Equipment fault diagnosis method and system

Abstract

The invention relates to an equipment fault diagnosis method and system. The method comprises the following steps: performing signal decomposition on an original vibration signal to obtain a group of intrinsic mode components; for each intrinsic mode component, calculating a kurtosis value and an order cyclic spectral density amplitude; based on the kurtosis value and the order cyclic spectral density amplitude, constructing a frequency band selection index used for representing fault impact energy and periodical modulation certainty; determining one or more intrinsic mode components enabling the frequency band selection index to be optimized, and determining an optimal filtering frequency band according to the center frequency and the bandwidth of the optimized intrinsic mode components; performing Hilbert transform on the filtered signal to obtain an analytic signal; extracting the instantaneous amplitude of the analysis signal to obtain an envelope signal; performing cepstrum editing on the envelope signal to obtain an enhanced envelope signal, and performing fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; and performing fault diagnosis based on the fault characteristic frequency in the enhanced envelope spectrum.

Description

Technical Field

[0001] The present invention relates to the field of fault diagnosis, and in particular to a method and system for diagnosing equipment faults. Background Art

[0002] Compressors, motors, and other critical machinery play a vital role in numerous industrial sectors, including petrochemicals, energy and power generation, refrigeration and air conditioning, and gas transportation. The stability and reliability of their operating conditions are directly linked to the efficiency, safety, and economic benefits of the entire production process. However, because these equipment often operate under harsh conditions such as high temperature, high pressure, high speed, and complex alternating loads for long periods of time, critical components such as bearings, valves, pistons, rotors, and cylinders are inevitably susceptible to various failures due to wear, fatigue, corrosion, poor assembly, or lubrication failure. If these early failures are not promptly and effectively diagnosed and addressed, they can lead to reduced compressor performance and increased energy consumption at best, or catastrophic component failure or even unplanned downtime, resulting in significant economic losses and safety hazards. Traditional compressor fault diagnosis methods rely primarily on manual inspections and regular preventive maintenance. This approach is not only inefficient but also struggles to detect even subtle early signs of failures. Failures are often only detected after they have already developed into more serious problems, missing the optimal time for repair.

[0003] Vibration signals carry rich dynamic information about the operating status of various components within a compressor. When a component fails, its vibration characteristics change accordingly. Envelope analysis is an effective method widely used to extract the periodic impulse signatures generated by local faults in rotating machinery, particularly rolling bearings and gears. In envelope analysis, accurately selecting the optimal filter frequency band is crucial and directly impacts the subsequent fault signature extraction. Spectral Kurtosis (SK) automatically selects the optimal resonant frequency band by evaluating the kurtosis of signals within different frequency bands. However, traditional kurtosis-based frequency band selection methods have inherent limitations. First, kurtosis is sensitive to all impulse components, including strong random pulses that are not periodically generated by the target fault or occasional impulses during operation. This may result in the selected frequency band not being truly excited by the periodic fault impulse. Second, in the presence of strong background noise or early, weak faults, the fault-induced impulse signal is weak, and its kurtosis signature may be less prominent or even overwhelmed by the noise, causing the spectral kurtosis method to fail or select suboptimal frequency bands. Even if an envelope signal is obtained through certain filtering methods, it may still be contaminated with residual broadband noise and non-periodic interference. These interferences can mask the weak fault characteristic frequencies in the envelope spectrum, making accurate fault identification difficult, especially when the signal-to-noise ratio is low or in the early stages of a fault. Therefore, further improving the selection of the optimal filtering frequency band and effectively enhancing the extracted envelope signal to highlight fault characteristics are currently urgent technical challenges in the field of compressor fault diagnosis. Summary of the Invention

[0004] In order to solve the above problems, in a first aspect of the present invention, a method for diagnosing equipment faults is provided, comprising the following steps: Obtaining an original vibration signal of the device to be diagnosed, performing signal decomposition on the original vibration signal to obtain a set of intrinsic modal components; calculating a kurtosis value and an order cyclic spectral density amplitude for each intrinsic modal component; and constructing a frequency band selection index for characterizing the fault impact energy and periodic modulation determinism based on the kurtosis value and the order cyclic spectral density amplitude; Determining one or more natural modal components that optimize the frequency band selection index, and determining an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural modal components; performing bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural modal components using the optimal filtering frequency band to obtain a filtered signal; Performing a Hilbert transform on the filtered signal to obtain an analytical signal; extracting the instantaneous amplitude of the analytical signal to obtain an envelope signal; applying cepstrum editing to the envelope signal to obtain an enhanced envelope signal; performing a fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; and performing fault diagnosis based on the fault characteristic frequency in the enhanced envelope spectrum.

[0005] Preferably, for each intrinsic modal component, the kurtosis value and the order cyclic spectral density amplitude are calculated as follows: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

[0006] Preferably, the frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

[0007] Preferably, the determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

[0008] Preferably, the determining of the optimal filtering frequency band according to the center frequency and bandwidth of the optimized natural mode component is specifically as follows: Using the center frequency of the optimized natural mode component as the center frequency of the optimal filtering frequency band; The width of the region where the spectrum energy of the optimized natural mode component is concentrated is used as the bandwidth of the optimal filtering frequency band.

[0009] Preferably, the applying cepstrum editing to the envelope signal to obtain the enhanced envelope signal is specifically: Calculating the real cepstrum of the envelope signal; In the real cepstrum domain, bandpass boosting is used to retain the cepstrum components associated with periodic fault impulses and to attenuate or remove the cepstrum components corresponding to broadband noise and non-periodic interference; The edited cepstrum is inversely transformed to obtain the enhanced envelope signal.

[0010] In a second aspect of the present invention, a device fault diagnosis system is provided, comprising the following modules: An indicator determination module is configured to obtain an original vibration signal of the device to be diagnosed, perform signal decomposition on the original vibration signal to obtain a set of intrinsic modal components; calculate a kurtosis value and an order cyclic spectral density amplitude for each intrinsic modal component; and construct a frequency band selection indicator for characterizing the fault impact energy and periodic modulation determinism based on the kurtosis value and the order cyclic spectral density amplitude; a filtering module, configured to determine one or more natural mode components that optimize the frequency band selection index, and determine an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural mode components; and perform bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural mode components using the optimal filtering frequency band to obtain a filtered signal; The diagnostic module is used to perform a Hilbert transform on the filtered signal to obtain an analytical signal; extract the instantaneous amplitude of the analytical signal to obtain an envelope signal; apply cepstrum editing to the envelope signal to obtain an enhanced envelope signal; perform a fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; and perform fault diagnosis based on the fault characteristic frequency in the enhanced envelope spectrum.

[0011] Preferably, for each intrinsic modal component, the kurtosis value and the order cyclic spectral density amplitude are calculated as follows: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

[0012] Preferably, the frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

[0013] Preferably, the determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

[0014] Preferably, the determining of the optimal filtering frequency band according to the center frequency and bandwidth of the optimized natural mode component is specifically as follows: Using the center frequency of the optimized natural mode component as the center frequency of the optimal filtering frequency band; The width of the region where the spectrum energy of the optimized natural mode component is concentrated is used as the bandwidth of the optimal filtering frequency band.

[0015] Preferably, the applying cepstrum editing to the envelope signal to obtain the enhanced envelope signal is specifically: Calculating the real cepstrum of the envelope signal; In the real cepstrum domain, bandpass boosting is used to retain the cepstrum components associated with periodic fault impulses and to attenuate or remove the cepstrum components corresponding to broadband noise and non-periodic interference; The edited cepstrum is inversely transformed to obtain the enhanced envelope signal.

[0016] The present invention adaptively decomposes the original vibration signal and constructs a frequency band selection index based on the kurtosis value of the inherent modal component and the order cyclic spectral density amplitude. This can more accurately identify and select the resonant frequency band dominated by the periodic fault impact of the equipment, effectively overcoming the defect of the traditional spectral kurtosis method that only relies on the impact intensity and is susceptible to random pulse interference, and improving the robustness and accuracy of the optimal filter frequency band selection. Moreover, by applying the cepstrum editing technology to the demodulated envelope signal, the broadband noise and non-periodic interference components are effectively suppressed in the cepstrum domain, while retaining or enhancing the components related to the periodic fault impact, thereby obtaining an enhanced envelope signal and enhanced envelope spectrum with a higher signal-to-noise ratio and more prominent fault characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of Example 1; Figure 2 Schematic diagram of the original vibration signal and the decomposed modal components; Figure 3 Schematic diagram of the spectrum diagram and band selection indicators of IMF2; Figure 4 Schematic diagram of the signal spectrum and the determined optimal filtering frequency band; Figure 5 Schematic diagram of the edited real cepstrum. DETAILED DESCRIPTION

[0018] In this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Figure 1 1 shows a flow chart of a first embodiment of the present invention, as shown in FIG. Figure 1 As shown, the following steps are included: S1, obtaining an original vibration signal of the device to be diagnosed, performing signal decomposition on the original vibration signal to obtain a set of intrinsic modal components; calculating a kurtosis value and an order cyclic spectral density amplitude for each intrinsic modal component; and constructing a frequency band selection index for characterizing the fault impact energy and periodic modulation determinism based on the kurtosis value and the order cyclic spectral density amplitude; Accelerometers, velocity sensors, or displacement sensors installed at key measurement points on the compressor housing, such as near the bearing seat, continuously collect vibration signals over a period of time during normal operation or under specific operating conditions. The acquisition system converts the analog signals into digital signals and stores them as time series data files.

[0021] Add white noise of different amplitudes to the original vibration signal multiple times to form multiple new signal sequences. Perform the standard EMD decomposition process independently on each signal sequence with white noise added, and decompose it into a series of intrinsic mode functions and a residual term. Figure 2 The original signal and the decomposed IMF1 and IMF2 are shown, where IMF1 is a low-frequency component and IMF2 is a high-frequency component.

[0022] For each decomposed intrinsic mode component, calculate its peak factor. The peak factor is the ratio of the maximum absolute value of the signal to its root mean square (RMS) value. The more significant the impact, the larger the peak factor. Obtain the device's instantaneous speed signal, for example, by calculating it from a key-phase pulse signal. For each IMF, first square it to obtain its instantaneous energy. The resulting instantaneous energy signal is then synchronously averaged based on the instantaneous speed signal. This means the signal is segmented according to each shaft revolution. Then, all periodic signal segments are aligned and averaged. This enhances periodic components synchronized with the speed and suppresses asynchronous noise and interference. A fast Fourier transform is performed on the synchronously averaged energy signal to obtain an order spectrum. From this order spectrum, the amplitude at specific predetermined orders associated with potential faults, such as the bearing outer race fault order and inner race fault order, is extracted as a measure of the order cyclic modulation strength of the IMF.

[0023] For example, for the intrinsic modal component IMF3, the amplitude sequence of some of its time-domain sampling points is [0.5, 0.8, -0.2, 0.1, 0.7, -0.3, ...]. The equipment synchronous speed signal is obtained by a key phase sensor installed on the rotating shaft, and the calculated instantaneous speed is 600 rpm. The sampling frequency of the vibration signal is assumed to be 1000 Hz. When calculating the instantaneous energy signal of IMF3, using the aforementioned sampling points as an example, the corresponding instantaneous energy sequence is [0.25, 0.64, 0.04, 0.01, 0.49, 0.09, ...]. Based on a 10 Hz rotational speed and a 1000 Hz sampling frequency, each shaft revolution corresponds to 100 sampling points. The instantaneous energy signal is segmented into segments of 100 sampling points. After all segments are aligned based on the key phase pulse position, a point-by-point arithmetic averaging is performed to obtain a synchronously averaged energy signal. Assume that after synchronous averaging, the sequence values are [0.28, 0.55, 0.06, 0.03, ...]. Then, a fast Fourier transform is performed on this synchronously averaged energy signal to obtain its order spectrum. If the predetermined characteristic order associated with a specific gear mesh fault is 4.2, the amplitude corresponding to order 4.2 is searched in the order spectrum. For example, analyzing the order spectrum reveals that the amplitude at order 4.0 is 0.15 and the amplitude at order 4.1 is 0.30. However, the amplitude at order 4.2, the characteristic order of the target fault, is 0.75, and the amplitude at order 4.3 is 0.25. The magnitude of the 4.2-order cyclic spectral density of IMF3 is determined to be 0.75. This value will be used as a measure to characterize the modulation intensity of this specific fault in IMF3 and will be used for subsequent frequency band selection and fault diagnosis.

[0024] All peak factors calculated from IMFs are sorted and assigned ordinal rankings. All order-specific order amplitudes calculated from IMFs are sorted and assigned ordinal rankings, and a comprehensive score is calculated such that the higher the ranking, the greater the comprehensive score. In one embodiment, a fuzzy logic-based rule system is used, with kurtosis and order cyclic spectral density amplitudes as inputs, to output a comprehensive fault indication level as a frequency band selection indicator.

[0025] S2, determining one or more natural modal components that optimize the frequency band selection index, and determining an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural modal components; using the optimal filtering frequency band to perform bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural modal components to obtain a filtered signal; The frequency band selection index values calculated for all intrinsic modal components are arranged according to the center frequencies of their corresponding modes. A local peak search algorithm is used to find significant local maxima within this arranged index value sequence. The intrinsic modal components corresponding to these local maxima are selected. For example, the IMFs corresponding to the two or three local peaks with the highest index values are selected as the optimized intrinsic modal components, thereby identifying multiple frequency bands that may contain fault information.

[0026] If multiple optimized natural mode components are selected, the approximate center frequency of each selected IMF is calculated and the frequency band where its energy is significantly concentrated , The lower cutoff frequency of the optimal filtering band is defined as the minimum value among the lower limits of all selected IMF band ranges. The upper cutoff frequency of the optimal filtering band is defined as the maximum value among the upper limits of all selected IMF band ranges. , then the optimal filtering frequency band is If only a single optimal IMF is selected, its center frequency can be used as the center of the filter, and the bandwidth can be determined based on its spectral characteristics, for example, the frequency width at which the energy is attenuated to a certain percentage of the peak value, such as -6dB. Figure 3 The spectrum diagram of IMF2 and the schematic diagram of the optimized IMF.

[0027] The signal obtained by simply adding and reconstructing one or more optimal natural mode components is selected for filtering. A digital finite impulse response filter is used, for example, a window function method such as a Hamming window or a Kaiser window is used, and the passband of the filter is set to the determined optimal filtering frequency band. , obtain the stopband attenuation and transition band width, and determine the filter order. Pass the reconstructed signal through the FIR bandpass filter to obtain the filtered signal. Figure 4 The spectrum of the original or reconstructed signal is displayed with the determined optimal filtering band superimposed.

[0028] S3, performing Hilbert transform on the filtered signal to obtain an analytical signal; extracting the instantaneous amplitude of the analytical signal to obtain an envelope signal; applying cepstrum editing to the envelope signal to obtain an enhanced envelope signal, performing fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; and performing fault diagnosis based on the fault characteristic frequency in the enhanced envelope spectrum.

[0029] The filtered signal obtained from S2 , calculate its Hilbert transform Then, transform back to the time domain. Construct the analytical signal , calculate the modulus of the analytical signal, that is, the instantaneous amplitude, as the envelope signal. Calculate the power inverse spectrum of the envelope signal env(t), specifically, first calculate the autocorrelation function of the envelope signal, then perform Fourier transform on the autocorrelation function to obtain the power spectrum, then take the logarithm of the power spectrum, and finally perform inverse Fourier transform. In the obtained power inverse spectrum domain, an adaptive threshold booster is used, which first estimates the noise baseline level in the inverse spectrum, and then sets a dynamic threshold that is several times higher than the baseline, such as 3 times the standard deviation. Only those inverse frequency components in the inverse spectrum whose amplitude exceeds this dynamic threshold are retained, which usually correspond to significant periodicity, and other inverse frequency components below the threshold are set to zero or greatly attenuated, such as Figure 5 The edited power cepstrum is inversely transformed, that is, Fourier transform is first performed, then the exponential is obtained, and then inverse Fourier transform is performed to reconstruct the enhanced envelope signal.

[0030] The obtained enhanced envelope signal is first windowed using the Blackman window function before performing a fast Fourier transform to reduce spectral leakage and improve frequency resolution and amplitude accuracy. An FFT is performed on the windowed signal to obtain an enhanced envelope spectrum. The obtained enhanced envelope spectrum is matched with a pre-established fault characteristic frequency template library. This library contains the theoretical fault characteristic frequency orders or ranges of different compressor components at different speeds. The enhanced envelope spectrum is searched for significant peaks and their harmonic components that match the frequencies in the template library, and their amplitudes and signal-to-noise ratios are calculated to determine the fault type and severity. Preferably, a rule-based reasoning system is used, such as "IF spectrum peak A appears AND spectrum peak B appears THEN fault type X."

[0031] In an optional embodiment, the kurtosis value and the order cyclic spectral density amplitude are calculated for each natural mode component, specifically as follows: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

[0032] For each natural mode component , where i is the modal number and t is the time. First, calculate the time series signal The mean and standard deviation Then, the fourth-order normalized statistical moment is used to calculate its kurtosis value , kurtosis value It can reflect the amount of impact components in the signal. A larger value usually indicates a stronger impact.

[0033] Obtain a speed signal synchronized with the compressor to be diagnosed. For example, a key phase sensor (such as a photoelectric sensor or Hall effect sensor) mounted on the compressor main shaft, combined with a specific marker, can be used to directly measure the pulse signal per revolution, and then calculate the instantaneous speed. Alternatively, if a key phase sensor is not available, the instantaneous speed can be indirectly estimated from the vibration signal itself, for example by tracking the dominant frequency component through time-frequency analysis.

[0034] Since the speed of the compressor may fluctuate during actual operation, in order to accurately analyze the modulation phenomenon related to the speed, each natural mode component Perform order tracking. After order tracking (angular domain resampling), the natural mode components ,in is the turning angle, calculate its cyclic spectral density , where o represents the order, that is, the ratio of the cyclic frequency to the fundamental frequency, Represents the carrier frequency. The cyclic spectrum can be calculated by various methods, such as averaging based on short-time Fourier transform or spectrum correlation algorithm based on periodogram. According to the compressor structure and potential fault types, such as inner ring, outer ring, rolling element, cage fault of rolling bearing, gear meshing fault, etc., one or more sets of theoretical fault characteristic order values are pre-determined. Or order range. Then in the calculated two-dimensional cyclic spectrum Find the order of these preset fault characteristics The corresponding cyclic spectral density amplitude. Specifically, over the entire carrier frequency range, corresponding to a specific order The integral value, maximum value or average value of the cyclic spectral density amplitude. This is what we are looking for, which characterizes the energy intensity generated by the specific order periodic modulation in the natural mode component.

[0035] In an optional embodiment, the frequency band selection index for characterizing the fault impulse energy and periodic modulation determinism is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

[0036] Collect the kurtosis values calculated for all natural mode components , where M is the total number of modes. Normalize it, for example, use maximum value normalization so that the normalized kurtosis value falls in the interval [0,1]. Similarly, the magnitude of the order cyclic spectral density calculated for all natural mode components is calculated . It is normalized, for example, by using maximum value normalization, so that the normalized order cyclic spectrum density amplitude Also falls in the interval [0,1]. The normalized kurtosis value and the normalized order cyclic spectral density amplitude Multiply them to get the frequency band selection index corresponding to the i-th natural mode component ,index The value of also falls in the range [0,1]. If an intrinsic mode component has both high kurtosis and high order cyclic spectral density amplitude, then its corresponding frequency band selection index is will be larger, indicating that the modal component is more likely to contain fault information.

[0037] In an optional embodiment, the determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

[0038] Compare the band selection index values of all natural mode components . Select the intrinsic modal component with the largest index value As the optimized natural mode component. Alternatively, set a preset threshold, such as , iterate over the frequency band selection index of all natural mode components , all intrinsic modal components greater than a threshold are selected as a set of optimized intrinsic modal components. This approach allows the selection of multiple intrinsic modal components that may contain fault information and is suitable for situations where a fault may manifest in multiple frequency bands or multiple faults may occur simultaneously.

[0039] In an optional embodiment, determining the optimal filtering frequency band according to the center frequency and bandwidth of the optimized natural mode component is specifically as follows: Using the center frequency of the optimized natural mode component as the center frequency of the optimal filtering frequency band; The width of the region where the spectrum energy of the optimized natural mode component is mainly concentrated is used as the bandwidth of the optimal filtering frequency band.

[0040] For each natural mode component selected for optimization If multiple are selected, process each one separately, or select the one with the highest index and determine its center frequency If the signal decomposition is performed using variational mode decomposition (VMD), the VMD algorithm itself will output the center frequency of each mode, which can be used directly. If the EMD and its improved algorithms are used, the IMF obtained does not have a direct center frequency parameter. In this case, Perform fast Fourier transform to obtain its spectrum, then find the frequency point with the largest amplitude in the spectrum and use this frequency point as the Center frequency Alternatively, calculate the instantaneous frequency of the IMF and take its average or dominant value.

[0041] For VMD modes, their bandwidth characteristics are relatively concentrated and can be estimated based on VMD parameters or by analyzing their spectral shape. For EMD-type IMFs, their spectrum is not strictly bandpass. Observe their spectrum and determine the frequency region where their energy is mainly concentrated. For example, find the center frequency The two frequency points above and below the half of the peak value, that is, the -3dB or -6dB point and , then the bandwidth is defined as Alternatively, it is defined as the minimum frequency range covered when the spectrum energy reaches a predetermined percentage of the total energy, such as 90%.

[0042] The center frequency of the optimal natural modal component determined is As the center frequency of the optimal filtering band, the width of the area where the spectrum energy of the optimal natural mode component is mainly concentrated is set to As the bandwidth of the optimal filtering band. Therefore, the optimal filtering band is , or directly use its upper and lower cutoff frequencies. If multiple optimal IMFs are selected and their frequency bands do not overlap, you need to construct filters separately or select the most important one.

[0043] In an optional embodiment, applying cepstrum editing to the envelope signal to obtain the enhanced envelope signal is specifically: Calculating the real cepstrum of the envelope signal; In the real cepstrum domain, bandpass boosting is used to retain the cepstrum components associated with periodic fault impulses and to attenuate or remove the cepstrum components corresponding to broadband noise and non-periodic interference; The edited cepstrum is inversely transformed to obtain the enhanced envelope signal.

[0044] Get the envelope signal env(t) obtained in the previous step. Calculate the inverse spectrum C(q), where q is the inverse frequency. Specifically, perform Fourier transform on the envelope signal env(t) to obtain its spectrum Env(f), and take the logarithm of the spectrum amplitude . Perform inverse fast Fourier transform or discrete cosine transform on the logarithmic amplitude spectrum to obtain the real cepstrum C(q). Periodic pulse signals will appear as a series of equally spaced peaks in the cepstrum. In order to retain the cepstrum components related to periodic fault shocks in the cepstrum domain, while attenuating or removing the cepstrum components corresponding to broadband noise and non-periodic interference, a bandpass lifting method is preferably used. Specifically, based on the periodicity of the expected fault characteristic frequency, for example, the characteristic period of the bearing fault or the gear meshing period, the corresponding cepstrum range in the cepstrum domain is estimated. The range does not include the very low frequency part of the cepstrum, which may correspond to the slow trend or DC component of the signal, and the very high frequency part, which is mainly contributed by noise. A window function is determined that is 1 or close to 1 within the above-mentioned cepstrum range and is 0 or decays rapidly to 0 outside the range, i.e., a lifter. The bandpass lifting window is multiplied by the calculated real cepstrum C(q) to obtain the edited cepstrum. Then, only the components in the cepstrum located within the selected cepstrum passband are retained, while other components are suppressed.

[0045] The edited cepstrum The enhanced envelope signal is converted back to the time domain through the corresponding inverse cepstrum transform process. Compared with the original envelope signal, the periodic fault impulse component of the enhanced envelope signal is more significant, while the noise and non-periodic interference are effectively suppressed.

[0046] The second embodiment provides a device fault diagnosis system, including the following modules: An indicator determination module is configured to obtain an original vibration signal of the device to be diagnosed, perform signal decomposition on the original vibration signal to obtain a set of intrinsic modal components; calculate a kurtosis value and an order cyclic spectral density amplitude for each intrinsic modal component; and construct a frequency band selection indicator for characterizing the fault impact energy and periodic modulation determinism based on the kurtosis value and the order cyclic spectral density amplitude; a filtering module, configured to determine one or more natural mode components that optimize the frequency band selection index, and determine an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural mode components; and perform bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural mode components using the optimal filtering frequency band to obtain a filtered signal; The diagnostic module is used to perform a Hilbert transform on the filtered signal to obtain an analytical signal; extract the instantaneous amplitude of the analytical signal to obtain an envelope signal; apply cepstrum editing to the envelope signal to obtain an enhanced envelope signal; perform a fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; and perform fault diagnosis based on the fault characteristic frequency in the enhanced envelope spectrum.

[0047] Preferably, for each intrinsic modal component, the kurtosis value and the order cyclic spectral density amplitude are calculated as follows: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

[0048] Preferably, the frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

[0049] Preferably, the determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

[0050] Preferably, the determining of the optimal filtering frequency band according to the center frequency and bandwidth of the optimized natural mode component is specifically as follows: Using the center frequency of the optimized natural mode component as the center frequency of the optimal filtering frequency band; The width of the region where the spectrum energy of the optimized natural mode component is mainly concentrated is used as the bandwidth of the optimal filtering frequency band.

[0051] Preferably, the applying cepstrum editing to the envelope signal to obtain the enhanced envelope signal is specifically: Calculating the real cepstrum of the envelope signal; In the real cepstrum domain, bandpass boosting is used to retain the cepstrum components associated with periodic fault impulses and to attenuate or remove the cepstrum components corresponding to broadband noise and non-periodic interference; The edited cepstrum is inversely transformed to obtain the enhanced envelope signal.

[0052] In addition, the present invention also provides a computer device, which includes at least a memory and a processor. A computer program is stored in the memory, and when the computer program is executed by the processor, it implements the method described in the first embodiment.

[0053] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using a general-purpose hardware platform, or alternatively, through a combination of hardware and software. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a computer product. The present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it, and other embodiments may also be used. Although the present invention has been described in detail with reference to the aforementioned embodiments, ordinary technicians in this field should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein with equivalents. 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 various embodiments of the present invention.

Claims

1. A device fault diagnosis method, characterized in that: The following steps are involved: Obtaining an original vibration signal of the device to be diagnosed, and performing signal decomposition on the original vibration signal to obtain a set of natural mode components; For each natural mode component, calculate the kurtosis value and order cyclic spectral density amplitude; Based on the kurtosis value and the order cyclic spectral density amplitude, constructing a frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty; Determining one or more natural modal components that optimize the frequency band selection index, and determining an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural modal components; performing bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural modal components using the optimal filtering frequency band to obtain a filtered signal; Performing a Hilbert transform on the filtered signal to obtain an analytical signal; extracting the instantaneous amplitude of the analytical signal to obtain an envelope signal; Applying cepstrum editing to the envelope signal to obtain an enhanced envelope signal, and performing fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; Fault diagnosis is performed based on the fault characteristic frequency in the enhanced envelope spectrum.

2. The method according to claim 1, wherein For each intrinsic modal component, the kurtosis value and the order cyclic spectral density amplitude are calculated, specifically: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

3. The method according to claim 1, wherein The frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

4. The method according to claim 1, wherein The determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

5. The method according to claim 1, wherein The optimal filtering frequency band is determined according to the center frequency and bandwidth of the optimized natural mode component, specifically: Using the center frequency of the optimized natural mode component as the center frequency of the optimal filtering frequency band; The width of the region where the spectrum energy of the optimized natural mode component is concentrated is used as the bandwidth of the optimal filtering frequency band.

6. The method according to claim 1, wherein The application of cepstrum editing to the envelope signal to obtain the enhanced envelope signal is specifically: Calculating the real cepstrum of the envelope signal; In the real cepstrum domain, bandpass boosting is used to retain the cepstrum components associated with periodic fault impulses and to attenuate or remove the cepstrum components corresponding to broadband noise and non-periodic interference; The edited cepstrum is inversely transformed to obtain the enhanced envelope signal.

7. A device fault diagnosis system, characterized in that: Includes the following modules: An indicator determination module is used to obtain an original vibration signal of the device to be diagnosed, and perform signal decomposition on the original vibration signal to obtain a set of natural mode components; For each natural mode component, calculate the kurtosis value and order cyclic spectral density amplitude; Based on the kurtosis value and the order cyclic spectral density amplitude, constructing a frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty; a filtering module, configured to determine one or more natural mode components that optimize the frequency band selection index, and determine an optimal filtering frequency band based on the center frequency and bandwidth of the optimized natural mode components; and perform bandpass filtering on the original vibration signal or a signal reconstructed from the optimized natural mode components using the optimal filtering frequency band to obtain a filtered signal; A diagnostic module, configured to perform a Hilbert transform on the filtered signal to obtain an analytical signal; and extract the instantaneous amplitude of the analytical signal to obtain an envelope signal; Applying cepstrum editing to the envelope signal to obtain an enhanced envelope signal, and performing fast Fourier transform on the enhanced envelope signal to obtain an enhanced envelope spectrum; Fault diagnosis is performed based on the fault characteristic frequency in the enhanced envelope spectrum.

8. The system according to claim 7, wherein: For each intrinsic modal component, the kurtosis value and the order cyclic spectral density amplitude are calculated, specifically: For each intrinsic modal component, the fourth-order normalized statistical moment of its time series is calculated as the kurtosis value; The synchronous speed signal of the equipment is obtained, and the order tracking and cyclic spectrum analysis are performed on each natural modal component to extract the cyclic spectrum density amplitude corresponding to the preset fault feature order.

9. The system according to claim 7, wherein: The frequency band selection index for characterizing the fault impulse energy and periodic modulation certainty is constructed based on the kurtosis value and the order cyclic spectral density amplitude, specifically: Normalize the kurtosis value and order cyclic spectral density amplitude of each intrinsic mode component; The frequency band selection index is obtained by multiplying the normalized kurtosis value and the normalized order cyclic spectral density amplitude.

10. The system according to claim 7, wherein: The determining of one or more natural mode components that optimize the frequency band selection index is specifically: The natural mode component with the largest frequency band selection index value is selected as the optimized natural mode component; or the natural mode component with the frequency band selection index value exceeding a preset threshold is selected as the optimized natural mode component.

Images