptive data processing of unstable signals according to claim 1, characterized in that: The step of designing a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension includes: Construct a time-varying frequency domain window function with a fixed window length and a center frequency that varies with time The window length and center frequency satisfy the following conditions: ； In the above formula, is the discretized index of time domain and frequency domain, Indicates the starting center frequency is At the moment The frequency domain window function is For a fixed window length, for time The center frequency of the window, is the window length of the window function, is the frequency in the time-frequency distribution.

5. The method for non-speed adaptive data processing of unstable signals according to claim 4, characterized in that: The method for determining the optimal time-varying frequency window includes: Calculate the maximum energy in the time-varying frequency window corresponding to each starting center frequency; Sum the maximum energy of all time-varying frequency windows to obtain the energy and corresponding to each starting center frequency , as follows: ； Select energy and The time-varying frequency window corresponding to the maximum starting center frequency is taken as the optimal time-varying frequency window, as shown in the following formula: ； In the above formula, is the sampling frequency of the original signal.

6. The method for non-speed adaptive data processing of unstable signals according to claim 5, characterized in that: The constructing a secondary sampling time series according to the instantaneous frequency distribution comprises: Determine the optimal instantaneous frequency of the original signal The characteristic expression is as follows: ； According to the instantaneous frequency Calculate the sampling interval at each moment as follows: ； In the above formula, and Respectively The instantaneous sampling frequency and sampling interval of the secondary sampling at the moment, is the number of sampling points, is a fixed sampling interval, is the average sampling frequency, is the sampling frequency of the original signal; Using the sampling interval to reverse the cubic spline to establish a secondary sampling time series , as follows: ； In the above formula, to Represents the regeneration of a uniform time series. A point in time, Indicates the corresponding time, is a fixed sampling interval.

7. The method for non-speed adaptive data processing of non-steady signals according to claim 6, characterized in that: The resampling of the original non-steady-state signal to obtain a new steady-state signal of the device under test includes: Get time series The signal amplitude corresponding to each moment in ; The original non-steady-state signal is interpolated by cubic spline and combined with the signal amplitude. Interpolate to get the new steady-state signal of the device under test , as follows: 。 8. A non-speed adaptive data processing system for non-steady signals, used to execute the non-speed adaptive data processing method according to any one of claims 1 to 7, characterized in that: The system comprises: A signal acquisition module, used for receiving the original non-steady-state signal of the device under test; A time-frequency distribution calculation module, used for calculating the time-frequency distribution spectrum of the original non-steady-state signal by using short-time Fourier transform; A time-varying frequency window design module, used to design a time-varying frequency window in the time-frequency distribution spectrum in a frequency dimension, wherein the number of points in the time-varying frequency window is determined according to the frequency domain width and sampling frequency of the original non-steady-state signal; A maximum energy calculation module, used to calculate the maximum energy in the current frequency window, and adjust the center frequency of the next frequency window according to the maximum energy; An optimal time-varying frequency window determination module is used to obtain the maximum frequency energy sum of all time-varying frequency windows, and determine the optimal time-varying frequency window according to the maximum frequency energy sum; An instantaneous frequency extraction module, used to extract the frequency ridge within the optimal time-varying frequency window to obtain the instantaneous frequency distribution of the original non-steady-state signal; The resampling module is used to construct a secondary sampling time series according to the instantaneous frequency distribution, resample the original non-steady-state signal, and obtain a new steady-state signal of the device under test.

9. The non-speed adaptive data processing system for unstable signals according to claim 8, characterized in that: The time-varying frequency window design module has a preset design rule, and the design rule is specifically that the center frequency of the time-varying frequency window varies with time, and the window length of the time-varying frequency window is fixed.

Images