A method and system for online monitoring of grinding mill bearing wear status

Abstract

The present invention discloses an online monitoring method and system for the wear status of a grinding mill bearing, which specifically relates to the technical field of industrial equipment status monitoring. The system is used to solve the problem of misjudgment caused by signal feature offset in existing grinding mill bearing vibration signal monitoring methods under dynamic load conditions. The system collects grinding pressure parameters and bearing vibration signals in real time, analyzes the dynamic correlation between the two and separates interference frequency band components, extracts the time-frequency ridge trajectory and curvature mutation density of the characteristic frequency band, generates a dynamic reference standard to adaptively adjust the amplitude-frequency threshold, and finally outputs the wear status assessment result through joint amplitude-frequency analysis, thereby realizing accurate monitoring of the bearing wear status under complex working conditions and significantly improving the reliability of the online monitoring system and the scientific nature of maintenance decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment status monitoring, and more particularly to an online monitoring method and system for the wear status of a grinding mill bearing. Background Art

[0002] In the flour processing industry, the long-term stable operation of the mill's bearings, a core piece of equipment, directly impacts production efficiency and equipment maintenance costs. Currently, monitoring mill bearing wear typically involves combining vibration signal analysis with temperature detection. This method collects real-time bearing operating data and compares it with preset thresholds to provide early warning of abnormalities. However, in actual operation, the mill must dynamically adjust the grinding pressure based on the raw material characteristics and processing requirements. Frequent changes in this process parameter lead to significant fluctuations in the bearing's internal stress distribution, which in turn causes unexpected shifts in the vibration signal characteristics.

[0003] In the existing technology, the bearing wear monitoring method based on vibration signals has insufficient adaptability under dynamic load conditions. Due to the strong coupling between grinding pressure adjustment and bearing stress distribution, the main frequency characteristics and amplitude of the vibration signal will deviate from the normal wear model as the process parameters change, resulting in a significantly increased risk of misjudgment. This problem is particularly prominent in industrial scenarios such as flour processing that require frequent load adjustments, seriously affecting the reliability of the online monitoring system. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for online monitoring of the wear status of a grinding mill bearing to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for online monitoring of the wear state of a grinding mill bearing comprises the following steps:

[0007] S1, obtaining the grinding pressure parameters and bearing vibration signals of the grinding mill during operation;

[0008] S2. Analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristic of the bearing vibration signal;

[0009] S3, separating the interference frequency band component and the characteristic frequency band component in the bearing vibration signal according to the main frequency characteristic offset;

[0010] S4. Perform time-frequency ridge tracking on the characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, calculate the curvature mutation density of the ridge trajectory, and determine the dynamic reference standard of the bearing wear characteristics based on the curvature mutation density;

[0011] S5. Adjusting the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard;

[0012] S6. Based on the adjusted amplitude threshold and frequency threshold, perform amplitude-frequency joint analysis on the characteristic frequency band components and output the bearing wear status assessment result.

[0013] In a preferred embodiment, obtaining the grinding pressure parameters and bearing vibration signals during operation of the grinding mill includes:

[0014] Grinding pressure parameters are collected through a pressure sensor installed on the grinding roller of the mill, and bearing vibration signals are collected through an acceleration sensor installed on the bearing seat of the mill;

[0015] The grinding pressure parameters and the bearing vibration signals are synchronously sampled to generate a time-aligned grinding pressure parameter sequence and a bearing vibration signal sequence.

[0016] In a preferred embodiment, the acceleration sensor is arranged in the horizontal and vertical directions of the bearing seat, and the pressure sensor is connected to the hydraulic adjustment mechanism of the grinding roller.

[0017] In a preferred embodiment, analyzing the corresponding relationship between the change in the grinding pressure parameter and the offset of the main frequency characteristic of the bearing vibration signal includes:

[0018] A dynamic time window is generated based on the time-aligned grinding pressure parameter sequence, and the interval length of the dynamic time window is inversely proportional to the change rate of the grinding pressure parameter;

[0019] Extract the main frequency energy distribution and phase synchronization index of the bearing vibration signal sequence within the dynamic time window;

[0020] According to the joint change trend of the main frequency energy distribution and the phase synchronization index, a nonlinear mapping rule between the grinding pressure parameter change and the main frequency characteristic offset is constructed;

[0021] The nonlinear mapping rule is used to indicate the confidence interval of the main frequency characteristic offset under dynamic load fluctuation.

[0022] In a preferred embodiment, separating the interference frequency band component and the characteristic frequency band component in the bearing vibration signal according to the main frequency characteristic offset includes:

[0023] Determine the dynamic frequency band division boundary based on the main frequency characteristic offset and nonlinear mapping rules;

[0024] According to the dynamic frequency band division boundary, the frequency band whose energy ratio in the bearing vibration signal is lower than the main frequency energy distribution threshold and whose phase synchronization index is negative is extracted as the interference frequency band component;

[0025] Extract the frequency band whose energy ratio is higher than the main frequency energy distribution threshold and whose phase synchronization index is positive as the characteristic frequency band component;

[0026] Complementary filtering is performed on the interference frequency band components and the characteristic frequency band components to generate a separated signal sequence aligned with the dynamic time window.

[0027] In a preferred embodiment, time-frequency ridge tracking is performed on the characteristic frequency band components, the ridge trajectory of the time-frequency energy distribution is extracted, and the curvature mutation density of the ridge trajectory is calculated. The dynamic reference standard of the bearing wear characteristics is determined based on the curvature mutation density, including:

[0028] The time-frequency ridge line of the time domain signal of the characteristic frequency band component is tracked in the dynamic time window to generate the ridge line trajectory of the time-frequency energy distribution;

[0029] Perform cubic spline interpolation on the ridge trajectory and calculate the curvature change rate of the interpolated trajectory; count the number of mutation points where the curvature change rate exceeds the set mutation threshold per unit time to generate the curvature mutation density;

[0030] Based on the mapping relationship between the curvature mutation density and the bearing wear degree in the historical wear data, the dynamic reference standard of the bearing wear characteristics in the current dynamic time window is determined;

[0031] The dynamic reference standards include the normal fluctuation range of curvature mutation density and the abnormality judgment threshold.

[0032] In a preferred embodiment, adjusting the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard includes:

[0033] Adjust the amplitude threshold of the bearing vibration signal according to the normal fluctuation range of the curvature mutation density in the dynamic reference standard;

[0034] Adjust the frequency threshold of the bearing vibration signal according to the abnormal judgment threshold of the curvature mutation density in the dynamic reference standard;

[0035] Generate a dynamic amplitude-frequency joint judgment rule based on the adjusted amplitude threshold and frequency threshold;

[0036] The dynamic amplitude-frequency joint judgment rule is matched with the bearing vibration signal in the current dynamic time window in real time, and the effective range of the amplitude threshold and frequency threshold is updated.

[0037] In a preferred embodiment, the characteristic frequency band components are subjected to an amplitude-frequency joint analysis based on the adjusted amplitude threshold and frequency threshold, and the bearing wear status assessment result is output, including:

[0038] Calculate the time domain energy mean and extract the frequency domain main frequency offset of the characteristic frequency band components in the dynamic time window;

[0039] Determine whether the time domain energy mean exceeds the adjusted amplitude threshold and whether the frequency domain main frequency offset exceeds the adjusted frequency threshold;

[0040] When both the time-domain energy mean and the frequency-domain main frequency offset exceed the limit, the bearing wear pattern is generated according to the combined coding of the exceeding amplitude threshold and the frequency threshold;

[0041] Based on the cumulative results of bearing wear patterns in the current dynamic time window and adjacent windows, the bearing wear status assessment results are output.

[0042] In a preferred embodiment, the bearing wear state assessment result is the proportion of occurrences of the same wear pattern within three consecutive dynamic time windows.

[0043] In another aspect, the present invention provides an online monitoring system for bearing wear of a grinding mill, comprising the following modules:

[0044] Signal acquisition module, used to obtain grinding pressure parameters and bearing vibration signals when the mill is running;

[0045] Dynamic correlation module, used to analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristics of the bearing vibration signal;

[0046] The frequency band separation module is used to separate the interference frequency band components and characteristic frequency band components in the bearing vibration signal according to the main frequency characteristic offset;

[0047] The curvature analysis module is used to track the time-frequency ridges of characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, and calculate the curvature mutation density of the ridge trajectory. The dynamic reference standard of the bearing wear characteristics is determined based on the curvature mutation density.

[0048] A threshold adjustment module, used to adjust the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard;

[0049] The condition assessment module is used to perform amplitude-frequency joint analysis on the characteristic frequency band components based on the adjusted amplitude threshold and frequency threshold, and output the bearing wear condition assessment results.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. By dynamically correlating the grinding pressure parameters with the main frequency characteristic offset of the bearing vibration signal, a real-time mapping relationship between load changes and vibration characteristics is established, effectively overcoming the misjudgment problem caused by signal offset under dynamic load in traditional monitoring methods. By separating the interference frequency band components and extracting the time-frequency ridge trajectory of the characteristic frequency band, the identification accuracy of the wear characteristic components in the vibration signal is significantly improved, avoiding the interference of background noise and process fluctuations on the monitoring results. At the same time, the dynamic reference standard generated based on the curvature mutation density can adapt to the changes in wear characteristics under different load conditions, making the threshold adjustment more in line with the actual operating status, thereby greatly improving the sensitivity and reliability of the monitoring system.

[0052] 2. Compared with static threshold and single signal analysis, the joint amplitude-frequency analysis strategy is adopted, combining the time-domain energy distribution and the frequency-domain harmonic attenuation characteristics to achieve a multi-dimensional comprehensive evaluation of the bearing wear status; by dynamically adjusting the amplitude and frequency thresholds, the system can accurately capture the weak characteristic signals of early wear and maintain stable judgment logic under complex working conditions; this closed-loop monitoring mechanism not only reduces the false alarm rate and missed alarm rate, but also provides a quantifiable decision-making basis for equipment maintenance, significantly extending the bearing service life and reducing unplanned downtime losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for online monitoring of bearing wear status of a grinding mill according to the present invention;

[0054] Figure 2 The present invention is a structural schematic diagram of an online monitoring system for the wear status of a grinding mill bearing. DETAILED DESCRIPTION

[0055] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1: Figure 1 The present invention provides an online monitoring method for the wear state of a grinding mill bearing, which comprises the following steps:

[0057] S1, obtaining the grinding pressure parameters and bearing vibration signals of the grinding mill during operation;

[0058] S2. Analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristic of the bearing vibration signal;

[0059] S3, separating the interference frequency band component and the characteristic frequency band component in the bearing vibration signal according to the main frequency characteristic offset;

[0060] S4. Perform time-frequency ridge tracking on the characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, calculate the curvature mutation density of the ridge trajectory, and determine the dynamic reference standard of the bearing wear characteristics based on the curvature mutation density;

[0061] S5. Adjusting the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard;

[0062] S6. Based on the adjusted amplitude threshold and frequency threshold, perform amplitude-frequency joint analysis on the characteristic frequency band components and output the bearing wear status assessment result.

[0063] S1. Obtaining the grinding pressure parameters and bearing vibration signals during the operation of the grinding mill, specifically implemented as follows:

[0064] Grinding pressure parameters are collected using a pressure sensor mounted on the grinding roller of the mill. The pressure sensor is attached to a flange connection on the oil inlet line of the grinding roller's hydraulic adjustment mechanism, where it is sealed to the hydraulic circuit. The pressure sensor's detection range covers the mill's operating pressure range. The output signal is converted to a digital signal by an analog-to-digital converter and transmitted to subsequent processing steps. The sampling frequency for collecting grinding pressure parameters is 100 times per second. During data preprocessing, the grinding pressure parameters are arranged chronologically according to the mill's operating cycle to generate a grinding pressure parameter sequence.

[0065] Bearing vibration signals are collected using accelerometers installed on the mill's bearing housing. These accelerometers consist of a horizontal accelerometer and a vertical accelerometer. The horizontal accelerometer is mounted on the outer surface of the bearing housing parallel to the roller axis, while the vertical accelerometer is mounted perpendicular to the roller axis. The accelerometers are secured by a magnetic base, and the contact surface between the base and the bearing housing is coated with a non-slip damping adhesive to reduce vibration signal distortion. The accelerometers have a range that covers the typical vibration range of mill bearings, with a sampling frequency of 100 times per second. The vibration signals are processed by an anti-aliasing filter and transmitted to subsequent processing steps to generate a bearing vibration signal sequence.

[0066] The grinding pressure parameters and bearing vibration signals are synchronously sampled to generate time-aligned grinding pressure parameter and bearing vibration signal sequences. A global clock synchronization mechanism is established to send a synchronization trigger signal to the pressure sensor and accelerometer. Based on the synchronization trigger signal, the pressure sensor and accelerometer initiate sampling at the same time. Each data point in the grinding pressure parameter sequence and the bearing vibration signal sequence contains the same timestamp tag, which is used to align the grinding pressure parameter sequence and the bearing vibration signal sequence with millisecond-level accuracy. After synchronous sampling, the time-aligned grinding pressure parameter sequence and bearing vibration signal sequence are stored in a buffer area for subsequent analysis.

[0067] Acceleration sensors are installed horizontally and vertically on the bearing seat. The horizontal acceleration sensor's detection direction is parallel to the radial force applied to the roller bearing, while the vertical acceleration sensor's detection direction is parallel to the axial force applied to the roller bearing. A pressure sensor is connected to the roller's hydraulic adjustment mechanism, whose hydraulic oil pressure is regulated via a proportional valve. The pressure sensor's detection end directly contacts the oil in the hydraulic oil pipeline, collecting the hydraulic oil pressure value in real time as the grinding pressure parameter.

[0068] During the generation of the grinding pressure parameter sequence and the bearing vibration signal sequence, data cleaning is performed on the original signals. This involves smoothing the grinding pressure parameter sequence using a sliding window mean filter. The sliding window is set to 10 sampling points, and the arithmetic mean of the data points within the window replaces the original value at the window center. A threshold comparison method is used to remove abnormal data points from the bearing vibration signal sequence. The threshold in this method is determined by statistically analyzing the maximum amplitude of the vibration signal in historical normal operation data. Data points exceeding this maximum amplitude are marked as outliers and replaced by linear interpolation.

[0069] Through the above-described implementation, the acquisition and synchronization of grinding pressure parameters and bearing vibration signals can accurately match real-time operating conditions under dynamic load changes. For example, when the pressure value in the grinding pressure parameter sequence increases from 5 MPa to 8 MPa within 1 second, the amplitude variation range and frequency distribution characteristics of the vibration signal at the corresponding moment in the synchronously acquired bearing vibration signal sequence can fully reflect the actual operating status of the bearing, avoiding the failure of the correlation between load changes and vibration characteristics due to time deviation.

[0070] S2. Analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristic of the bearing vibration signal. The specific implementation is as follows:

[0071] A dynamic time window is generated based on the time-aligned grinding pressure parameter sequence. The length of the dynamic time window is dynamically adjusted based on the grinding pressure parameter's rate of change, which is calculated by calculating the ratio of the absolute value of the difference between two adjacent grinding pressure parameter data points to the time interval. For example, when the grinding pressure parameter increases from 5 MPa to 7 MPa in 0.1 seconds, the rate of change is (7-5) / 0.1 = 20 MPa / s, so the dynamic time window length is set to 0.05 seconds. If the rate of change is 5 MPa / s, the window length is set to 0.2 seconds. The dynamic time window is generated according to the following rule: window length = baseline window length × (baseline change rate / current change rate), where the baseline window length is set to 0.1 seconds and the baseline change rate is set to 10 MPa / s based on the average pressure change rate of historical operating data. This rule ensures that the length of the dynamic time window is inversely proportional to the grinding pressure parameter's rate of change, ensuring that the analysis frequency increases when the load fluctuates rapidly and that the analysis window is extended to capture long-period features when the load is stable.

[0072] The main frequency energy distribution and phase synchronization index of the bearing vibration signal sequence are extracted within the dynamic time window. The main frequency energy distribution extraction process includes: performing fast Fourier transform on the bearing vibration signal sequence within the dynamic time window, identifying the frequency component with the highest energy in the spectrum as the main frequency; calculating the proportion of the main frequency component in the total spectrum energy as the main frequency energy distribution value. The phase synchronization index extraction process includes: performing Hilbert transform on the bearing vibration signal sequence within the dynamic time window to obtain the instantaneous phase sequence of the signal; calculating the sliding window covariance of the instantaneous phase sequence and the grinding pressure parameter change rate as the phase synchronization index. For example, the sliding window length is set to 1 / 5 of the dynamic time window length. When the covariance calculation result is greater than zero, it indicates that the phase change is synchronized with the pressure change trend. When it is less than zero, it indicates phase loss.

[0073] Based on the joint variation trend of the main frequency energy distribution and the phase synchronization index, a nonlinear mapping rule is constructed between the grinding pressure parameter change and the main frequency characteristic offset. The specific construction process includes: statistically analyzing the joint distribution range of the main frequency energy distribution and the phase synchronization index in historical normal operation data as a benchmark reference interval; during real-time monitoring, if the main frequency energy distribution value is lower than the lower limit of the benchmark reference interval and the phase synchronization index is continuously negative, the main frequency characteristic offset is determined to be outside the normal range; if the main frequency energy distribution value is within the benchmark reference interval and the phase synchronization index is positive, the main frequency characteristic offset is determined to be a normal fluctuation. The nonlinear mapping rule is defined by a two-dimensional decision boundary, which is jointly determined by the main frequency energy distribution threshold and the phase synchronization index threshold. The threshold is dynamically adjusted according to the statistical quantiles of the historical data. For example, the main frequency energy distribution threshold is set to the 10th percentile of the historical data, and the phase synchronization index threshold is set to the 25th percentile of the historical data.

[0074] The nonlinear mapping rule is used to indicate the confidence interval of the main frequency characteristic offset under dynamic load fluctuations. The confidence interval update logic includes: when the grinding pressure parameter change rate is in a low fluctuation state, the upper and lower limits of the confidence interval are expanded to allow a larger main frequency characteristic offset; when the grinding pressure parameter change rate is in a high fluctuation state, the confidence interval is compressed to improve detection sensitivity. For example, the low fluctuation state is defined as a change rate of less than 5 MPa / s, at which time the normal range of the main frequency characteristic offset is relaxed to ±15% of the historical baseline value; the high fluctuation state is defined as a change rate greater than 15 MPa / s, at which time the normal range is compressed to ±5% of the historical baseline value.

[0075] When extracting the primary frequency energy distribution within a dynamic time window, if the window length is too short, resulting in insufficient spectral resolution, an overlapping window analysis method is used to compensate. The overlapping window step size is set to 50% of the window length. The overlapping data of adjacent windows are fused using a weighted average, with weight coefficients dynamically assigned based on the signal-to-noise ratio of the data within the window. For example, window regions with a signal-to-noise ratio above 20 decibels are assigned a weight of 0.7, while regions below 20 decibels are assigned a weight of 0.3 to suppress noise interference with the primary frequency energy distribution.

[0076] In the process of constructing nonlinear mapping rules, trend fitting is performed on the joint change trend of the main frequency energy distribution and the phase synchronization index. The trend fitting method includes: discretizing the relationship curve between the main frequency energy distribution and the phase synchronization index in the historical data into multiple piecewise linear intervals, each linear interval corresponding to a specific load change stage; matching the real-time data with the corresponding linear interval according to the current grinding pressure parameter change rate, and calculating the predicted value of the main frequency characteristic offset. When the residual between the predicted value and the actual value exceeds the set threshold, the confidence interval update mechanism is triggered. For example, the residual threshold is set to 3 times the standard deviation of the historical data residual. When this threshold is exceeded, the benchmark reference interval is recalculated.

[0077] Through this implementation, the correspondence between changes in the grinding pressure parameter and the dominant frequency characteristic offset can adapt to complex dynamic load conditions. For example, during a rapid increase in the grinding pressure parameter (rate of change of 20 MPa / s), the dynamic time window is shortened to 0.05 seconds. The dominant frequency energy distribution value increases significantly due to the enhanced high-frequency vibration, and the phase synchronization index briefly becomes negative due to the sudden load change. At this time, the nonlinear mapping rule combines the combined trends of energy increase and phase loss to determine that the dominant frequency characteristic offset enters the transition state confidence interval, avoiding misjudgment of abnormal wear.

[0078] S3, separating the interference frequency band component and the characteristic frequency band component in the bearing vibration signal according to the main frequency characteristic offset, specifically implemented as follows:

[0079] The dynamic frequency band division boundaries are determined based on the main frequency characteristic offset and nonlinear mapping rules. The logic for generating dynamic frequency band division boundaries includes dividing the main frequency characteristic offset into a normal offset range and an abnormal offset range based on the confidence interval defined in the nonlinear mapping rules. The normal offset range is determined by the statistical distribution of the main frequency characteristic offset in historical normal operation data, for example, the 5th to 95th percentile of historical data is used as the normal offset range. The abnormal offset range is the area outside the normal offset range. When the main frequency characteristic offset is within the normal offset range, the dynamic frequency band division boundaries are set within a frequency band of ±5% of the main frequency component. When the main frequency characteristic offset is within the abnormal offset range, the dynamic frequency band division boundaries are extended to a frequency band of ±10% of the main frequency component. For example, if the nonlinear mapping rules indicate that the confidence interval of the current main frequency characteristic offset is ±8%, the dynamic frequency band division boundaries are set within a frequency band of ±8% of the main frequency center frequency. Frequency bands outside this range are considered interference or characteristic components to be separated. The dynamic frequency band division boundary update period is consistent with the length of the dynamic time window, ensuring that the frequency band division within each time window adapts to load changes in real time.

[0080] Based on the dynamic frequency band demarcation boundaries, frequency bands in the bearing vibration signal whose energy contribution is below the main frequency energy distribution threshold and whose phase synchronization index is negative are extracted as interference frequency band components. The extraction criteria for interference frequency band components include: within the dynamic frequency band demarcation boundaries, the bearing vibration signal sequence is divided into multiple sub-bands according to frequency resolution, and the energy contribution of each sub-band is calculated (sub-band energy / total energy). The energy contribution is calculated by integrating the vibration signal within the sub-band to obtain the sub-band energy value, and then dividing the sub-band energy value by the total energy value. If the energy contribution of a sub-band is below the main frequency energy distribution threshold (for example, the 10th percentile of historical data) and the corresponding phase synchronization index is negative (the phase change is out of lock with the load change trend), the sub-band is determined to be an interference frequency band component. For example, if the energy contribution of a sub-band is 3% (below the 5% threshold) and the phase synchronization index is -0.2, the sub-band is classified as an interference frequency band component. The interference frequency band components are marked by setting the amplitude of the corresponding sub-band to zero in the frequency domain or filtering it out through a band-stop filter in the time domain.

[0081] Frequency bands whose energy percentage exceeds the main frequency energy distribution threshold and whose phase synchronization index is positive are extracted as characteristic frequency band components. The extraction criteria for characteristic frequency band components include: within the dynamic frequency band demarcation boundary, if the energy percentage of a sub-band exceeds the main frequency energy distribution threshold and the phase synchronization index is positive (the phase change is synchronized with the load change trend), the sub-band is determined to be a characteristic frequency band component. For example, if the energy percentage of a sub-band is 12% (above the threshold of 5%) and the phase synchronization index is 0.6, the sub-band is classified as a characteristic frequency band component. Characteristic frequency band components are extracted by retaining the amplitude and phase information of the corresponding sub-band in the frequency domain or by using a bandpass filter in the time domain to retain the target frequency band. The phase synchronization index of the characteristic frequency band component is obtained by calculating the sliding covariance between the instantaneous phase of the sub-band and the rate of change of the grinding pressure parameter. The sliding window length is consistent with the length of the dynamic time window.

[0082] Complementary filtering is performed on the interference frequency band components and the characteristic frequency band components to generate a separated signal sequence aligned with the dynamic time window. The complementary filtering process is implemented by subtracting the signal components corresponding to the interference frequency band components from the original bearing vibration signal sequence, and retaining the signal components corresponding to the characteristic frequency band components. The specific steps are: performing an inverse Fourier transform on the interference frequency band components to generate a time domain interference signal, and subtracting the original vibration signal sequence from the interference signal sequence one by one at each time point to obtain the time domain signal of the characteristic frequency band components. Alternatively, the characteristic frequency band components can be directly extracted from the original signal and the interference frequency band components can be suppressed. For example, in the frequency domain, the amplitude of the interference frequency band components is set to zero and then an inverse Fourier transform is performed. The signal sequence after complementary filtering is strictly aligned with the timestamp of the dynamic time window to ensure the timing consistency of subsequent analysis. The output of the complementary filtering process is the separated characteristic frequency band signal sequence and the interference frequency band signal sequence, which together constitute a complete separated signal sequence.

[0083] During the adjustment process of the dynamic frequency band division boundary, if the main frequency feature offset continues to exceed the normal offset range in multiple consecutive dynamic time windows, the boundary adaptive expansion mechanism is triggered. The boundary adaptive expansion mechanism includes: associating the expansion amplitude of the dynamic frequency band division boundary with the degree of deviation of the main frequency feature offset, and the degree of deviation is quantified by the absolute value of the difference between the current offset and the historical benchmark value. For example, if the main frequency feature offset exceeds the normal range for three consecutive windows and the average deviation value is 20% of the historical benchmark, the dynamic frequency band division boundary is extended to the ±15% frequency band of the main frequency component. The expanded boundary is gradually shrunk in subsequent windows according to the return of the offset to the normal range, and the shrinkage rate is 1% of the frequency band per window until it returns to the initial set range.

[0084] When extracting interference and characteristic frequency band components, smooth transition processing is performed on transition bands near the boundaries. This smooth transition is achieved by linearly weighting the energy proportions of the interference and characteristic frequency bands within a ±2% band of the dynamic frequency band boundary. The weight coefficients are dynamically assigned based on the distance between the frequency band and the boundary. For example, the sub-band at a frequency point -2% from the boundary has a weight of 0.2 (interference contribution 80%), while the sub-band at a frequency point +2% from the boundary has a weight of 0.8 (characteristic contribution 80%). This eliminates signal distortion caused by sudden frequency band changes.

[0085] The formula for calculating the sub-band energy proportion after weighted fusion is: Fusion Energy = Interference Energy × Weight + Feature Energy × (1 - Weight). This method ensures a smooth transition of signal energy at band boundaries, avoiding high-frequency noise caused by rigid divisions.

[0086] Through the above-mentioned implementation, the separation process of interference frequency band components and characteristic frequency band components can adapt to the fluctuations of the main frequency characteristics under dynamic loads. For example, when the grinding pressure parameters fluctuate violently and cause the main frequency characteristic offset to enter the abnormal range, the dynamic frequency band division boundary automatically expands to ensure that the high-frequency interference components are effectively separated; at the same time, the positive and negative judgment of the phase synchronization index is combined with the energy ratio condition to avoid misjudging transient load impacts as bearing wear characteristics. During the stable stage of the grinding pressure parameters, the dynamic frequency band division boundary shrinks and the extraction accuracy of the characteristic frequency band components is improved, thereby accurately identifying early signs of bearing wear.

[0087] S4. Track the time-frequency ridges of the characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, and calculate the curvature mutation density of the ridge trajectory. Determine the dynamic reference standard of the bearing wear characteristics based on the curvature mutation density. The specific implementation is as follows:

[0088] The time-frequency ridge line tracking of the time domain signal of the characteristic frequency band component is performed within the dynamic time window to generate the ridge line trajectory of the time-frequency energy distribution. The implementation method of the time-frequency ridge line tracking includes: performing continuous wavelet transform on the time domain signal of the characteristic frequency band component within the dynamic time window to generate a time-frequency energy distribution matrix; searching for the frequency value corresponding to the energy maximum in the time-frequency energy distribution matrix at each time point to form a time-frequency ridge line trajectory. The length of the dynamic time window is aligned with the time of the separated signal sequence generated in step S3 to ensure that the ridge line tracking range strictly matches the load fluctuation stage. For example, for a signal segment with a dynamic time window length of 0.1 seconds, each time point is traversed at intervals of milliseconds, and the frequency with the highest energy in the wavelet energy spectrum is selected as the ridge line point, and all ridge line points are connected to generate the time-frequency ridge line trajectory. The starting time of the dynamic time window is synchronized with the key time point of the change of the grinding pressure parameter, such as the starting time of the pressure surge or plunge, to ensure that the ridge line tracking covers the vibration characteristics of the load mutation stage.

[0089] The ridge trajectory is interpolated using cubic spline interpolation, and the curvature rate of the interpolated trajectory is calculated. The specific steps of cubic spline interpolation include: inputting the discrete point sequence of the time-frequency ridge trajectory into the interpolation algorithm to generate a smooth, continuous curve. The interpolation algorithm parameters include node spacing and a smoothing factor. The node spacing is set to 1% of the dynamic time window length; for example, for a 0.1-second window length, the node spacing is 1 millisecond. The smoothing factor is dynamically adjusted based on the amplitude of the local curvature fluctuation of the ridge trajectory; the larger the fluctuation, the smaller the smoothing factor. Samples are taken at fixed intervals along the interpolated curve, and the curvature value of each sampling point is calculated. The curvature value is calculated by calculating the ratio of the rate of change of the tangent angle of the curve at a point to the arc length, using the definition of curvature in differential geometry. The curvature rate is calculated by taking the difference between the curvature values ​​of adjacent sampling points and dividing it by the sampling interval. For example, if the sampling interval is 1 millisecond and the curvature difference between two points is 0.05, then the curvature rate of change is 0.05 / 0.001 = 50 seconds.

[0090] The curvature mutation density is calculated by counting the number of mutation points whose curvature change rate exceeds a set mutation threshold per unit time. The mutation threshold is set based on the 95th percentile of the statistical distribution of curvature change rates in historical normal operation data. During real-time monitoring, the baseline mutation threshold is dynamically adjusted by ±20% based on the average curvature change rate within the current dynamic time window. For example, if the historical baseline threshold is 40 per second and the average curvature change rate within the current window is 35 per second, the real-time threshold is adjusted to 35 × 1.2 = 42 per second. The method for counting mutation points per unit time is to iterate over all sampling points within the dynamic time window, count the number of points whose curvature change rate exceeds the threshold, and then divide the count by the window length. For example, if 5 mutation points are detected within a 0.1-second window, the curvature mutation density is 5 / 0.1 = 50 per second. During the statistical process, if the same sampling point exceeds the threshold for multiple consecutive intervals, it is counted as only one mutation to avoid double counting.

[0091] Based on the mapping relationship between curvature mutation density and bearing wear severity in historical wear data, a dynamic reference standard for bearing wear characteristics within the current dynamic time window is determined. This mapping process involves collecting historical bearing lifecycle data from normal to severe wear, recording curvature mutation density values ​​at each stage. Linear regression analysis is then used to fit a quantitative relationship between curvature mutation density and wear severity, generating a reference standard curve. For example, if the fitting results show that every 10-second increase in curvature mutation density is associated with a 0.01 mm increase in wear depth, the normal fluctuation range in the dynamic reference standard is set to ±3 standard deviations of the historical normal operating density, and the abnormality threshold is set to 1.5 times the upper limit of the normal range. During real-time monitoring, if the current curvature mutation density falls within the normal fluctuation range, the bearing is considered normal; if it exceeds the abnormality threshold, a wear risk is identified. The input data for the linear regression analysis must be normalized by dividing the curvature mutation density value by the historical maximum value and the wear depth value by the maximum allowable wear of the bearing material to ensure data consistency.

[0092] The dynamic reference standard includes the normal fluctuation range of the curvature mutation density and the abnormal judgment threshold. The update logic of the normal fluctuation range is: based on the curvature mutation density data of the latest 100 dynamic time windows, recalculate the mean and standard deviation, and dynamically adjust the normal range. For example, if the mean of the latest 100 windows is 30 per second and the standard deviation is 5 per second, the normal range is updated to 30±3×5=15 to 45 per second. The abnormal judgment threshold is set according to the critical wear data in the equipment maintenance record. For example, the maximum curvature mutation density of the 10 windows before the historical critical wear event is taken as the threshold. The update cycle of the dynamic reference standard is synchronized with the equipment maintenance cycle, such as every 24 hours or after each grinding task is completed, to ensure that the detection standard adapts to the aging trend of the equipment.

[0093] During the curvature rate of change calculation process, if the interpolated trajectory exhibits local distortion, a trajectory smoothing correction mechanism is triggered. This smoothing correction mechanism involves checking whether the sudden change in the curvature rate of change exceeds twice the historical maximum sudden change. If so, a median filter is applied to the curvature values ​​of the five sampling points before and after the distortion point, replacing the original curvature value at the distorted point. For example, if the curvature rate of change at a point is 120 per second (the historical maximum is 60 per second), the median value of the 11 points before and after it is taken to eliminate the interference of abnormal sudden changes. The median filter window length is adaptively adjusted based on the number of distorted points. If three consecutive points exceed the historical threshold, the window length is extended to 15 points to cover a longer abnormal period. The corrected curvature rate of change is then re-incorporated into the sudden change point count to ensure data reliability.

[0094] Through the above implementation, the curvature mutation density of the time-frequency ridge trajectory can effectively characterize the dynamic characteristics of bearing wear. For example, in the early stage of bearing wear, the time-frequency ridge trajectory will experience high-frequency jitter due to local stress concentration, and the curvature mutation density will rise from the normal value of 30 per second to 45 per second, triggering an early warning; in the severe wear stage, the ridge trajectory will show periodic fractures, and the curvature mutation density will exceed 60 per second, and it will be determined that immediate shutdown maintenance is required. The adaptive update mechanism of the dynamic reference standard ensures that the detection standard is dynamically adjusted with the aging trend of the equipment to avoid misjudgment. In the stage of stable grinding pressure parameters, the dynamic time window is extended, the smoothness of the ridge trajectory is improved, and the curvature mutation density is stabilized within the normal range; in the stage of severe load fluctuations, the window is automatically shortened, and the high-frequency mutation characteristics are accurately captured, thereby achieving real-time and accurate assessment of the health status of the bearing.

[0095] S5. Adjust the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard. The specific implementation is as follows:

[0096] The amplitude threshold of the bearing vibration signal is adjusted based on the normal fluctuation range of the curvature mutation density in the dynamic reference standard. The specific rules for adjusting the amplitude threshold include mapping the upper limit of the normal fluctuation range to the upper limit of the amplitude threshold, and mapping the lower limit to the lower limit of the amplitude threshold. The upper and lower limits of the normal fluctuation range are obtained by fitting historical data. The fitting method is to record the maximum and minimum values ​​of the curvature mutation density during the historical normal operation period and use them as the initial values ​​of the upper and lower limits, respectively. The upper and lower limits are then dynamically adjusted based on the distribution of the curvature mutation density within the current dynamic time window. For example, if the historical normal fluctuation range is 15 to 45 seconds, and the curvature mutation density in the current window is concentrated between 20 and 40 seconds, the upper limit of the adjusted amplitude threshold corresponds to 40 seconds, and the lower limit corresponds to 20 seconds. The mapping relationship between the amplitude thresholds is achieved through linear interpolation, with the input of the curvature mutation density value and the output being the vibration signal amplitude threshold.

[0097] The frequency threshold of the bearing vibration signal is adjusted based on the abnormality threshold for the curvature mutation density in the dynamic reference standard. The specific rules for adjusting the frequency threshold include mapping the curvature mutation density value corresponding to the abnormality threshold to the trigger point of the frequency threshold. The abnormality threshold is obtained by recording the vibration signal's main frequency offset when the curvature mutation density reaches a preset critical value (e.g., 60 per second) during the accelerated wear test of the bearing and using this as the frequency threshold. For example, when the curvature mutation density reaches 60 per second, the main frequency offset is typically ±50 Hz, so the frequency threshold is set to ±50 Hz. During real-time monitoring, the frequency threshold is adjusted synchronously with the dynamic update of the abnormality threshold, and the adjustment range is determined by the standard deviation of the main frequency offset within the current window. For example, if the standard deviation of the main frequency offset is 5 Hz, the frequency threshold is adjusted to ±(50 + 5) = ±55 Hz.

[0098] Generate dynamic amplitude-frequency joint judgment rules based on the adjusted amplitude and frequency thresholds. The generation logic of the dynamic amplitude-frequency joint judgment rules includes: defining the combination conditions of the amplitude threshold and the frequency threshold, including three judgment scenarios: amplitude overlimit, frequency overlimit, and composite overlimit. The amplitude overlimit judgment condition is that the vibration signal amplitude exceeds the upper limit of the amplitude threshold or is lower than the lower limit; the frequency overlimit judgment condition is that the main frequency offset exceeds the frequency threshold; the composite overlimit judgment condition is that both the amplitude and frequency exceed the limit. For example, if the amplitude threshold is 0.8 to 2.5 mm / s² and the frequency threshold is ±50 Hz, then the composite overlimit condition is that the amplitude is >2.5 mm / s² and the main frequency offset is >+50 Hz. The priority of the dynamic amplitude-frequency joint judgment rule is set to: composite overlimit > amplitude overlimit > frequency overlimit to ensure that high-risk conditions are given priority alarms.

[0099] The dynamic amplitude-frequency joint judgment rule is matched with the bearing vibration signal in the current dynamic time window in real time, and the effective interval of the amplitude threshold and frequency threshold is updated. The updating method of the effective interval includes setting the effective action time of the threshold according to the sliding step of the dynamic time window and the load change trend. For example, if the sliding step of the dynamic time window is 0.05 seconds and the grinding pressure is in the rising stage, the effective interval is set to within 0.03 seconds after the end of the current window; if the pressure tends to be stable, the effective interval is extended to 0.08 seconds. During the real-time matching process, the amplitude and main frequency offset of the vibration signal are compared with the threshold point by point in each effective interval, and the matching result is output. The update cycle of the effective interval is strictly synchronized with the sliding step of the dynamic time window to ensure that the threshold action time matches the load fluctuation in real time.

[0100] During the adjustment of the amplitude threshold, if the normal fluctuation range of the curvature mutation density changes suddenly, a threshold smoothing mechanism is triggered. This smoothing mechanism is implemented by calculating the difference between the old and new thresholds, evenly distributing the difference across a preset number of transition windows, and gradually transitioning to the new threshold. For example, if the upper amplitude threshold is adjusted from 2.5 mm / s² to 3.0 mm / s², and the number of transition windows is set to 5, the adjustment increment for each window is (3.0 - 2.5) / 5 = 0.1 mm / s², and this increment continues until the new threshold is reached. The number of transition windows is dynamically set based on the rate of change of the curvature mutation density; the faster the rate of change, the fewer transition windows are used. For example, if the rate of change exceeds 10 per second, the number of transition windows is set to 3; if the rate is less than 5 per second, the number of windows is set to 10.

[0101] When generating dynamic amplitude-frequency joint judgment rules, the judgment weights for the amplitude and frequency thresholds are dynamically assigned. The weight assignment rules include adjusting the judgment weights for amplitude and frequency based on the signal-to-noise ratio of the vibration signal within the current dynamic time window. The signal-to-noise ratio is calculated as the ratio of total signal energy to noise energy (total energy minus energy in the characteristic frequency band), expressed in decibels after taking the logarithm. For example, when the signal-to-noise ratio is 20 decibels, the amplitude weight is set to 0.6 and the frequency weight is set to 0.4; when the signal-to-noise ratio is 10 decibels, the amplitude weight is reduced to 0.4 and the frequency weight is increased to 0.6. The weight coefficients are determined by minimizing the false positive rate in historical data. Specifically, different weight combinations are iterated through the historical data set and the weight that results in the lowest false positive rate is selected as the current weight.

[0102] Through the above implementation, the dynamic adjustment of the amplitude threshold and frequency threshold can accurately adapt to the real-time changes in the bearing wear state. For example, during the stage of drastic fluctuations in the grinding pressure parameters, the density of curvature mutations increases rapidly, the upper limit of the amplitude threshold is increased to 3.0 mm / s² to tolerate transient shocks, and the frequency threshold is simultaneously tightened to ±55 Hz to capture high-frequency anomalies; during the pressure stable stage, the lower limit of the amplitude threshold is reduced to 0.5 mm / s² to improve detection sensitivity, and the frequency threshold is relaxed to ±60 Hz to reduce false alarms. The dynamic amplitude-frequency joint judgment rule and the real-time matching mechanism of the effective interval ensure the stability and reliability of the online monitoring system under complex working conditions.

[0103] S6. Based on the adjusted amplitude threshold and frequency threshold, perform amplitude-frequency joint analysis on the characteristic frequency band components and output the bearing wear status assessment result. The specific implementation is as follows:

[0104] The time domain energy mean is calculated for the characteristic frequency band components within the dynamic time window, and the frequency domain main frequency offset is extracted. The method for calculating the time domain energy mean includes: integrating the time domain signal of the characteristic frequency band components within the dynamic time window, with the integration interval being the start and end time of the current dynamic time window, and dividing the integration result by the window length to generate the time domain energy mean. For example, if the dynamic time window length is 0.1 seconds, and the total energy of the time domain signal amplitude of the characteristic frequency band components integrated within the window is 0.25 (unit: mm² / s³), then the time domain energy mean is 0.25 / 0.1 = 2.5 mm² / s². The method for extracting the frequency domain main frequency offset includes: performing a fast Fourier transform on the characteristic frequency band components within the dynamic time window, setting the spectral resolution to the inverse of the window length, identifying the frequency component with the highest energy in the spectrum as the main frequency, and calculating the difference between the main frequency and the device calibration reference frequency as the main frequency offset. For example, if the base frequency is 1000 Hz and the current main frequency is 1050 Hz, the main frequency offset is +50 Hz.

[0105] Determine whether the time-domain energy mean exceeds the adjusted amplitude threshold and whether the frequency-domain dominant frequency offset exceeds the adjusted frequency threshold. The amplitude threshold is determined by the time-domain energy mean being greater than the upper or lower limit of the adjusted amplitude threshold; the frequency threshold is determined by the absolute value of the dominant frequency offset being greater than the absolute value of the adjusted frequency threshold. For example, if the adjusted amplitude threshold is 0.8 to 2.5 mm / s² and the frequency threshold is ±50 Hz, if the time-domain energy mean is 3.0 mm / s² and the dominant frequency offset is +60 Hz, both limits are exceeded. The judgment logic requires that the limit-exceeding condition be met for a period of time: the limit-exceeding condition must persist for more than 50% of the dynamic time window length to avoid false positives due to transient interference. For example, within a 0.1-second window length, the limit-exceeding condition must persist for more than 0.05 seconds to trigger a valid judgment.

[0106] When both the time-domain energy mean and the frequency-domain main frequency offset exceed limits simultaneously, a bearing wear pattern is generated based on a combined encoding of the amplitude and frequency thresholds. The combined encoding is generated by combining the amplitude excess direction (exceeding the upper limit or falling below the lower limit) with the frequency excess direction (positive or negative) into a four-bit code in the format of "amplitude direction + frequency direction + threshold level." For example, if the amplitude exceeds the upper limit and the frequency exceeds the limit in the positive direction, the code is "A1+"; if the amplitude falls below the lower limit and the frequency exceeds the limit in the negative direction, the code is "B2-." The bearing wear pattern is defined by mapping the codes to a predefined wear type library. The wear type library is trained using historical fault data. The training method is to count the occurrence frequency of each code in known fault cases and associate the codes with frequencies exceeding the set threshold with the fault type. For example, if the code "A1+" appears in 90% of outer ring wear cases, it is mapped to "outer ring wear."

[0107] Based on the cumulative results of bearing wear patterns in the current dynamic time window and adjacent windows, the bearing wear condition assessment result is output. The statistical method for the cumulative results includes using a sliding window counter to record the wear patterns in the current window and the two previous windows. The counter has a capacity of 3, and new data overwrites the oldest data as it enters. The proportion of occurrences of the same wear pattern in the three windows is calculated using the formula: number of occurrences / 3 × 100%. For example, if the current window is "outer ring wear" and the previous two windows are "outer ring wear" and "inner ring wear," respectively, the proportion is 2 / 3, which is approximately 66.7%. The output rule for the assessment result is: if the proportion of identical patterns exceeds 70%, the wear condition is determined to be confirmed; if the proportion is between 50% and 70%, the wear condition is suspected; and if it is below 50%, the wear condition is considered normal fluctuation. The thresholds of 70% and 50% are determined based on the distribution intervals between normal and abnormal conditions in historical data. For example, under normal conditions, the proportion of identical patterns is typically less than 30%, while under abnormal conditions, it exceeds 70%.

[0108] When generating bearing wear patterns, if multiple overrun combinations occur within the same window, a priority determination rule is used to determine the final pattern. The priority rule defines the order of the combination codes by overrun severity, with severity determined by the weighted sum of the amplitude and frequency overruns, with a weighting of 0.6 for amplitude and 0.4 for frequency. For example, if the code "A1+" has an amplitude overrun of 3.0-2.5 = 0.5 mm / s² and a frequency overrun of 60-50 = 10 Hz, the severity score is 0.5 × 0.6 + 10 × 0.4 = 4.3. The code "B2-" has a score of 0.3 × 0.6 + 8 × 0.4 = 3.38. Therefore, the higher-scoring "A1+" is selected as the current pattern. The priority rule weighting is determined through regression analysis of historical fault data, with the goal of minimizing the rate of pattern misjudgment.

[0109] The updating and storage logic for the cumulative results includes using a circular buffer to store the wear patterns of the last three dynamic time windows, with the buffer index updated on a first-in, first-out basis. For example, buffer index 0 represents the earliest data, and index 2 represents the latest data. When new data enters, the data at index 0 is overwritten, and indexes 1 and 2 move forward in sequence. The output cycle of the evaluation results is strictly synchronized with the sliding step of the dynamic time window. For example, if the window sliding step is 0.05 seconds, the evaluation results are updated every 0.05 seconds. The output results are in the form of text descriptions (such as "outer ring wear"), numerical levels (such as "wear level 3"), and control signals (such as "stop command"). The control signals are transmitted to the equipment control system via an industrial communication protocol.

[0110] Through the above implementation, the assessment results of bearing wear status can reflect wear trends under dynamic loads in real time. For example, if the "outer ring wear" pattern appears twice within three consecutive dynamic time windows and the time-domain energy mean continues to rise, the assessment output is "Increased outer ring wear, recommended shutdown for inspection." If "outer ring wear," "inner ring wear," and "outer ring wear" appear in three windows, and the proportion is 66.7%, the output is "Combined wear warning, further diagnosis required." The verification mechanism for the assessment results includes recording the actual wear status during equipment maintenance, comparing it with the system output, calculating the false positive rate, and dynamically optimizing the threshold and weight coefficients. For example, if outer ring wear is detected during actual maintenance but the system does not trigger an alarm, the threshold is adjusted from 70% to 65% to increase sensitivity.

[0111] Example 2: Figure 2 The present invention provides a structural schematic diagram of an online monitoring system for the wear status of a grinding mill bearing, which includes the following modules:

[0112] Signal acquisition module, used to obtain grinding pressure parameters and bearing vibration signals when the mill is running;

[0113] Dynamic correlation module, used to analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristics of the bearing vibration signal;

[0114] The frequency band separation module is used to separate the interference frequency band components and characteristic frequency band components in the bearing vibration signal according to the main frequency characteristic offset;

[0115] The curvature analysis module is used to track the time-frequency ridges of characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, and calculate the curvature mutation density of the ridge trajectory. The dynamic reference standard of the bearing wear characteristics is determined based on the curvature mutation density.

[0116] A threshold adjustment module, used to adjust the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard;

[0117] The condition assessment module is used to perform amplitude-frequency joint analysis on the characteristic frequency band components based on the adjusted amplitude threshold and frequency threshold, and output the bearing wear condition assessment results.

[0118] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to actual conditions.

[0119] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0120] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0121] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0122] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0123] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0124] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for online monitoring of bearing wear status of a grinding mill, characterized in that: The steps include: S1, obtaining the grinding pressure parameters and bearing vibration signals of the grinding mill during operation; S2. Analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristic of the bearing vibration signal; S3, separating the interference frequency band component and the characteristic frequency band component in the bearing vibration signal according to the main frequency characteristic offset; S4. Track the time-frequency ridges of the characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, and calculate the curvature mutation density of the ridge trajectory. Determine the dynamic reference standard of the bearing wear characteristics based on the curvature mutation density, including: The time-frequency ridge line of the time domain signal of the characteristic frequency band component is tracked in the dynamic time window to generate the ridge line trajectory of the time-frequency energy distribution; Perform cubic spline interpolation on the ridge trajectory and calculate the curvature change rate of the interpolated trajectory; count the number of mutation points where the curvature change rate exceeds the set mutation threshold per unit time to generate the curvature mutation density; Based on the mapping relationship between the curvature mutation density and the bearing wear degree in the historical wear data, the dynamic reference standard of the bearing wear characteristics in the current dynamic time window is determined; Dynamic reference standards include the normal fluctuation range of curvature mutation density and the abnormality judgment threshold; S5. Adjust the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard, including: Adjust the amplitude threshold of the bearing vibration signal according to the normal fluctuation range of the curvature mutation density in the dynamic reference standard; Adjust the frequency threshold of the bearing vibration signal according to the abnormal judgment threshold of the curvature mutation density in the dynamic reference standard; Generate a dynamic amplitude-frequency joint judgment rule based on the adjusted amplitude threshold and frequency threshold; The dynamic amplitude-frequency joint judgment rule is matched with the bearing vibration signal in the current dynamic time window in real time, and the effective range of the amplitude threshold and frequency threshold is updated; S6. Based on the adjusted amplitude threshold and frequency threshold, perform amplitude-frequency joint analysis on the characteristic frequency band components and output the bearing wear status assessment result.

2. A method for online monitoring of bearing wear status of a grinding mill according to claim 1, characterized in that: Obtain grinding pressure parameters and bearing vibration signals during mill operation, including: Grinding pressure parameters are collected through a pressure sensor installed on the grinding roller of the mill, and bearing vibration signals are collected through an acceleration sensor installed on the bearing seat of the mill; The grinding pressure parameters and the bearing vibration signals are synchronously sampled to generate a time-aligned grinding pressure parameter sequence and a bearing vibration signal sequence.

3. A method for online monitoring of bearing wear status of a grinding mill according to claim 2, characterized in that: The acceleration sensor is arranged in the horizontal and vertical directions of the bearing seat, and the pressure sensor is connected with the hydraulic adjustment mechanism of the grinding roller.

4. The method for online monitoring of bearing wear status of a grinding mill according to claim 1, wherein: Analyze the corresponding relationship between the change of grinding pressure parameters and the main frequency characteristic offset of the bearing vibration signal, including: A dynamic time window is generated based on the time-aligned grinding pressure parameter sequence, and the interval length of the dynamic time window is inversely proportional to the change rate of the grinding pressure parameter; Extract the main frequency energy distribution and phase synchronization index of the bearing vibration signal sequence within the dynamic time window; According to the joint change trend of the main frequency energy distribution and the phase synchronization index, a nonlinear mapping rule between the grinding pressure parameter change and the main frequency characteristic offset is constructed; The nonlinear mapping rule is used to indicate the confidence interval of the main frequency characteristic offset under dynamic load fluctuation.

5. The method for online monitoring of bearing wear status of a grinding mill according to claim 1, wherein: The interference frequency band components and characteristic frequency band components in the bearing vibration signal are separated according to the main frequency characteristic offset, including: Determine the dynamic frequency band division boundary based on the main frequency characteristic offset and nonlinear mapping rules; According to the dynamic frequency band division boundary, the frequency band whose energy ratio in the bearing vibration signal is lower than the main frequency energy distribution threshold and whose phase synchronization index is negative is extracted as the interference frequency band component; Extract the frequency band whose energy ratio is higher than the main frequency energy distribution threshold and whose phase synchronization index is positive as the characteristic frequency band component; Complementary filtering is performed on the interference frequency band components and the characteristic frequency band components to generate a separated signal sequence aligned with the dynamic time window.

6. The method for online monitoring of bearing wear status of a grinding mill according to claim 1, characterized in that: Based on the adjusted amplitude and frequency thresholds, the characteristic frequency band components are analyzed in a combined amplitude and frequency manner to output the bearing wear status assessment results, including: Calculate the time domain energy mean and extract the frequency domain main frequency offset of the characteristic frequency band components in the dynamic time window; Determine whether the time domain energy mean exceeds the adjusted amplitude threshold and whether the frequency domain main frequency offset exceeds the adjusted frequency threshold; When both the time-domain energy mean and the frequency-domain main frequency offset exceed the limit, the bearing wear pattern is generated according to the combined coding of the exceeding amplitude threshold and the frequency threshold; Based on the cumulative results of bearing wear patterns in the current dynamic time window and adjacent windows, the bearing wear status assessment results are output.

7. The method for online monitoring of bearing wear status of a grinding mill according to claim 6, characterized in that: The bearing wear status evaluation result is the proportion of the occurrence of the same wear pattern in three consecutive dynamic time windows.

8. An online monitoring system for the wear state of a grinding mill bearing, used to implement the online monitoring method for the wear state of a grinding mill bearing according to any one of claims 1 to 7, characterized in that: Includes the following modules: Signal acquisition module, used to obtain grinding pressure parameters and bearing vibration signals when the mill is running; Dynamic correlation module, used to analyze the corresponding relationship between the change of grinding pressure parameters and the offset of the main frequency characteristics of the bearing vibration signal; The frequency band separation module is used to separate the interference frequency band components and characteristic frequency band components in the bearing vibration signal according to the main frequency characteristic offset; The curvature analysis module is used to track the time-frequency ridges of characteristic frequency band components, extract the ridge trajectory of the time-frequency energy distribution, and calculate the curvature mutation density of the ridge trajectory. The dynamic reference standard of the bearing wear characteristics is determined based on the curvature mutation density. A threshold adjustment module, used to adjust the amplitude threshold and frequency threshold of the bearing vibration signal according to the dynamic reference standard; The condition assessment module is used to perform amplitude-frequency joint analysis on the characteristic frequency band components based on the adjusted amplitude threshold and frequency threshold, and output the bearing wear condition assessment results.

Images