A method and device for monitoring the failure of a ball flexible steel cable drawing machine

Abstract

This invention relates to a fault monitoring method and device for a flexible steel cable drawing machine, comprising the following steps: acquiring the time-domain waveform of the drawing machine's vibration, obtaining intrinsic modal components through frequency band division; calculating kurtosis values ​​and plotting the waveform, selecting peak values ​​to obtain sensitive modal components; demodulating the waveform using Hilbert spectroscopy to obtain the envelope spectrum, and extracting a set of characteristic peak frequencies. If the frequency exceeds a preset threshold, the fault type is determined and an alarm is generated. This method effectively eliminates noise and accurately identifies weak faults, solving the technical problem that conventional methods struggle to effectively remove background noise masking weak fault signals, leading to impure or offset characteristic frequencies in the extracted signals.

Description

Technical Field

[0001] This invention relates to the field of fault monitoring technology, and in particular to a fault monitoring method and device for a ball-type flexible steel cable drawing machine. Background Technology

[0002] In the field of operation and maintenance of flexible steel cable drawing machines, existing fault monitoring technologies mainly rely on the acquisition and analysis of machine vibration signals. Conventional solutions typically use accelerometers to acquire vibration data in real time, directly converting the time-domain signal into a frequency-domain signal using Fourier transform, and then extracting characteristic frequencies to determine the equipment status; or they use algorithms such as empirical mode decomposition to decompose the vibration waveform, attempting to separate components of different frequency bands to identify anomalies. These methods are widely used in the condition monitoring of various mechanical transmission equipment, aiming to provide early warning of potential mechanical faults through changes in vibration amplitude or specific frequency components, and are currently the mainstream monitoring methods in industrial sites.

[0003] However, the vibration signals generated by ball-type flexible steel cable drawing machines under high-speed, heavy-load conditions exhibit significant non-stationarity and strong noise interference characteristics, and existing technologies have obvious limitations in processing such complex signals. Due to the lack of adaptive and fine-grained frequency band division for vibration signals and a sensitive component filtering mechanism for impact characteristics, conventional methods struggle to effectively remove the masking effect of background noise on weak fault signals, resulting in extracted feature frequencies that are often impure or shifted. This insufficient signal processing precision prevents the system from accurately capturing the minute impact components caused by early faults, easily leading to errors or missed detections in fault feature frequency identification. Consequently, the accuracy of fault diagnosis is low, and reliable alarm information cannot be generated in a timely manner to guide maintenance operations. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a fault monitoring method and device for a ball-type flexible steel cable drawing machine.

[0005] The technical solution adopted in this invention is:

[0006] On one hand, embodiments of the present invention provide a fault monitoring method for a flexible steel cable drawing machine, comprising the following steps:

[0007] The vibration of the machine body during the operation of the ball flexible steel cable drawing machine is collected in real time to obtain the time-domain vibration waveform;

[0008] The time-domain vibration waveform is divided into frequency bands to obtain multiple intrinsic mode components, and the kurtosis values ​​of the multiple intrinsic mode components are calculated to obtain a kurtosis curve.

[0009] Peak filtering is performed on the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components, and Hilbert envelope demodulation is performed on the sensitive mode components to obtain envelope spectra.

[0010] Peak frequencies are extracted from the envelope spectrum to obtain a set of characteristic peak frequencies. If any fault characteristic frequency in the set of characteristic peak frequencies is not within the preset fault characteristic frequency threshold range corresponding to the drawing machine, it is determined that the ball flexible steel cable drawing machine has a corresponding type of fault state, and a fault diagnosis alarm message is generated.

[0011] Furthermore, the real-time acquisition of the machine body vibration during the operation of the ball-flexible steel cable drawing machine to obtain the time-domain vibration waveform includes:

[0012] During operation, the acceleration sensor installed in the bearing seat of the ball flexible steel cable drawing machine reads the signal to obtain the original vibration sequence, and the original vibration sequence is preprocessed by removing DC and trend terms to obtain the purified vibration waveform.

[0013] The purification vibration waveform is sampled at equal intervals using a preset sampling frequency to obtain analysis data segments. The analysis data segments are then normalized to obtain an amplitude normalization sequence. Based on the amplitude normalization sequence, the waveform is plotted to obtain a time-domain vibration waveform.

[0014] Furthermore, the frequency band division of the time-domain vibration waveform yields multiple intrinsic mode components, including:

[0015] Based on a preset number of modes, the time-domain vibration waveform is iteratively searched for the center frequency in the frequency domain to obtain multiple center frequency values. Then, Wiener filtering is performed on the time-domain vibration waveform based on the multiple center frequency values ​​to obtain an initial modal component set.

[0016] For each initial modal component in the initial modal component set, the frequency band overlap is checked. When there is frequency band overlap between adjacent modes, the preset number of modes is adjusted and the center frequency iterative search is performed again until the frequency band separation condition is met, resulting in multiple independent modal components.

[0017] The plurality of independent modal components are used as the plurality of intrinsic modal components, wherein the plurality of independent modal components are used to display a pure impulse response waveform with non-overlapping frequency bands.

[0018] Furthermore, the step of calculating the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve includes:

[0019] For each of the multiple intrinsic mode components, the amplitude probability distribution is statistically analyzed to obtain an amplitude distribution histogram. The ratio of the fourth moment to the second moment of the amplitude distribution histogram is calculated to obtain a single kurtosis value.

[0020] Based on the arrangement order of the multiple intrinsic mode components, the single kurtosis value of each intrinsic mode component is sequentially mapped onto the frequency band coordinate axis to obtain discrete kurtosis points, and curve fitting is performed on the discrete kurtosis points to obtain an initial kurtosis curve.

[0021] Local maxima are identified on the initial kurtosis curve to obtain the curve peak points, and peak marks are made on the initial kurtosis curve based on the curve peak points to obtain a kurtosis curve graph.

[0022] Furthermore, the step of peak filtering of the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components includes:

[0023] The peak coordinates of the kurtosis curve are read to obtain the peak frequency band position, and the multiple intrinsic mode components are back-indexed according to the peak frequency band position to obtain candidate mode components;

[0024] The candidate modal components are compared with kurtosis thresholds. Candidate modal components with kurtosis values ​​lower than a preset screening threshold are marked as noise modes and removed to obtain the retained modal components.

[0025] The correlation between the retained modal components is calculated to obtain a modal correlation coefficient matrix. Based on the modal correlation coefficient matrix, the retained modal components are redundantly merged. Multiple retained modal components with correlation exceeding a preset merging threshold are merged into a representative mode to obtain a sensitive modal component.

[0026] Furthermore, the step of performing Hilbert envelope demodulation on the sensitive mode components to obtain the envelope spectrum includes:

[0027] The sensitive mode components are subjected to Hilbert transform to obtain orthogonal phase components. A complex analytic signal is constructed based on the sensitive mode components and the orthogonal phase components. The amplitude of the complex analytic signal is calculated to obtain the amplitude envelope waveform.

[0028] The amplitude envelope waveform is subjected to a fast Fourier transform to obtain the envelope spectrum. The amplitude of the envelope spectrum is then normalized to obtain a normalized envelope spectrum. Based on the normalized envelope spectrum, a spectrum is constructed to obtain an envelope spectrum diagram.

[0029] The present invention also provides a fault monitoring device for a flexible steel cable drawing machine, comprising:

[0030] The acquisition module is used to acquire the vibration of the machine body in real time during the operation of the ball flexible steel cable drawing machine, and obtain the time-domain vibration waveform;

[0031] The partitioning module is used to partition the time-domain vibration waveform into frequency bands to obtain multiple intrinsic mode components, and to calculate the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve.

[0032] The filtering module is used to perform peak filtering on the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components, and to perform Hilbert envelope demodulation on the sensitive mode components to obtain an envelope spectrum.

[0033] The extraction module is used to extract the peak frequency from the envelope spectrum to obtain a set of characteristic peak frequencies. If any fault characteristic frequency in the set of characteristic peak frequencies is not within the preset fault characteristic frequency threshold range corresponding to the drawing machine, it is determined that the ball flexible steel cable drawing machine has a corresponding type of fault state and a fault diagnosis alarm message is generated.

[0034] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above methods.

[0035] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the above methods.

[0036] This invention provides a fault monitoring method for a flexible steel cable drawing machine, comprising the following steps: real-time acquisition of machine body vibration during operation to obtain a time-domain vibration waveform; frequency band division of the time-domain vibration waveform to obtain multiple intrinsic modal components, and kurtosis calculation of the multiple intrinsic modal components to obtain a kurtosis curve; peak filtering of the multiple intrinsic modal components based on the kurtosis curve to obtain sensitive modal components, and Hilbert envelope demodulation of the sensitive modal components to obtain an envelope spectrum; and further processing of the envelope spectrum. Peak frequency extraction is performed to obtain a set of characteristic peak frequencies. If any fault characteristic frequency in the set of characteristic peak frequencies is not within the preset fault characteristic frequency threshold range corresponding to the drawing machine, the ball flexible steel cable drawing machine is determined to have a corresponding type of fault state, and a fault diagnosis alarm is generated. This solves the technical problem that conventional methods are difficult to effectively remove the masking of weak fault signals by background noise, resulting in the extracted characteristic frequencies often being impure or shifted. In this embodiment, accurate extraction and state identification of weak fault characteristics under complex working conditions of the ball flexible steel cable drawing machine are achieved. Specifically, the non-stationary time-domain vibration waveform is decomposed into multiple intrinsic mode components by using frequency band division, which effectively overcomes the limitations of traditional Fourier transform in processing nonlinear and non-stationary signals and avoids the dispersion of fault impact energy in a wide frequency band. Furthermore, kurtosis calculation is introduced as a statistical measure to quantify the impulse characteristics of each component. Peak filtering based on kurtosis curves can automatically lock and separate the sensitive mode components containing the richest fault impact information, eliminating a large amount of low-energy background noise and normal operation interference without fault characteristics from the source, solving the problem of noise-induced distortion in conventional methods. This approach addresses the challenge of masking features, which can lead to impure or shifted feature frequency extraction. By applying Hilbert envelope demodulation to the sensitive modal components, high-frequency carrier oscillations are successfully removed, transforming the periodic impact hidden in the modulated signal into a clear low-frequency envelope trajectory. This trajectory is then mapped to a high signal-to-noise ratio envelope spectrum via Fast Fourier Transform, highlighting the weak fault feature frequencies that were previously submerged in strong background noise. Finally, by logically comparing the extracted feature peak frequency set with a preset threshold range for the fault feature frequencies of the drawing machine, automated qualitative judgment of fault types and alarm generation are achieved. This solution not only significantly improves the accuracy and anti-interference capability of fault feature frequency identification, ensuring the detectability of early weak faults (such as early bearing spalling and gear micro-pitting) in strong noise environments, but also effectively avoids false alarms or missed alarms caused by feature frequency shifts. It greatly improves the reliability, real-time performance, and engineering practicality of fault diagnosis for ball-and-wire flexible steel cable drawing machines, providing solid data support and decision-making basis for predictive maintenance of equipment. Attached Figure Description

[0037] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0038] Figure 1 This is a flowchart illustrating the steps of the fault monitoring method for a flexible steel cable drawing machine in an embodiment of the present invention.

[0039] Figure 2 This is a structural block diagram of a fault monitoring device for a flexible steel cable drawing machine according to an embodiment of the present invention;

[0040] Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0041] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0042] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0043] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0044] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0045] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0046] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0047] Reference Figure 1This invention provides a fault monitoring method for a flexible steel cable drawing machine, comprising the following steps:

[0048] Step S1: Real-time acquisition of the vibration of the ball flexible steel cable drawing machine during operation to obtain the time-domain vibration waveform.

[0049] Specifically, a high-frequency accelerometer is used to continuously capture simulated mechanical vibration signals at a fixed sampling frequency by closely monitoring key stress points on the body of the flexible steel cable drawing machine. These signals are then quantized by an analog-to-digital converter to directly generate a discrete time-domain vibration waveform data stream. This process does not simply record amplitude values; it strictly synchronizes with the equipment's operating clock to ensure that the time-domain vibration waveform retains its impact transient characteristics even during sudden changes in drawing speed or load fluctuations. For example, when the steel cable vibrates periodically as it passes through the die, the voltage pulse sequence at the corresponding moment in the waveform is losslessly locked and stored. This real-time capture and digital mapping at the hardware level avoids data loss caused by signal transmission delays, ensuring the integrity of the original vibration information. In this embodiment, the above scheme, through high-fidelity acquisition of the time-domain vibration waveform, provides an undistorted raw data foundation for subsequent frequency band division and kurtosis calculation, effectively eliminating spectral aliasing errors caused by asynchronous sampling, thereby significantly improving the signal-to-noise ratio and identification accuracy of fault feature extraction.

[0050] Step S2: Divide the time-domain vibration waveform into frequency bands to obtain multiple intrinsic mode components, and calculate the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve.

[0051] Specifically, following the previously acquired time-domain vibration waveform, the system first employs a variational mode decomposition algorithm to adaptively decompose the signal into several narrowband sequences based on a preset center frequency, thereby obtaining multiple intrinsic mode components. This process avoids the endpoint effect of traditional empirical mode decomposition. Subsequently, for each intrinsic mode component, the fourth moment of its amplitude distribution is statistically analyzed to quantify the impact intensity, i.e., kurtosis value calculation is performed. For example, when early spalling occurs in the bearing of a drawing machine, the corresponding high-frequency component will exhibit a large-amplitude pulse, and its calculated kurtosis value is significantly higher than that of the stationary background noise component. By mapping all component indices as the abscissa and the corresponding kurtosis values ​​as the ordinate, a kurtosis curve that intuitively reflects the impact characteristics of each component can be obtained. In this embodiment, the above scheme effectively separates the weak fault impact components in a strong noise background by combining fine frequency band division with kurtosis quantitative characterization. The kurtosis curve is used to quickly locate the mode containing the richest fault information, providing a reliable quantitative basis for the accurate screening of subsequent sensitive components and solving the problem of difficulty in extracting fault features under complex working conditions.

[0052] Step S3: Based on the kurtosis curve, peak filtering is performed on the multiple intrinsic mode components to obtain sensitive mode components, and Hilbert envelope demodulation is performed on the sensitive mode components to obtain envelope spectra.

[0053] Specifically, based on the kurtosis curve, the index corresponding to the numerical maximum point is located, and the specific component is extracted from the multiple intrinsic mode components as the sensitive mode component. This operation aims to lock the frequency band where the impact energy is most concentrated. For example, in the monitoring of a flexible steel cable pulling machine, only the high-frequency component containing bearing spalling impact is retained while low-frequency mechanical noise is filtered out. Then, a Hilbert transform is performed on the sensitive mode component to construct a complex analytic signal, and its modulus is calculated to obtain the instantaneous amplitude, i.e., the envelope waveform. Then, a fast Fourier transform is performed to map the time-domain pulse sequence to the frequency-domain spectrum, and finally the envelope spectrum is obtained. In this embodiment, the above scheme accurately removes irrelevant noise modes through kurtosis statistics and uses demodulation technology to transform the weak periodic impacts in the high-frequency carrier into visualized characteristic spectral peaks, significantly improving the signal-to-noise ratio and identification accuracy of the fault frequency under strong background interference.

[0054] Step S4: Extract peak frequencies from the envelope spectrum to obtain a set of characteristic peak frequencies. If any fault characteristic frequency in the set of characteristic peak frequencies is not within the preset fault characteristic frequency threshold range corresponding to the drawing machine, it is determined that the ball flexible steel cable drawing machine has a corresponding type of fault state, and a fault diagnosis alarm message is generated.

[0055] Specifically, based on the previously generated envelope spectrum, the system first executes a peak search algorithm to identify local maxima in the spectrum whose amplitude exceeds a preset noise threshold. The corresponding frequency coordinates are extracted and summarized to obtain a set of characteristic peak frequencies. Subsequently, the values ​​in this set are compared one by one with the pre-stored theoretical fault characteristic frequency threshold ranges for key components of the drawing machine, such as bearings and gears. For example, if an extracted frequency value is 125Hz, and the preset bearing outer ring fault frequency threshold is 120Hz±2Hz, the value falls within this range and is considered normal. Conversely, if the extracted value is 150Hz and no preset threshold covers this range, or if a specific theoretical frequency is missing, the logic determines that the ball-and-wire steel cable drawing machine has a corresponding type of fault. Once the determination is successful, the control unit immediately triggers an interrupt program, generating a fault diagnosis alarm message containing the fault type and occurrence time, and pushes it to the monitoring terminal. In this embodiment, the above scheme, through an automated frequency matching and threshold verification mechanism, eliminates the subjective error of manual spectrum interpretation, achieving quantitative identification and immediate alarm of early, minor faults in the drawing machine, significantly reducing the risk of equipment damage due to missed detections.

[0056] In a specific embodiment, the real-time acquisition of the machine body vibration during the operation of the ball-flexible steel cable drawing machine to obtain the time-domain vibration waveform includes:

[0057] During operation, the acceleration sensor installed in the bearing seat of the ball flexible steel cable drawing machine reads the signal to obtain the original vibration sequence, and the original vibration sequence is preprocessed by removing DC and trend terms to obtain the purified vibration waveform.

[0058] The purification vibration waveform is sampled at equal intervals using a preset sampling frequency to obtain analysis data segments. The analysis data segments are then normalized to obtain an amplitude normalization sequence. Based on the amplitude normalization sequence, the waveform is plotted to obtain a time-domain vibration waveform.

[0059] Specifically, the sensing front end is first constructed using an accelerometer installed in the bearing housing of the ball-and-wire steel cable drawing machine. This sensor is rigidly coupled to the outer ring of the bearing via a magnetic base or bolts, directly sensing the transient impacts and continuous vibrations generated by the mechanical structure during the drawing operation. The data acquisition card drives the analog-to-digital converter interface with a high-frequency clock to continuously scan and read the charge or voltage signals output by the sensor, thereby obtaining the original vibration sequence containing a large amount of environmental noise and device zero drift. Given that the sensor's own temperature drift effect and installation stress may introduce non-zero mean DC components, and that thermal deformation during low-speed operation of the equipment will cause the signal baseline to drift slowly, forming a trend term, direct analysis would severely mask weak fault characteristics. Therefore, the original vibration sequence must undergo DC removal and trend term preprocessing. Operationally, the arithmetic mean of the sequence is first calculated and subtracted point by point to eliminate DC bias. Then, a polynomial fitting or wavelet decomposition algorithm is used to extract the low-frequency trend curve, which is then removed from the signal, finally obtaining a purified vibration waveform that retains only dynamic vibration information. This step effectively eliminates baseline tilt interference caused by temperature changes.

[0060] After acquiring the purification vibration waveform, to meet the periodicity assumption of frequency domain analysis and reduce spectral leakage, the purification vibration waveform needs to be truncated at equal intervals using a preset sampling frequency. For example, setting the sampling rate to 25.6kHz, according to the Nyquist sampling theorem, a continuous data segment with a duration of 1 second is selected as the analysis data segment to ensure coverage of at least several complete equipment rotation cycles and potential fault impact repetition cycles. Considering that the vibration amplitude of the drawing machine fluctuates drastically under different load conditions, directly comparing waveforms at different times is easily misled by differences in energy magnitude. Therefore, amplitude normalization processing is then performed on the analysis data segment. That is, the maximum absolute value in the sequence is found, and all data points in the sequence are divided by this extreme value, so that the signal amplitude is mapped to the interval [-1, 1], thereby obtaining an amplitude normalized sequence that eliminates the influence of dimensions. Finally, with time as the horizontal axis and normalized amplitude as the vertical axis, the graphics engine is called to draw the waveform based on the amplitude normalized sequence, intuitively presenting the fluctuation pattern of vibration over time, and obtaining a standard time-domain vibration waveform. During this process, if the wear of the drawing machine die causes periodic impacts, after the above processing, the time-domain vibration waveform will clearly show equally spaced pulse spikes without visual distortion caused by baseline drift or amplitude scaling. In this embodiment, the above scheme, through progressive signal conditioning and standardization, not only eliminates the interference of sensor zero drift, temperature drift, and changes in operating load on the vibration signal, but also ensures the consistency and purity of the input data for subsequent feature extraction algorithms, significantly improving the identifiability of fault features in the time-domain waveform, and laying a high signal-to-noise ratio data foundation for subsequent frequency band division and kurtosis calculation.

[0061] In a specific embodiment, the step of dividing the time-domain vibration waveform into frequency bands to obtain multiple intrinsic mode components includes:

[0062] Based on a preset number of modes, the time-domain vibration waveform is iteratively searched for the center frequency in the frequency domain to obtain multiple center frequency values. Then, Wiener filtering is performed on the time-domain vibration waveform based on the multiple center frequency values ​​to obtain an initial modal component set.

[0063] For each initial modal component in the initial modal component set, the frequency band overlap is checked. When there is frequency band overlap between adjacent modes, the preset number of modes is adjusted and the center frequency iterative search is performed again until the frequency band separation condition is met, resulting in multiple independent modal components.

[0064] The multiple independent modal components are referred to as the multiple intrinsic modal components.

[0065] Specifically, firstly, based on prior knowledge or spectral pre-analysis results of the vibration characteristics of the ball-and-flexible steel cable drawing machine, an initial preset number of modes is set. This value determines how many independent narrowband components the complex vibration signal is expected to be decomposed into. Subsequently, the system constructs a variational constraint model and performs an iterative search for the center frequency of the time-domain vibration waveform in the frequency domain based on the preset number of modes. This process is essentially finding the optimal solution in the frequency domain space that minimizes the sum of the bandwidths of each mode. The center position of each component is continuously corrected through the alternating direction multiplier method until convergence, thereby obtaining multiple center frequency values. These frequency values ​​precisely correspond to several frequency bands in the signal where the energy is most concentrated. Next, a series of adaptive Wiener filter banks are constructed using the obtained center frequency values, and the time-domain vibration waveform is subjected to Wiener filtering based on the multiple center frequency values. This filter is a Gaussian window function with the center frequency as the peak value in the frequency domain, which can effectively suppress out-of-band noise and extract signals in specific frequency bands, thereby obtaining an initial modal component set. The initial modal component set is used to display the vibration waveforms of different frequency bands that have been initially separated. At this time, although the components have been initially separated, due to the coarseness of the initial parameter setting, there may be spectral aliasing.

[0066] To ensure the rigor of the decomposition, the frequency band overlap of each initial modal component in the initial modal component set must be checked. Specifically, the power spectral density curves of two adjacent modal components are calculated, and the area of ​​their spectral overlap region is quantified. If this area exceeds a preset threshold (e.g., 5%), it is determined that there is frequency band overlap between adjacent modes, indicating that the initially set number of modes is insufficient to distinguish closely adjacent fault characteristic frequencies or mechanical resonance peaks. For example, when the fault frequency of the inner ring of the traction wheel bearing of a drawing machine is extremely close to the gear meshing frequency, if the preset number of modes is too small, the two may be forcibly merged into a single wideband component, resulting in blurred fault characteristics. Once overlap is detected, the system immediately adjusts the preset number of modes, typically by increasing the number of modes to refine the frequency band resolution, and re-performs the center frequency iterative search, re-executes the Wiener filtering and verification process, forming a closed-loop optimization mechanism, until the frequency band separation condition is met, i.e., the spectral overlap of all adjacent modes is below the allowable threshold. At this point, multiple independent modal components are obtained, which do not interfere with each other in the frequency domain and each carries a single physical vibration mode. Finally, the multiple independent modal components are used as the multiple intrinsic modal components. These multiple independent modal components are used to display pure impact response waveforms with non-overlapping frequency bands, ensuring the accuracy of subsequent kurtosis calculations and avoiding false impact features caused by mode aliasing. In this embodiment, the above scheme overcomes the limitations of traditional fixed parameter decomposition methods in processing non-stationary, strongly coupled vibration signals by introducing a frequency band overlap verification and dynamic iteration mechanism. It achieves adaptive fine frequency band division for complex working conditions of ball-flexible steel cable drawing machines, effectively solving the problem of weak fault impacts being submerged by strong background noise or adjacent frequency components. This provides a high-purity, non-aliased modal data foundation for subsequent accurate extraction of fault feature frequencies, significantly improving the robustness and early warning capability of the fault diagnosis system.

[0067] In a specific embodiment, the step of calculating the kurtosis values ​​of the plurality of intrinsic mode components to obtain a kurtosis curve includes:

[0068] For each of the multiple intrinsic mode components, the amplitude probability distribution is statistically analyzed to obtain an amplitude distribution histogram. The ratio of the fourth moment to the second moment of the amplitude distribution histogram is calculated to obtain a single kurtosis value.

[0069] Based on the arrangement order of the multiple intrinsic mode components, the single kurtosis value of each intrinsic mode component is sequentially mapped onto the frequency band coordinate axis to obtain discrete kurtosis points, and curve fitting is performed on the discrete kurtosis points to obtain an initial kurtosis curve.

[0070] Local maxima are identified on the initial kurtosis curve to obtain the curve peak points, and peak marks are made on the initial kurtosis curve based on the curve peak points to obtain a kurtosis curve graph.

[0071] Specifically, the system first iterates through the multiple intrinsic mode components obtained from the decoupling steps. For each intrinsic mode component, the system constructs an amplitude probability density function to quantify its statistical characteristics. Operationally, this involves accumulating and statistically analyzing the frequency of amplitudes falling into different intervals at all sampling points of the component, thereby obtaining an amplitude distribution histogram. This histogram intuitively reflects the aggregation pattern of vibration energy in the amplitude domain. Next, based on the statistical definition of kurtosis, the ratio of the fourth moment to the second moment is calculated on the amplitude distribution histogram; that is, the signal is first calculated... The mean of the fourth power of the amplitude is taken as the fourth central moment, and the square of the mean of the square of the second power of the signal amplitude is taken as the square of the second central moment. The two are divided to obtain the single kurtosis value. The single kurtosis value is used to quantify the kurtosis of the corresponding modal component waveform. If a transient impact caused by the breakage of the steel cable of the drawing machine is mixed in a certain modal component, its amplitude distribution will show a long tail feature, which will cause the fourth moment to increase sharply, and thus make the calculated kurtosis value much higher than 3.0 of the normal distribution background. Conversely, if it is only stationary noise, the value will be close to 3.

[0072] After obtaining the statistical indicators of all components, based on the arrangement order of the multiple intrinsic modal components, typically arranged sequentially from low to high center frequency or by decomposition number, the individual kurtosis value of each intrinsic modal component is sequentially mapped onto the frequency band coordinate axis. The horizontal axis represents the center frequency or frequency band number corresponding to the mode, and the vertical axis represents the kurtosis magnitude, thus obtaining discrete kurtosis points. These points exhibit a non-uniform distribution in the coordinate system. Considering that discrete points cannot intuitively reflect the continuous law of impact energy change with frequency, curve fitting is required for the discrete kurtosis points. Cubic spline interpolation or Gaussian kernel smoothing algorithms are used to connect the discrete points to eliminate the minor jitter caused by data truncation, resulting in an initial kurtosis curve. The initial kurtosis curve is used to show the continuous trend of impact intensity change with frequency band, clearly revealing which frequency band contains the most significant fault impact component. For example, in the operation of a ball-and-wire drawing machine, if the impact generated by die wear mainly excites high-frequency resonance, the initial kurtosis curve will show a significant bulge in the high-frequency band, while remaining flat in the low-frequency band.

[0073] To further pinpoint the fault source, the system needs to identify local maxima of the initial kurtosis curve. By searching for points where the first derivative is zero and the second derivative is less than zero, the system accurately captures the peak positions on the curve, obtaining the peak points. Based on these peak points, the initial kurtosis curve is marked with peaks, typically by adding arrows, highlighted blocks, or numerical labels at the peaks, ultimately resulting in a kurtosis curve graph. This graph displays the distribution of impact characteristic intensity at the marked candidate fault frequency band locations. Maintenance personnel can directly deduce the corresponding intrinsic mode components based on the marked peak positions in the graph, using them as the optimal input for subsequent envelope analysis. In this embodiment, the above-mentioned scheme transforms the complex time-domain impact characteristics into a kurtosis peak distribution in the frequency domain. By utilizing the high sensitivity of the fourth-order moment to pulse signals, it effectively overcomes the shortcomings of traditional spectrum analysis in identifying weak early faults in the context of strong noise. It realizes the transformation from "blind search across the entire frequency band" to "precise positioning in sensitive frequency bands". This not only significantly reduces the amount of data processing, but also significantly improves the accuracy and timeliness of fault diagnosis for ball flexible steel cable drawing machines, providing an intuitive quantitative basis for preventive maintenance.

[0074] In a specific embodiment, the step of calculating the ratio of the fourth moment to the second moment of the amplitude distribution histogram to obtain a single kurtosis value includes:

[0075] The amplitude distribution histogram is divided into amplitude intervals to obtain multiple amplitude intervals, and the number of sample points in the multiple amplitude intervals is counted to obtain an interval counting sequence;

[0076] The expected value of the amplitude distribution histogram is calculated based on the interval counting sequence to obtain the distribution mean. The fourth and second central moments are then calculated based on the distribution mean to obtain the fourth and second moment values.

[0077] The initial kurtosis number is obtained by ratioing the squares of the fourth-order moment value and the second-order moment value, and then the initial kurtosis number is normalized to obtain a single kurtosis value.

[0078] Specifically, firstly, a discretization and quantization operation is performed on the amplitude distribution histogram generated in the previous step. A fixed step width is set according to the dynamic range of the vibration signal, and the continuous amplitude axis is cut into several sub-regions of equal or unequal width, thereby obtaining multiple amplitude intervals. This step is essentially the process of discretizing the continuous probability density function into histogram columnar units. Subsequently, the system traverses all sampling points of the intrinsic modal components obtained in the previous step, determines which amplitude interval the amplitude of each sampling point falls into, and accumulates the counts of the sample points falling into the same interval, finally obtaining the interval counting sequence. The interval counting sequence is used to show the frequency of vibration amplitude occurrence in each amplitude interval. This sequence constitutes the basic data source for subsequent moment calculation. If the drawing machine runs smoothly, the interval counting sequence will show a bell-shaped distribution with high values ​​in the middle and low values ​​at both ends. However, if there is an impact fault, the count values ​​of the large intervals at both ends will be abnormally high.

[0079] After obtaining the frequency statistics results, the expected value of the amplitude distribution histogram needs to be calculated based on the interval counting sequence. Specifically, the center value of each amplitude interval is multiplied by the count value of that interval and summed, then divided by the total number of samples to obtain the distribution mean. This mean represents the statistical equilibrium position of the vibration signal. Next, using this distribution mean as a benchmark, the fourth and second central moments of the interval counting sequence are calculated. Operationally, the square of the difference between the center value of each interval and the distribution mean is calculated first. After weighted summation, the second moment value, i.e., variance, is obtained, which reflects the dispersion of signal energy. Then, the fourth power of the difference between the center value of each interval and the distribution mean is calculated, and the fourth moment value is obtained by weighted summation. The fourth moment value is extremely sensitive to extreme large amplitudes far from the mean and can capture transient impact peaks hidden in background noise. For example, when the bearing of the traction wheel of the ball flexible steel cable drawing machine peels off, the periodic impact will cause the number of large amplitude sampling points to increase, resulting in an exponential increase in the fourth moment value, while the change in the second moment value is relatively gradual.

[0080] Subsequently, the ratio of the fourth-order moment value to the square of the second-order moment value is calculated, i.e., the fourth-order moment value is divided by the second-order moment value raised to the power of two. This eliminates the influence of dimensions and standardizes the energy level, yielding the initial kurtosis number. This value directly reflects the sharpness of the waveform relative to the Gaussian distribution. Considering the scale differences that may be introduced by different sensor gains or signal conditioning circuits, the initial kurtosis number also needs to be normalized. This is usually done by subtracting a reference value (such as 3.0, representing the theoretical kurtosis of Gaussian white noise) or dividing by the maximum theoretical threshold, mapping it to a standard range, and finally obtaining a single kurtosis value. This process ensures the comparability of calculation results under different operating conditions. In this embodiment, the above-mentioned scheme avoids numerical overflow errors that may be caused by directly performing high-order operations on the original time-domain data through a refined calculation strategy based on histogram statistical moments. It effectively smooths out quantization noise interference by using interval counting sequences. At the same time, through the nonlinear ratio amplification mechanism of the fourth moment and the second moment, it greatly enhances the sensitivity to the early weak fault impact characteristics of the ball flexible steel cable drawing machine. This allows for the accurate extraction of a single kurtosis value that characterizes the severity of the fault, even in environments with strong background noise. This provides a high-confidence quantitative criterion for the automatic identification of subsequent fault frequency bands.

[0081] In a specific embodiment, the step of peak filtering of the plurality of intrinsic mode components based on the kurtosis curve to obtain sensitive mode components includes:

[0082] The peak coordinates of the kurtosis curve are read to obtain the peak frequency band position, and the multiple intrinsic mode components are back-indexed according to the peak frequency band position to obtain candidate mode components;

[0083] The candidate modal components are compared with kurtosis thresholds. Candidate modal components with kurtosis values ​​lower than a preset screening threshold are marked as noise modes and removed to obtain the retained modal components.

[0084] The correlation between the retained modal components is calculated to obtain a modal correlation coefficient matrix. Based on the modal correlation coefficient matrix, the retained modal components are redundantly merged. Multiple retained modal components with correlation exceeding a preset merging threshold are merged into a representative mode to obtain a sensitive modal component.

[0085] Specifically, firstly, relying on the previously generated kurtosis curve as the screening criterion, the system automatically traverses all peaks on the curve, reads the peak coordinates of the kurtosis curve, and accurately extracts the abscissa value corresponding to the local maximum value of the ordinate, thus obtaining the peak frequency band position. This position directly maps the frequency range where the vibration energy is most concentrated and the impact characteristics are most significant. Then, based on the one-to-one correspondence between frequency and mode number, the multiple intrinsic mode components are back-indexed according to the peak frequency band position, and one or more components located within this frequency band are retrieved from the original decomposition results, thereby obtaining candidate mode components. This step realizes rapid positioning from frequency domain features to time domain signals. However, selecting based solely on peak position may lead to false peaks caused by random noise. Therefore, it is necessary to compare the kurtosis thresholds of the candidate modal components. A preset screening threshold based on statistical experience or background noise level (e.g., 4.5) is set, and the actual kurtosis value of each candidate component is compared one by one. If the kurtosis value of a component is lower than the preset screening threshold, it is determined that although it is in the peak frequency band, the impact intensity is insufficient, and it is very likely an accidental aggregation of broadband noise. Therefore, it is marked as a noise mode and eliminated. Only those components with kurtosis values ​​significantly higher than the threshold are retained as retained modal components. The retained modal components are used to display the high-confidence impact waveform after passing the energy threshold screening, ensuring the effectiveness of subsequent analysis objects. For example, in the operation of a ball-flexible steel cable drawing machine, if a small bulge appears in a certain high-frequency band, but the kurtosis value of its corresponding component is only 3.2, close to the Gaussian distribution characteristics, then the component will be decisively discarded to avoid interfering with fault judgment.

[0086] After obtaining the high-confidence retained modal components, considering that algorithms such as variational mode decomposition often suffer from mode aliasing or over-decomposition when processing complex non-stationary signals, leading to the splitting of features of the same fault source into components in two or more adjacent frequency bands, resulting in information redundancy and analysis dispersion, it is necessary to perform correlation calculation on the retained modal components. Specifically, this involves calculating the Pearson correlation coefficient or the maximum value of the cross-correlation function between any two retained components, constructing a symmetric modal correlation coefficient matrix. This matrix quantifies the similarity of the waveforms of each component in the time domain. Then, based on the modal correlation coefficient matrix, the retained modal components are analyzed... Redundancy merging of modal components is performed by setting a preset merging threshold (e.g., 0.85). The element pairs in the scanning matrix that exceed this threshold are scanned. If the correlation between two or more remaining modal components exceeds the preset merging threshold, it indicates that they carry highly repetitive fault information and originate from the same physical excitation source. At this time, these components are linearly superimposed or weighted averaged and merged into a representative mode, thereby eliminating redundant calculations and concentrating fault energy, and finally obtaining the sensitive modal component. The sensitive modal component is used to display the only core fault analysis object after redundancy is eliminated. This component completely retains the temporal details of the fault impact and has no spectral leakage interference. In this embodiment, the above-mentioned scheme, through a three-level progressive screening mechanism of "peak location - threshold filtering - correlation deduplication", not only effectively avoids the risk of false alarms caused by random noise, but also cleverly solves the problem of feature fragmentation caused by over-decomposition of modes. It re-aggregates weak fault impacts scattered in multiple frequency bands, significantly improves the signal-to-noise ratio and identification of fault features of ball flexible steel cable drawing machine, and provides a single, pure and complete key data input for subsequent envelope spectrum analysis and fault type diagnosis, greatly reducing the computational load and false judgment rate of the diagnostic system.

[0087] In a specific embodiment, the step of inverting the plurality of intrinsic mode components based on the peak frequency band position to obtain candidate mode components includes:

[0088] Based on the peak frequency band position, the center frequencies of the multiple intrinsic mode components are matched to obtain the matched mode number, and waveform data is extracted from the multiple intrinsic mode components according to the matched mode number to obtain the initial selected mode waveform set.

[0089] Perform time-domain envelope calculation on each initial modal waveform in the initial modal waveform set to obtain the waveform envelope line, and mark the peak start and end times of the waveform envelope line to obtain the envelope start and end time pair;

[0090] Based on the start and end times of the envelope, the initial selected modal waveform set is truncated to obtain multiple impact segment waveforms. The waveform morphology of the multiple impact segment waveforms is compared, and impact segment waveforms with similar morphology are grouped into the same impact category to obtain candidate modal components.

[0091] Specifically, firstly, a mapping bridge between frequency domain features and time domain signals is established. Based on the peak frequency band position, frequency band matching is performed on the center frequencies of the multiple intrinsic mode components. This process involves calculating the Euclidean distance between the peak frequency band center value and the center frequency of each mode, selecting the mode number with the smallest distance that falls within a preset tolerance range, thereby obtaining the matched mode number. Then, waveform data is extracted from the multiple intrinsic mode components according to the matched mode number, and the complete time series data of the corresponding number is loaded into memory to form a preliminary selected mode waveform set. The preliminary selected mode waveform set is used to display the original vibration waveform segments corresponding to the peak frequency band of the kurtosis curve. At this time, the data still contains background noise and non-fault-related random fluctuations. To further remove invalid information and pinpoint the actual mechanical impact event, time-domain envelope calculations are performed on each initial modal waveform in the initial modal waveform set. Typically, Hilbert transform is used to construct an analytical signal and extract the modulus value, or rectification followed by low-pass filtering is used to precisely outline the slow change trend of the signal amplitude, resulting in a waveform envelope. This envelope eliminates the oscillation details of the high-frequency carrier, retaining only the energy fluctuation profile. Next, the peak start and end times of the waveform envelope are marked. A dynamic threshold (e.g., 3 times the envelope mean) or a slope abrupt change detection algorithm is set to identify the moment when the rising edge of the envelope exceeds the threshold as the starting point and the moment when the falling edge falls back below the threshold as the ending point, thus obtaining an envelope start and end time pair. This pair is used to show the start and end positions of a single impact event on the time-domain waveform. For example, in a flexible steel cable drawing machine, when a steel wire rope experiences a transient impact as it passes through a wear die, its envelope will exhibit a distinct single-peak spike. The start and end time pair precisely defines the duration of this impact process.

[0092] Subsequently, based on the start and end times of the envelope, impact segments are extracted from the initially selected modal waveform set. Each pair of start and end times is used as a clipping window to cut independent short-time signal segments from the original waveform, resulting in multiple impact segment waveforms. Each of these segments represents a potential mechanical impact event. Given that actual working conditions involve occasional external interference (such as material falling and impacting the casing) or non-periodic noise, their waveform morphology often differs significantly from regular fault impacts (such as bearing balls passing through a spalling point). Therefore, it is necessary to compare the waveform morphology of the multiple impact segment waveforms using dynamic time... The Dynamic Normalization (DTW) algorithm calculates the similarity distance between any two segments, or extracts the skewness, kurtosis, and dominant frequency components of the waveform to form a feature vector for cluster analysis. Waveforms of similar morphology are grouped into the same impact category, and abnormal segments that exist in isolation, have morphological distortion, or are too far from the main category are eliminated. Finally, the most representative waveform set is selected to obtain candidate modal components. The candidate modal components are used to display the typical fault impact waveform set after eliminating occasional interference, ensuring that the subsequent analysis object purely reflects the internal mechanical fault of the equipment rather than external environmental noise. In this embodiment, the above-mentioned scheme, through a progressive strategy of "frequency band positioning - envelope segmentation - morphological clustering", not only achieves refined extraction from broadband signals to single fault impact events, but also effectively overcomes the misdiagnosis problem caused by the inability of traditional methods to distinguish between regular fault impacts and random interference. Especially in the complex environment of ball flexible steel cable drawing machine with strong noise and multiple impact sources, it can accurately separate typical waveforms that characterize the core fault features, greatly improving the purity and reliability of fault feature extraction, and providing high-fidelity data samples for subsequent fault mode recognition and life prediction.

[0093] In a specific embodiment, performing Hilbert envelope demodulation on the sensitive mode components to obtain the envelope spectrum includes:

[0094] The sensitive mode components are subjected to Hilbert transform to obtain orthogonal phase components. A complex analytic signal is constructed based on the sensitive mode components and the orthogonal phase components. The amplitude of the complex analytic signal is calculated to obtain the amplitude envelope waveform.

[0095] The amplitude envelope waveform is subjected to a fast Fourier transform to obtain the envelope spectrum. The amplitude of the envelope spectrum is then normalized to obtain a normalized envelope spectrum. Based on the normalized envelope spectrum, a spectrum is constructed to obtain an envelope spectrum diagram.

[0096] Specifically, firstly, a Hilbert transform operation is performed on the sensitive mode components selected in the previous steps. This operation essentially lags the phase of all frequency components of the time-domain signal by 90 degrees, thereby obtaining the orthogonal phase components that are strictly orthogonal to the original signal in the time domain. Next, a complex analytic signal is constructed based on the sensitive mode components and the orthogonal phase components. The construction formula needs to be clearly defined here. Physical definitions of the symbols in the code: The generated complex analytic signal serves as the mathematical carrier for subsequent demodulation. The generated complex analytic signal serves as the mathematical carrier for subsequent demodulation. The orthogonal phase component obtained after transformation, as the imaginary part of the complex number, provides a phase orthogonal reference; The imaginary unit is used to construct an orthogonal coordinate system in the complex plane. By introducing the imaginary part, negative frequency components are canceled out, so that the spectrum retains only single-sideband characteristics. Subsequently, the amplitude of the complex analytic signal is calculated using the formula... Calculate the magnitude of a complex vector, where The amplitude represents the instantaneous amplitude after demodulation. The square root operation is essentially to calculate the geometric length of the vector in the complex plane. This operation strips away the phase oscillation details of the high-frequency carrier and extracts only the amplitude modulation trend caused by mechanical impact, thus obtaining the amplitude envelope waveform. The amplitude envelope waveform is used to show the undulating trajectory of the periodic impact force carried by high-frequency vibration over time. For example, in the monitoring scenario of a ball-and-flexible steel cable pulling machine, when there is a small spalling in the raceway of the traction wheel bearing, the ball will excite a high-frequency structural resonance every time it passes the defect point. The original time-domain signal appears as a dense and chaotic high-frequency oscillation train, while the amplitude envelope waveform obtained by the above demodulation clearly shows a series of uniformly spaced pulse peaks. The peak value corresponds to the impact time, and the reciprocal of the pulse interval is the fault characteristic frequency.

[0097] After acquiring the time-domain signal reflecting the trend of impact energy changes, to reveal the periodicity of the impact event, a Fast Fourier Transform (FFT) needs to be performed on the amplitude envelope waveform to map the discrete time-domain sequence to the frequency domain, calculate its spectral density distribution, and obtain the envelope spectrum. At this point, the peak frequency in the spectrum directly corresponds to the recurrence frequency of the mechanical component's fault. Considering that differences in gain between different acquisition channels or signal transmission attenuation can cause fluctuations in the absolute amplitude of the spectrum, interfering with the quantitative assessment of the severity of the fault, the envelope spectrum must be normalized. This is typically done by dividing the amplitude at each frequency point by the maximum peak value of the spectrum or the total energy norm. The normalized envelope spectrum is obtained by mapping to a unified dimension range and eliminating the influence of system gain. Finally, a spectrum is constructed based on the normalized envelope spectrum, and a two-dimensional visualization curve is plotted with frequency as the abscissa and normalized amplitude as the ordinate. The frequency values ​​are marked at significant spectral peaks, and the envelope spectrum is finally obtained. The envelope spectrum is used to display the characteristic frequencies and their relative energy distribution corresponding to periodic impacts. If there are high-frequency impacts induced by cavitation in the hydraulic system of the drawing machine, significant normalized spectral peaks will appear at the blade passage frequency and its harmonics in the figure, while the background noise energy is suppressed to a low-level baseline. In this embodiment, the above scheme utilizes the nonlinear demodulation mechanism of Hilbert transform to successfully separate weak periodic fault impulses submerged in strong background noise and high-frequency carrier waves. Through frequency domain transformation, the pulse intervals that are difficult to be directly identified in the time domain are transformed into clear and sharp characteristic spectral lines in the frequency domain. This effectively overcomes the defects of traditional spectrum analysis, which suffers from blurred fault frequencies and complex sidebands due to modulation effects. It achieves accurate qualitative diagnosis of early weak faults in ball flexible steel cable drawing machines, significantly improving the signal-to-noise ratio of fault characteristic frequencies and the engineering credibility of diagnostic conclusions.

[0098] In a specific embodiment, constructing a complex analytic signal based on the sensitive mode component and the orthogonal phase component includes:

[0099] The real number sequence of the sensitive modal components is read to obtain the real part vibration sequence, and the imaginary number sequence is mapped to the orthogonal phase components to obtain the imaginary part vibration sequence;

[0100] The initial analytic signal is obtained by combining the real part vibration sequence and the imaginary part vibration sequence in the complex domain, and the conjugate symmetry of the initial analytic signal is checked to obtain the complex analytic signal.

[0101] Specifically, the sensitive modal components are first read as a real number sequence. This operation essentially loads the discrete-time vibration data obtained from the pre-decomposition from the memory buffer into the processor register point by point, forming a one-dimensional real number array, i.e., the real part vibration sequence, where each sampling point... Strictly corresponding to the instantaneous displacement or acceleration values ​​of mechanical vibration captured by the sensor at a specific moment, the real part vibration sequence is used to characterize the original amplitude change of the fault impact waveform. For example, in the monitoring of the traction wheel bearing of a ball-and-flexible steel cable drawing machine, the transient high-frequency oscillation peak generated when the ball rolls over the inner ring spalling point is directly mapped to the maximum point in this sequence. Simultaneously, the orthogonal phase components are mapped using an imaginary sequence, that is, another discrete data stream after Hilbert transform is multiplied by an imaginary unit. (satisfy The mathematical operations of the real number domain are projected onto the imaginary axis of the complex plane to generate the imaginary vibration sequence. The imaginary vibration sequence is used to characterize the orthogonal amplitude change that is 90 degrees out of phase with the original waveform. Its physical meaning is to convert the cosine component of the original signal into a sine component, so that at the same moment, the real part and the imaginary part form a strictly orthogonal vector pair in the phase space. If the original signal is at a peak at a certain moment, the corresponding imaginary sequence value should theoretically cross zero. This phase complementarity characteristic is the cornerstone for the subsequent construction of the single-sideband spectrum.

[0102] Subsequently, a complex field combination is performed based on the real part vibration sequence and the imaginary part vibration sequence, according to the definition of complex numbers. Index the same time The real and imaginary values ​​are merged into a complex unit. All complex units are arranged in chronological order to obtain the initial analytic signal. In digital signal processors, this process is usually manifested as packing two real number arrays into a complex structure array, realizing the dimensional expansion of the signal from a one-dimensional real axis to a two-dimensional complex plane. Next, the initial analytic signal is checked for conjugate symmetry, which is a key step to ensure the purity of the analytic signal. According to the properties of Fourier transform, the spectrum of a real signal has conjugate symmetry (i.e., it has negative frequency components), while the spectrum of an ideal complex analytic signal should only contain positive frequency components, and its negative frequency components must be zero. The verification process verifies the correctness of the construction by calculating the spectrum of the initial signal and checking whether the energy in its negative frequency region approaches machine zero or the theoretical antisymmetric relationship. If a significant negative frequency residue is detected, it indicates that there is a phase deviation or truncation error in the Hilbert transform filter design. The filter order or window function parameters need to be readjusted until the conjugate antisymmetry condition is met. The signal that is finally confirmed to be correct is the complex analytic signal. Taking a flexible steel cable drawing machine as an example, if the original vibration signal contains strong power frequency interference, after the above construction and verification, the complex analytical signal can effectively suppress the negative frequency image component caused by the power frequency, ensuring that the subsequent envelope extraction only targets the high-frequency modulation band caused by bearing failure. In this embodiment, the above scheme eliminates the aliasing interference of negative frequency components in traditional real signal processing through strict orthogonal construction of real and imaginary parts and conjugate symmetry verification, and constructs a mathematically rigorous single-sideband complex analytical signal model. This not only ensures the accuracy of amplitude calculation and avoids envelope waveform distortion caused by phase mismatch, but also significantly improves the separability of weak fault impacts in a strong noise background, laying a solid signal foundation for subsequent accurate extraction of characteristic frequencies.

[0103] In a specific embodiment, the step of extracting peak frequencies from the envelope spectrum to obtain a set of characteristic peak frequencies includes:

[0104] The envelope spectrum is compared point by point, and the positions where the amplitude is simultaneously greater than the left and right adjacent frequency points are marked as local maxima, thus obtaining a set of candidate peak points;

[0105] The candidate peak point set is screened out by noise baseline threshold, and candidate peak points with amplitudes lower than a preset signal-to-noise ratio threshold are removed to obtain the effective peak frequencies;

[0106] The effective peak frequencies are subjected to pairwise frequency ratio calculations to obtain a frequency ratio matrix. Based on the frequency ratio matrix, the effective peak frequencies are determined to have an integer multiple relationship. Multiple effective peak frequencies with an integer multiple relationship and sharing the same fundamental frequency component are grouped into the same frequency combination to obtain a set of characteristic peak frequencies.

[0107] Specifically, firstly, a discrete spectrum data traversal scan is performed on the previously generated envelope spectrum. The system reads the amplitude data of each frequency sampling point in the spectrum and compares the amplitude of the envelope spectrum point by point. It determines whether the amplitude of the current frequency point is strictly greater than the adjacent frequency point to its left and simultaneously greater than the adjacent frequency point to its right. If this condition is met, the point is determined to be a peak, and its frequency coordinates and amplitude information are recorded and marked as a local maximum. After traversing the entire frequency band, a candidate peak point set is obtained. This step initially locks down all frequency positions where energy is relatively concentrated, but it inevitably contains false peaks caused by random noise fluctuations. Subsequently, in order to distinguish between real fault characteristics and background noise, the candidate peak point set needs to be noise-controlled. The acoustic baseline threshold screening first estimates the noise floor level of the spectrum (e.g., using the statistical mean or median of the low amplitude region of the spectrum). Combined with a preset signal-to-noise ratio threshold (e.g., set to 6dB or an amplitude ratio of 2:1), a dynamic cutoff threshold is calculated. The amplitude of each point in the candidate peak point set is examined one by one. Any candidate peak point with an amplitude lower than the preset signal-to-noise ratio threshold is considered a false peak caused by noise interference and is resolutely eliminated. Only those points that are significantly higher than the noise baseline are retained as effective peak frequencies, ensuring the authenticity of the subsequent analysis objects. For example, in the monitoring scenario of a ball-flexible steel cable drawing machine, if an isolated small amplitude spike appears in a high-frequency band but its signal-to-noise ratio is only 3dB, which is far below the set threshold, then this point will be filtered out to avoid misjudging it as early bearing damage.

[0108] After obtaining the effective peak frequencies with high confidence, considering that the vibration signals excited by mechanical faults (such as bearing raceway spalling and gear tooth breakage) usually have harmonic characteristics, that is, the fault characteristic frequency and its harmonics will appear simultaneously in the spectrum, in order to identify this inherent periodicity from discrete frequency points, it is necessary to perform frequency ratio calculations on the effective peak frequencies pairwise to construct a symmetrical frequency ratio matrix. Each element in the matrix represents the quotient of two effective peak frequencies. Then, based on the frequency ratio matrix, the effective peak frequencies are judged for integer harmonic relationships. A tolerance range (such as ±2%) is set, and elements in the matrix whose values ​​are close to integers (such as 1, 2, 3, 4...) are scanned. If it is found that the ratio between multiple effective peak frequencies stably falls within the integer multiple relationship range, and these frequency points can be traced back to a common lowest frequency component (i.e., fundamental frequency), these frequency points are classified into the same frequency combination and regarded as a harmonic family originating from the same fault source. Finally, all the identified frequency combinations are integrated, and their fundamental frequency and main harmonic components are extracted to obtain a set of characteristic peak frequencies. This set completely characterizes the repetitive impact pattern of the core fault source inside the equipment. For example, if the theoretical characteristic frequency of the outer ring fault of the traction wheel bearing of the flexible steel cable drawing machine is 125Hz, after the above processing, the characteristic peak frequency concentration will include a group of frequency points that are integer multiples of 125Hz, 250Hz, and 375Hz. Even if the amplitude of a certain harmonic is low due to attenuation in the transmission path, it can still be accurately captured by correlating it with the multiples of other harmonics. In this embodiment, the above scheme, through a three-level logical architecture of "local extreme value location - signal-to-noise ratio filtering - harmonic family clustering", not only effectively eliminates the false spectral peak interference caused by random noise, but also cleverly utilizes the harmonic modulation characteristics of mechanical fault vibration, reorganizing the single spectral lines dispersed in the frequency domain into a fault characteristic frequency family with clear physical meaning. This significantly improves the accuracy and robustness of identifying weak fault characteristics of the flexible steel cable drawing machine in the context of strong noise, providing accurate and reliable frequency domain fingerprint basis for subsequent automatic fault type matching and severity quantification assessment.

[0109] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. It should be noted that the information interaction, execution process, etc. between the above devices / units are based on the same concept as the method embodiments of this application. Their specific functions and technical effects can be found in the embodiment section of the control device, and will not be repeated here.

[0110] Please see Figure 2 , Figure 2 This is a schematic diagram of the framework of an embodiment of the broadband focusing detection device based on a MEMS high-precision ultrasonic sensor according to this application. Figure 2As shown, the broadband focusing detection device based on MEMS high-precision ultrasonic sensor includes a receiving module 1, which is used to transmit broadband ultrasonic pulses to the object under test through MEMS ultrasonic sensor array, and receive ultrasonic echo signals reflected from different depth positions of the object under test, and obtain the original radio frequency data matrix composed of the ultrasonic echo signals.

[0111] The superposition module 2 is used to perform delay superposition processing on the echo data in the original radio frequency data matrix based on multiple preset focusing depths to obtain multiple sets of focused beam data corresponding to different focusing depths, and to extract the envelope of each set of focused beam data to obtain multiple sets of envelope feature curves.

[0112] The stitching module 3 is used to spatially stitch the multiple sets of envelope feature curves according to their corresponding focusing depths to form a continuous full-depth focusing feature curve covering the range from the near field to the far field, and to determine the boundary position of the internal defect of the object under test based on the amplitude abrupt change points on the continuous full-depth focusing feature curve.

[0113] The above module is used to perform the steps of the broadband focusing detection method based on MEMS high-precision ultrasonic sensor.

[0114] Reference Figure 3 This invention also provides a computer device whose internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0115] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0116] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0117] This application provides a computer program product that, when run on an electronic device, enables the electronic device to perform the functions of the various structures of the control device described above, or to implement the steps in the various method embodiments described above.

[0118] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0119] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0120] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0123] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

Claims

1. A ball flexible steel cable drawbench failure monitoring method characterized by, Includes the following steps: The vibration of the machine body during the operation of the ball flexible steel cable drawing machine is collected in real time to obtain the time-domain vibration waveform; The time-domain vibration waveform is divided into frequency bands to obtain multiple intrinsic mode components, and the kurtosis values ​​of the multiple intrinsic mode components are calculated to obtain a kurtosis curve. Peak filtering is performed on the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components, and Hilbert envelope demodulation is performed on the sensitive mode components to obtain envelope spectra. Peak frequencies are extracted from the envelope spectrum to obtain a set of characteristic peak frequencies. If any fault characteristic frequency in the set of characteristic peak frequencies is within the preset fault characteristic frequency threshold range corresponding to the drawing machine, it is determined that the ball flexible steel cable drawing machine has a corresponding type of fault state, and fault diagnosis alarm information is generated. The step of calculating the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve includes: For each of the multiple intrinsic mode components, the amplitude probability distribution is statistically analyzed to obtain an amplitude distribution histogram. The ratio of the fourth moment to the second moment of the amplitude distribution histogram is calculated to obtain a single kurtosis value. Based on the arrangement order of the multiple intrinsic mode components, the single kurtosis value of each intrinsic mode component is sequentially mapped onto the frequency band coordinate axis to obtain discrete kurtosis points, and curve fitting is performed on the discrete kurtosis points to obtain an initial kurtosis curve. Local maxima are identified on the initial kurtosis curve to obtain the curve peak points, and peak marks are made on the initial kurtosis curve based on the curve peak points to obtain a kurtosis curve graph; The calculation of the ratio of the fourth moment to the second moment on the amplitude distribution histogram to obtain a single kurtosis value includes: The amplitude distribution histogram is divided into amplitude intervals to obtain multiple amplitude intervals, and the number of sample points in the multiple amplitude intervals is counted to obtain an interval counting sequence; The expected value of the amplitude distribution histogram is calculated based on the interval counting sequence to obtain the distribution mean. The fourth and second central moments are then calculated based on the distribution mean to obtain the fourth and second moment values. The initial kurtosis number is obtained by ratioing the squares of the fourth moment value and the second moment value, and then the initial kurtosis number is normalized to obtain a single kurtosis value. The process of peak filtering of the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components includes: The peak coordinates of the kurtosis curve are read to obtain the peak frequency band position, and the multiple intrinsic mode components are back-indexed according to the peak frequency band position to obtain candidate mode components; The candidate modal components are compared with kurtosis thresholds. Candidate modal components with kurtosis values ​​lower than a preset screening threshold are marked as noise modes and removed to obtain the retained modal components. The correlation between the retained modal components is calculated to obtain a modal correlation coefficient matrix. Based on the modal correlation coefficient matrix, the retained modal components are redundantly merged. Multiple retained modal components with correlation exceeding a preset merging threshold are merged into a representative mode to obtain a sensitive modal component.

2. The ball flexible steel cable drawbench failure monitoring method of claim 1, wherein, The real-time acquisition of the machine body vibration during the operation of the ball-flexible steel cable drawing machine yields a time-domain vibration waveform, including: During operation, the acceleration sensor installed in the bearing seat of the ball flexible steel cable drawing machine reads the signal to obtain the original vibration sequence, and the original vibration sequence is preprocessed by removing DC and trend terms to obtain the purified vibration waveform. The purification vibration waveform is sampled at equal intervals using a preset sampling frequency to obtain analysis data segments. The analysis data segments are then normalized to obtain an amplitude normalization sequence. Based on the amplitude normalization sequence, the waveform is plotted to obtain a time-domain vibration waveform.

3. The ball flexible steel cable drawbench failure monitoring method of claim 1, wherein, The frequency band division of the time-domain vibration waveform yields multiple intrinsic mode components, including: Based on a preset number of modes, the time-domain vibration waveform is iteratively searched for the center frequency in the frequency domain to obtain multiple center frequency values. Then, Wiener filtering is performed on the time-domain vibration waveform based on the multiple center frequency values ​​to obtain an initial modal component set. For each initial modal component in the initial modal component set, the frequency band overlap is checked. When there is frequency band overlap between adjacent modes, the preset number of modes is adjusted and the center frequency iterative search is performed again until the frequency band separation condition is met, resulting in multiple independent modal components. The plurality of independent modal components are used as the plurality of intrinsic modal components, wherein the plurality of independent modal components are used to display a pure impulse response waveform with non-overlapping frequency bands.

4. The fault monitoring method for a flexible steel cable drawing machine according to claim 1, characterized in that, The step of performing Hilbert envelope demodulation on the sensitive mode components to obtain the envelope spectrum includes: The sensitive mode components are subjected to Hilbert transform to obtain orthogonal phase components. A complex analytic signal is constructed based on the sensitive mode components and the orthogonal phase components. The amplitude of the complex analytic signal is calculated to obtain the amplitude envelope waveform. The amplitude envelope waveform is subjected to a fast Fourier transform to obtain the envelope spectrum. The amplitude of the envelope spectrum is then normalized to obtain a normalized envelope spectrum. Based on the normalized envelope spectrum, a spectrum is constructed to obtain an envelope spectrum diagram.

5. A fault monitoring device for a flexible steel cable drawing machine, characterized in that, The method for fault monitoring of a flexible steel cable drawing machine according to any one of claims 1 to 4 includes: The acquisition module is used to acquire the vibration of the machine body in real time during the operation of the ball flexible steel cable drawing machine, and obtain the time-domain vibration waveform; The partitioning module is used to partition the time-domain vibration waveform into frequency bands to obtain multiple intrinsic mode components, and to calculate the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve. The filtering module is used to perform peak filtering on the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components, and to perform Hilbert envelope demodulation on the sensitive mode components to obtain an envelope spectrum. The extraction module is used to extract the peak frequency from the envelope spectrum to obtain a set of characteristic peak frequencies. When any fault characteristic frequency in the set of characteristic peak frequencies is within the preset fault characteristic frequency threshold range corresponding to the drawing machine, it is determined that the ball flexible steel cable drawing machine has a corresponding type of fault state and generates fault diagnosis alarm information. The step of calculating the kurtosis values ​​of the multiple intrinsic mode components to obtain a kurtosis curve includes: For each of the multiple intrinsic mode components, the amplitude probability distribution is statistically analyzed to obtain an amplitude distribution histogram. The ratio of the fourth moment to the second moment of the amplitude distribution histogram is calculated to obtain a single kurtosis value. Based on the arrangement order of the multiple intrinsic mode components, the single kurtosis value of each intrinsic mode component is sequentially mapped onto the frequency band coordinate axis to obtain discrete kurtosis points, and curve fitting is performed on the discrete kurtosis points to obtain an initial kurtosis curve. Local maxima are identified on the initial kurtosis curve to obtain the curve peak points, and peak marks are made on the initial kurtosis curve based on the curve peak points to obtain a kurtosis curve graph; The amplitude distribution histogram is divided into amplitude intervals to obtain multiple amplitude intervals, and the number of sample points in the multiple amplitude intervals is counted to obtain an interval counting sequence; The expected value of the amplitude distribution histogram is calculated based on the interval counting sequence to obtain the distribution mean. The fourth and second central moments are then calculated based on the distribution mean to obtain the fourth and second moment values. The initial kurtosis number is obtained by ratioing the squares of the fourth moment value and the second moment value, and then the initial kurtosis number is normalized to obtain a single kurtosis value. The process of peak filtering of the multiple intrinsic mode components based on the kurtosis curve to obtain sensitive mode components includes: The peak coordinates of the kurtosis curve are read to obtain the peak frequency band position, and the multiple intrinsic mode components are back-indexed according to the peak frequency band position to obtain candidate mode components; The candidate modal components are compared with kurtosis thresholds. Candidate modal components with kurtosis values ​​lower than a preset screening threshold are marked as noise modes and removed to obtain the retained modal components. The correlation between the retained modal components is calculated to obtain a modal correlation coefficient matrix. Based on the modal correlation coefficient matrix, the retained modal components are redundantly merged. Multiple retained modal components with correlation exceeding a preset merging threshold are merged into a representative mode to obtain a sensitive modal component.

6. A computer device, characterized in that, The method includes a memory and a processor that are coupled to each other. The memory stores program instructions, and the processor executes the program instructions to implement the fault monitoring method for a ball-flexible steel cable drawing machine according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The device stores program instructions that can be executed by a processor, the program instructions being used to implement the fault monitoring method for a ball-and-flexible steel cable drawing machine according to any one of claims 1 to 4.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, enables the implementation of the steps of the fault monitoring method for the ball-and-flexible steel cable drawing machine as described in any one of claims 1 to 4.

Images