A bearing fault early warning data processing method based on operational state constraints
By constructing a fault-sensitive and operational disturbance segment pair matrix under operational state constraints, effective bearing fault early warning indicators are screened out, solving the problem of misjudgment in bearing fault early warning methods when equipment operating conditions change, and achieving accurate fault early warning.
Patent Information
- Application Number
- CN202611579169.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-13
- Publication Date
- 2026-11-10
AI Technical Summary
Existing bearing fault early warning methods have difficulty effectively distinguishing between bearing fault-sensitive changes and operational disturbances when equipment operating conditions change, leading to misjudgments or information loss.
By establishing operational state constraints, constructing fault-sensitive segment pairs and operational disturbance segment pairs, we can screen out bearing fault early warning indicators that are less affected by operational disturbances and have effective distinguishing capabilities, and confirm the early warning based on the monitoring results under the same operational state.
It effectively distinguishes between bearing failure-sensitive changes and operational disturbance changes, reduces misjudgments, and ensures the accuracy and reliability of early warning indicators.
Smart Images

Figure CN123065756A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of bearing condition monitoring data processing technology, specifically a bearing fault early warning data processing method based on operational condition constraints. Background Technology
[0002] Bearings are important supporting components in rotating equipment. Existing bearing fault early warning methods typically extract time-domain, frequency-domain, or time-frequency-domain indicators from monitoring time-series data such as vibration and current, and determine the bearing's operating status based on changes in these indicators relative to normal operating reference values.
[0003] Besides being affected by changes in the bearing's own condition, bearing monitoring indicators also change with operating conditions such as equipment speed, load, and control status. Especially in equipment with start-stop, acceleration, deceleration, reversing, or braking processes, differences in operating conditions or disturbances during operation can cause variations in vibration amplitude, impact intensity, and frequency band energy. Even when the bearing condition remains unchanged, some candidate indicators may show significant differences. While existing methods employ techniques such as operating condition classification, feature normalization, reference benchmark establishment, and threshold judgment to reduce the impact of changes in operating conditions, simply selecting indicators based on the magnitude of differences or sensitivity may still lead to the selection of indicators primarily reflecting operational disturbances as bearing fault early warning indicators.
[0004] For indicators that simultaneously exhibit both bearing fault sensitivity and operational disturbance response, direct deletion may result in the loss of fault information, while direct retention may lead to misjudgment. Furthermore, using uniform reference values or adjacent continuous windows for early warning confirmation is easily affected by normal fluctuations under different operational conditions or by the same operational disturbance. Therefore, it is necessary to distinguish between bearing fault sensitivity changes and operational disturbance changes within the differences in candidate indicators under operational condition constraints, and to confirm fault early warnings based on monitoring results under the same operational conditions. Summary of the Invention
[0005] To address the aforementioned issues, this invention provides a bearing fault early warning data processing method based on operational state constraints. This method distinguishes between bearing fault-sensitive changes and operational disturbance changes among candidate indicator differences, confirms overlapping indicator differences, filters bearing fault early warning indicators that are less affected by operational disturbances and have effective distinguishing capabilities, and confirms fault early warnings based on monitoring results under the same operational state.
[0006] To achieve the above objectives, the present invention provides the following technical solution, which specifically includes the following steps: Steps S1 to S4 are used to establish reference information and filter bearing fault early warning indicator groups, while steps S5 and S6 are used to perform online early warning processing on the current status monitoring window.
[0007] S1: Acquire the monitoring time-series data of the bearing to be monitored and the equipment operating status data. Divide the monitoring time-series data into status monitoring windows, and determine the operating status of each status monitoring window based on the equipment operating status data. Mark each status monitoring window according to the operating status, defining one status monitoring window as a segment, or merging multiple status monitoring windows with the same operating status and consecutive time into a segment. Extract data from each status monitoring window. A candidate indicator set is formed from 10 candidate indicators. ,in, Indicates the first One candidate indicator, .
[0008] The equipment operating status data includes state variables characterizing the equipment's operating conditions, such as speed, torque, radial force, load, current level, displacement, and control or braking status. When a work cycle identifier is available, the equipment operating status data also includes the work cycle identifier. This equipment operating status data is used to determine the work status and to construct a work disturbance segment pair matrix. State variables directly characterizing the equipment's operating conditions are not considered as candidate indicators in the coverage search. Indicators extracted from vibration signals, current signals, or other bearing monitoring time-series data are used as candidate indicators in the coverage search.
[0009] S2: Calculate the segment difference statistics of each candidate indicator between different segments, and construct the candidate indicator segment difference matrix. Construct a fault-sensitive segment pair matrix based on bearing fault-sensitive reference information. Construct operation disturbance segments based on equipment operating status data to pair the matrix. The difference threshold for candidate indicators is determined based on normal operation reference segments under the same operating conditions. The operation disturbance segment pair matrix is used to characterize the segment pair differences caused by differences in equipment operating conditions or operation process disturbances.
[0010] S3: Match the segment differences of the candidate indicators with the fault-sensitive segment pair matrix and the operational disturbance segment pair matrix, respectively, to obtain fault-sensitive matching and operational disturbance matching. Confirm the segment differences that overlap between the two, and construct the effective sensitive segment matrix based on the matching results and confirmation results. .
[0011] S4: Based on the effective sensitive segment matrix of each candidate index For fault-sensitive segments to the matrix A constrained coverage search was performed to obtain the bearing fault early warning index set. Coverage search determines the admission criteria for candidate indicators based on the number of newly added coverage indicators, the proportion of unconfirmed overlaps in operational disturbances, and the effective discrimination. The selection order of candidate indicators is determined by the number of newly added coverage indicators, and the weight of each candidate indicator is determined based on its number of newly added coverage indicators, effective discrimination, and the proportion of unconfirmed overlaps in operational disturbances. State variables that directly characterize equipment operating conditions are not considered as candidate indicators in the coverage search.
[0012] S5: During online monitoring, the current status monitoring window will be displayed. Compared with the normal operating reference window under the same operating conditions, based on the bearing fault early warning index group The deviation of each indicator from the normal operating reference value is calculated within the fault deviation window. The fault sensitivity matching for the current window is determined based on the standardized deviation of the selected indicators, the operational disturbance matching for the current window is determined based on the differences in equipment operating status data, and the overlap difference confirmation result for the current window is determined according to the method described in step S3. Based on this, the early warning confirmation quantity is calculated. .
[0013] S6: Based on window fault deviation , number of confirmed early warnings The system confirms the current status of the monitored window with the corresponding work status. When the current status of the monitored window meets the warning conditions, and the number of windows with the same status that meet the warning conditions reaches a preset confirmation number, a warning level is output; otherwise, a warning level or a normal level is output based on the window's fault deviation.
[0014] As a preferred technical solution, step S2 specifically includes the following steps: S2.1: For any candidate index Calculate fragments With fragments Segment difference statistics The expression is as follows: in, and Each represents a segment and fragments , The statistical method for representing the difference between two segment index sequences is based on the sum of the values of each segment pair. Composition of candidate indicator fragment difference matrix The expression is as follows: Candidate indicators were determined based on the statistical differences in normal reference segments under the same operating conditions. Difference threshold The expression is as follows: in, and Each represents a segment and fragments The corresponding job status Representing fragments With fragments This is a normal operating reference fragment pair. Represents the quantile function. ; S2.2: Establish a fault-sensitive segment pair matrix based on preset bearing fault-sensitive reference information. ; The bearing fault sensitive reference information includes preset bearing fault characteristic frequency band energy, local peak value of envelope spectrum, frequency rotation and its harmonic energy, frequency rotation sideband energy, high frequency impact energy or periodic impact interval. The fault-sensitive segment pair matrix is used to characterize segments that exhibit changes in bearing fault sensitivity under the same operating conditions. Let the first segment be... Bearing-like fault-sensitive reference information in fragments With fragments The difference between them is ,but: in, Determine the threshold for fault-sensitive segments; S2.3: Establish a work disturbance segment pair matrix based on the equipment operating status data. ; The operation disturbance segment pair matrix is used to characterize the segment pair differences caused by differences in equipment operating conditions or operation process disturbances. Let the first segment pair be... The equipment operating status data in the fragment With fragments The difference in response between them is ,but: in, Determine the threshold for the disturbance segment of the operation.
[0015] As a preferred technical solution, step S3 specifically includes the following steps: S3.1: Constructing candidate indices Fault-sensitive matching matrix The expression is as follows: in, Indicate candidate indicators In fragment pairs The difference in the fragments reaches a threshold, and the fragment pair belongs to a fault-sensitive fragment pair. This is a statistic of segment differences. For fault-sensitive segment pairs matrix, , The index number of the segment; S3.2: Constructing candidate indices The operation perturbation matching matrix The expression is as follows: in, Indicate candidate indicators In fragment pairs The difference in the data reaches the threshold, and the segment pair belongs to the operational disturbance segment pair. For the perturbation fragment pair matrix, The difference threshold; S3.3: For and For fragment pairs, calculate the fragment pair confirmation count. The expression is as follows: in, Indicates the type of bearing fault-sensitive reference information. This is the amount of information used to confirm the location of the characteristic frequency band. This is the confirmed amount for the periodic impact interval. This refers to the number of confirmations that occur repeatedly under the same work status or in different work cycles. The relative enhancement confirmation value for the reference measuring point. , , and Non-negative weights It is a positive number; according to Constructing the overlap difference confirmation matrix The expression is as follows: in, To confirm the threshold; right and When the segments are compared and the differences are confirmed, And the number of characteristic frequency band locations confirmed Or periodic impact interval confirmation quantity If one of the conditions is met, the confirmation quantity occurs repeatedly in the same work status or different work cycles. Or the relative enhancement confirmation value of the reference measurement point When one of them satisfies the corresponding condition, it makes And retain the candidate index differences of the segment pair as effective sensitive differences. If the confirmation conditions are not met, make ; The confirmation conditions include: at least one of the characteristic frequency band location confirmation quantity or periodic impact interval confirmation quantity meets the corresponding condition, and at least one of the repeated occurrence confirmation quantity or the relative enhancement confirmation quantity of the reference measurement point meets the corresponding condition; S3.4: Construct candidate indices according to the following formula Effective sensitive segment matrix The expression is as follows: in, This represents element-wise multiplication. This indicates an element-wise logical OR operation. This indicates that the segment will be... Differences in candidate indicators are retained as valid sensitive differences.
[0016] As a preferred technical solution, step S4 specifically includes the following steps: S4.1: Will Fragment pairs as the set to be covered and will Initialize to an empty set; S4.2: For candidate indicators that were not selected Calculate its set to be covered New coverage And calculate the proportion of unconfirmed overlap of operational disturbances and the proportion of confirmed overlap differences. and effective discrimination ; Calculate its set to be covered New coverage The expression is as follows: in, , The index number of the segment. For effective sensitive segment matrix; The formula for calculating the proportion of unconfirmed overlap of work disturbances is as follows: The formula for calculating the overlap difference confirmation ratio is as follows: The formula for calculating effective discrimination is as follows: in, Indicate candidate indicators The proportion of fault-sensitive matching that overlaps with operational disturbances and fails to pass confirmation. Indicate candidate indicators The proportion of overlapping operational disturbances that has been confirmed. Indicate candidate indicators Compared to the ability to effectively distinguish normal fluctuations in the same working state. Representing fragments With fragments This is a normal operating reference fragment pair. It is a positive number; when the candidate index There is no satisfying When the fragment pairs are equal, the effective discrimination of the candidate indicator is improved. Furthermore, this candidate metric is not included in subsequent coverage searches.
[0017] S4.3: When satisfying , and Among the candidate indicators, select The largest candidate indicator was added to the bearing failure early warning indicator group. and update the set to be covered. ,in, This indicates the threshold for the proportion of overlapping operational disturbances that have not been confirmed. Indicates the effective discrimination threshold; S4.4: Remove already covered fragment pairs from the set to be covered. Remove from the middle and calculate the coverage ratio. : in, Bearing Fault Early Warning Indicator Group The corresponding effective sensitive segment matrix is composed of Effective sensitive segment matrix of each candidate indicator The element-wise logical OR operation yields the following expression: Indicates fragment pairs Bearing Failure Early Warning Indicator Group At least one candidate indicator should be used as an effective sensitive difference coverage. It is a positive number.
[0018] when The number of selected indicators has reached the preset upper limit. Or there is no such condition. , and When candidate indicators are found, stop the coverage search and obtain the bearing fault early warning indicator group. ,in, This indicates the coverage ratio threshold.
[0019] Compared with the prior art, the present invention has the following beneficial effects: 1) This invention matches the segment differences of candidate indicators with fault-sensitive segment pairs and operational disturbance segment pairs respectively, and confirms the segment differences that overlap between the two. This allows the invention to retain the indicator differences that are confirmed to still be sensitive to bearing faults, thereby reducing the number of indicators that mainly reflect operational disturbances entering the bearing fault early warning indicator group.
[0020] 2) This invention calculates the window fault deviation based on the normal operation reference value under the same operating conditions, and calculates the early warning confirmation amount based on the effective sensitive confirmation result of the current window. Then, it uses the window under the same operating conditions for confirmation, so that the monitoring windows under different operating conditions adopt the corresponding reference information and avoids directly merging and counting windows under different operating conditions. Attached Figure Description
[0021] Figure 1 This is a schematic diagram of the overall process of the method proposed in this invention; Figure 2 This is a schematic diagram illustrating the changes in window fault deviation, early warning confirmation, and bearing condition level with the condition monitoring window of the present invention. Figure 2 (a) is a schematic diagram of the window fault deviation. Figure 2 (b) is a schematic diagram of the early warning confirmation quantity. Figure 2 (c) is a schematic diagram showing how the bearing condition level changes with the condition monitoring window. Detailed Implementation
[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
[0023] like Figure 1 The diagram shown is a flowchart illustrating a specific implementation method for bearing fault early warning data processing based on operational state constraints. The method of this invention includes the following steps: S1: Monitoring data acquisition, status monitoring window division and segment determination; Acquire the monitoring time-series data of the bearing to be monitored and the equipment operating status data. Divide the monitoring time-series data into status monitoring windows, and determine the operating status of each status monitoring window based on the equipment operating status data. Mark each status monitoring window according to the operating status, and define a status monitoring window as a segment, or merge multiple status monitoring windows with the same operating status and consecutive time into a segment.
[0024] In this embodiment, the monitoring time series data is the vibration monitoring data of the bearing to be monitored.
[0025] The equipment operating status data is used to characterize the equipment operating conditions, including speed, torque, radial force, load, current level, displacement, control status, or braking status.
[0026] The operational status can be determined based on one piece of equipment operational status data or a combination of multiple pieces of equipment operational status data. When the equipment has repeatable operating processes, a work cycle identifier is set for the status monitoring window to distinguish the same operational status in different work cycles.
[0027] Extract from each status monitoring window A set of candidate indicators is formed from 10 candidate indicators: in, Indicates the first One candidate indicator, , The total number of candidate indicators. This serves as an index for candidate indicators. Candidate indicators include time-domain statistical indicators, frequency-domain energy indicators, envelope spectrum indicators, frequency conversion and its harmonics indicators, frequency conversion sideband indicators, high-frequency impulse indicators, or periodic impulse indicators. State variables that directly characterize equipment operating conditions are used to determine the operating status and operational disturbances and are not included as candidate indicators in subsequent coverage searches.
[0028] The length of the condition monitoring window is determined based on the sampling frequency, bearing speed, and required frequency resolution. Adjacent condition monitoring windows are set consecutively or according to a preset step size. When multiple condition monitoring windows are merged into a segment, each condition monitoring window within the same segment has the same operating status.
[0029] S2: Fragment Difference and Fragment Pair Matrix Construction: S2.1: Construct the difference matrix of candidate index segments Fault-sensitive segments on the matrix and operation perturbation fragments to matrix The specific steps are as follows: For any candidate index Calculate fragments With fragments Segment difference statistics : in, and Each represents a segment and fragments , A statistical method for representing the difference between two segment index sequences.
[0030] From each fragment pair Composition of candidate indicator fragment difference matrix : Candidate indicators were determined based on the statistical differences in normal reference segments under the same operating conditions. Difference threshold : in, and Each represents a segment and fragments The corresponding job status Representing fragments With fragments This is a normal operating reference fragment pair. Represents the quantile function. , , The index number of the segment; S2.2: Establish a fault-sensitive segment pair matrix based on preset bearing fault-sensitive reference information. The fault-sensitive segment pair matrix is used to characterize segments that exhibit changes in bearing fault sensitivity under the same operating conditions.
[0031] The fault-sensitive segment pairs matrix The fault-sensitive reference information is determined based on the statistical difference between the fault-sensitive reference information and the segment difference statistics of the current candidate indicator to be evaluated. This fault-sensitive reference information is determined from other monitoring features, status records, maintenance records, or preset bearing fault feature information. When using normal operation reference data, the fault-sensitive segment pair matrix is determined based on the change of the bearing fault-sensitive reference information relative to the normal operation reference data. ; Record No. Bearing-like fault-sensitive reference information in fragments With fragments The difference between them is Then the fault-sensitive segment pair matrix: in, Determine the threshold for fault-sensitive segments. Indicates the first Bearing-like fault-sensitive reference information in fragments With fragments The difference between them This indicates the type of bearing fault-sensitive reference information. The bearing fault-sensitive reference information is independent of the segment differences of the current candidate indicator being evaluated, and includes preset bearing fault characteristic frequency band energy, local peak values in the envelope spectrum, rotational frequency and its harmonics energy, rotational frequency sideband energy, high-frequency impact energy, or periodic impact intervals. When using normal operation reference data, the bearing fault-sensitive reference information is determined based on its variation relative to the normal operation reference data. .
[0032] S2.3: Establish a work disturbance segment pair matrix based on the equipment operating status data. The operation disturbance segment pair matrix is used to characterize the segment pair differences caused by variations in equipment operating conditions or operation process disturbances. Let the first... The equipment operating status data in the fragment With fragments The difference in response between them is ,but: in, Determine the threshold for the operation disturbance segment. For the first The equipment operating status data in the fragment With fragments The difference in response between them This indicates the type index of the device operating status data. Response variance. The difference between two segments is determined based on the variations in speed, torque, radial force, load, current level, displacement, and control or braking status. For equipment operating status data with different dimensions, standardization is first performed based on normal operating data, and then the response difference between segments is calculated.
[0033] Standardized mean difference can be used. A standardized variation relative to the normal operating reference data can be used. Standardized differences can be used for continuous state variables, and state difference identifiers can be used for discrete state variables.
[0034] S3: Construction of the effective sensitive segment matrix: Based on the difference matrix of candidate index segments Statistical measure of the difference between each segment and its corresponding segment Combining fault-sensitive segments with matrix and operation perturbation fragments to matrix The study identifies fault-sensitive matching and operational disturbance matching among the differences in candidate indicators, confirms the overlapping segments of these two types of differences, and constructs candidate indicators. Effective sensitive segment matrix The specific method is as follows: S3.1: Candidate Indicators Fault-sensitive matching matrix for: S3.2: Candidate Indicators The operation perturbation matching matrix for: S3.3: For and For fragment pairs, calculate the fragment pair confirmation count. The calculation expression is as follows: in, This indicates the type of bearing fault-sensitive reference information. The characteristic frequency band location confirmation quantity is determined based on the consistency between the frequency band where the candidate index abnormally enhances and the preset bearing fault characteristic frequency band. The periodic impact interval confirmation value is determined based on the consistency between the impact interval in the monitoring signal and the bearing fault characteristic period. The confirmation quantity for repeated occurrences of the same work status or different work cycles is determined based on the proportion of repeated occurrences of abnormal enhancement of candidate indicators in the same work status or different work cycles. The relative enhancement confirmation value for the reference measuring point is determined based on the degree of response enhancement of the bearing measuring point relative to the reference measuring point. , , , Non-negative weights It is a positive number. When no reference measuring point is set, let... and set the corresponding weights .
[0035] according to Constructing the overlap difference confirmation matrix : in, To confirm the threshold. Confirmation conditions include: characteristic frequency band location confirmation quantity. Or periodic impact interval confirmation quantity At least one of the corresponding conditions must be met, and the confirmation quantity must occur repeatedly in the same work status or different work cycles. Or the relative enhancement confirmation value of the reference measurement point At least one of the conditions must be met; S3.4: Construct candidate indices according to the following formula Effective sensitive segment matrix : in, This represents element-wise multiplication. This indicates an element-wise logical OR operation. This indicates that the segment will be... Differences in candidate indicators are retained as valid sensitive differences. Differences that coincide with operational disturbances and have been confirmed are retained, while differences that coincide with operational disturbances but have not been confirmed are excluded.
[0036] S4: Bearing Failure Early Warning Indicator Group Screening: Based on the effective sensitive segment matrix of each candidate index Fault-sensitive segments pair matrix A constrained coverage search was performed to obtain the bearing fault early warning index set. The specific method is as follows: S4.1: Will Fragment pairs as the set to be covered and will Initialize to an empty set. For candidate indicators that were not selected... Calculate its set to be covered New coverage The expression is as follows: S4.2: Calculate the proportion of unconfirmed overlap of operational disturbances, the proportion of confirmed overlap differences, and the effective discrimination index. in, Indicate candidate indicators The proportion of fault-sensitive matching that overlaps with operational disturbances and fails to pass confirmation. Indicate candidate indicators The proportion of overlapping operational disturbances that has been confirmed. Indicate candidate indicators The ability to effectively distinguish between normal fluctuations in the same operational state. When candidate indicators... There is no satisfying When the fragment pairs are compared, their effective distinguishability is improved. This candidate metric will not participate in subsequent coverage searches.
[0037] S4.3: When satisfying , and Among the candidate indicators, select The largest candidate indicator was added to the bearing failure early warning indicator group. and update the set to be covered. When multiple candidate indicators have the same number of newly added coverages, the candidate indicators with higher effective discrimination, lower proportion of unconfirmed overlap of operational disturbances, and higher proportion of confirmed overlap differences are selected in that order.
[0038] S4.4: Remove already covered fragment pairs from the set to be covered. Remove from the middle and calculate the coverage ratio. The expression is as follows: in, Bearing Fault Early Warning Indicator Group The corresponding effective sensitive segment matrix is composed of Effective sensitive segment matrix of each candidate indicator The element-wise logical OR operation yields the following expression: Indicates fragment pairs Bearing Failure Early Warning Indicator Group At least one candidate indicator should be used as an effective sensitive difference coverage. The value is positive. Coverage search stops when the coverage ratio reaches the preset coverage ratio, the number of selected indicators reaches the preset maximum number, or there are no candidate indicators that meet the admission conditions and have new coverage.
[0039] The weights of candidate indicators are determined jointly based on the number of newly added coverage areas, effective discrimination, and the proportion of unconfirmed overlaps in operational disturbances. In this embodiment, unnormalized weights are calculated. The expression is as follows: Again Normalization is performed to obtain the normalized weights, as shown in the following expression: In the formula, Candidate indicators The weights; When each When all values are zero, the same weight is assigned to each candidate indicator in the bearing fault early warning indicator group.
[0040] When multiple candidate indicators When the two are the same, choose the one that takes precedence. higher Lower Highly regarded candidate indicators. The weights of candidate indicators are determined based on the number of new coverage instances, effective discrimination, and the proportion of unconfirmed overlaps in operational disturbances. .
[0041] S5: Calculation of window fault deviation and early warning confirmation amount: During online monitoring, the current status monitoring window is used. According to the bearing failure early warning index group The deviation of each indicator from the normal operating reference value under the same operating conditions is used to calculate the window fault deviation. The expression is as follows: in, For status monitoring window Internal candidate indicators The deviation from the reference value under the same operating conditions. Candidate indicators The weights; For candidate indicators that change in the direction of increasing bearing failure, the deviation is calculated according to the following formula. The calculation expression is as follows: in, Indicates candidate indicators within the current status monitoring window. The value of , and These represent the work status respectively. Candidate indicators in the normal operation reference window The median and median absolute deviation. For candidate indicators that change towards decreasing bearing failure, the numerator is changed to For the two-way deviation index, the absolute value of the standardized difference is taken.
[0042] If the number of normal operation reference windows in the current operation state is insufficient, select an operation state with similar operating conditions, and determine the backup reference value after correcting for differences in operating conditions.
[0043] View the current status monitoring window By comparing the bearing with a normal operating reference window under the same operating conditions, and following the methods described in S2 and S3, determine the bearing fault early warning index group. Fault-sensitive matching flags for each indicator in the current window Current window job disturbance matching flag Confirmation flag for overlap with current window And calculate the current window's valid sensitivity confirmation flag. The calculation expression is as follows: The current window overlap difference confirmation flag is determined according to the confirmation method in S3, based on the confirmation amount of the characteristic frequency band position of the current state monitoring window, the confirmation amount of the periodic impact interval, the confirmation amount of repeated occurrence of the same working state or different working cycles, and the confirmation amount of the relative enhancement of the reference measuring point when setting the reference measuring point.
[0044] Calculate the number of warning confirmations based on the current window's valid sensitive confirmation flags. The calculation expression is as follows: Window fault deviation threshold and warning confirmation threshold Determined based on a normal operating reference window under the same working conditions. In this embodiment, the window fault deviation threshold is... Reference window for normal operation under the same working conditions The statistical distribution is determined. The threshold for early warning confirmation is determined. According to the normal operation reference window The statistical distribution or the pre-set effective sensitivity confirmation weight ratio is determined.
[0045] S6: Same as operation status confirmation and bearing status level output: Based on window fault deviation , number of confirmed early warnings and status monitoring window The corresponding operating status is confirmed, and the bearing status level is output. Specifically: Select the window that matches the current status monitoring window from the historical status monitoring window. Belonging to the same work status The windows form a set of confirmation windows with the same status. When a work cycle identifier exists, select the appropriate status monitoring window from different work cycles. Windows belonging to the same job status are selected for confirmation, and at most one status monitoring window is selected per job cycle. If no job cycle identifier exists, a preset number of windows with the same job status are selected from the historical status monitoring windows.
[0046] The current status monitoring window meets the requirements. and Furthermore, after including the current status monitoring window, when the number of windows in the current status monitoring window and the set of windows with the same status confirmation window that meet the above conditions reaches a preset confirmation number m, an early warning level is output. The current status monitoring window meets... If the conditions for issuing a warning level are not met, an alert level will be issued. The current status monitoring window meets the requirements. When the time is right, output the normal level.
[0047] When a work cycle identifier exists, the count is based on the number of different work cycles that meet the warning conditions under the same work status; when no work cycle identifier exists, the count is based on the number of historical status monitoring windows that meet the warning conditions under the same work status. The preset confirmation quantity m is determined based on the required confirmation reliability and the number of available historical windows under the same work status.
[0048] To illustrate the parameter settings and the changes in window fault deviation, early warning confirmation, and bearing condition level during the monitoring process in this embodiment, this embodiment uses publicly available LDM / UPB rolling bearing life-cycle monitoring data. This data is obtained from the 61806-2RS rolling bearing operating under variable speed and load conditions until failure, including bearing housing vibration monitoring data and speed, static load, and dynamic load data corresponding to each vibration sampling. The sampling time for each vibration measurement is 1.6 seconds, and the measurement interval is typically about 12 seconds. In this embodiment, the data obtained from each vibration measurement is used as a condition monitoring window, and the condition monitoring windows are arranged according to the measurement time.
[0049] In step S1, the vibration monitoring data for each state monitoring window undergoes a Fourier transform, and 200 spectral amplitudes are retained according to the preprocessing method of publicly available data. These spectral amplitudes are then divided into 20 frequency bands, and the maximum amplitude within each band is selected as a candidate index. The operating state is then determined based on the average rotational speed, average static load, and peak dynamic load corresponding to each state monitoring window.
[0050] In steps S2 to S4, eight historical bearing monitoring sequences are selected to establish reference information. The first 60% of the status monitoring windows in each historical bearing monitoring sequence are used as normal operation reference windows, and the last 10% are used as degradation reference windows. Eighty status monitoring windows are then evenly selected from each of these two parts. Windows belonging to the same operating state within the same historical bearing monitoring sequence and originating from the normal operation reference window and degradation reference window, respectively, are grouped into fault-sensitive segment pairs to establish a fault-sensitive segment pair matrix. The normal operation reference window and degradation reference window are predetermined based on their positions in the historical full-lifetime monitoring sequence, and are not determined based on the segment differences of the current candidate indicators to be evaluated.
[0051] The segment difference threshold for candidate indicators is taken as the 99th percentile of the segment difference between the normal operating reference segment pairs under the same operating conditions. The operating condition difference between the normal operating reference windows is calculated based on the average rotational speed, average static load, and peak dynamic load, and the 90th percentile of this operating condition difference is used as the operating disturbance judgment threshold. Fault sensitivity matching, operating disturbance matching, and their overlap differences are confirmed according to the method described in step S3 to obtain the effective sensitive segment matrix for each candidate indicator.
[0052] The preset coverage ratio for the coverage search is set to 1, and the maximum number of selected indicators is set to 3. The coverage search is performed according to the method described in step S4, resulting in a bearing fault early warning indicator group composed of amplitude indicators from the 12th, 3rd, and 5th frequency bands. According to the weight calculation method described in step S4, the normalized weights of the amplitude indicators from the 12th, 3rd, and 5th frequency bands are 0.9945, 0.0035, and 0.0020, respectively.
[0053] In steps S5 and S6, the B13 bearing's full-life monitoring sequence, containing 19417 condition monitoring windows, is selected for online sequential processing. Window fault deviation threshold. The threshold for warning confirmation is determined based on the statistical distribution of window fault deviations in historical normal operation reference windows; in this embodiment, it is set to 6.8517. Set the value to 0.5. The length of the same operation status confirmation window range is set to 12 status monitoring windows, and the preset confirmation quantity m is set to 3.
[0054] For the current status monitoring window, first determine its operating status based on the corresponding average speed, average static load, and peak dynamic load. Then, call the normal operating reference value under the same operating status to calculate the window fault deviation. Following the confirmation method described in step S3, the effective sensitive confirmation results of each selected indicator in the bearing fault early warning indicator group within the current status monitoring window are re-determined, and the early warning confirmation quantity is calculated. .
[0055] when If the window fault deviation threshold is not exceeded, the bearing condition is determined to be at the normal level. Exceeding the window fault deviation threshold, but If the warning confirmation threshold is not reached, or if the number of confirmations for the same operating condition does not reach the preset confirmation number, the bearing condition will be determined as a warning level. Exceeding the window fault deviation threshold When the warning confirmation threshold is reached, and the number of windows meeting the above conditions in the most recent 12 working status monitoring windows, including the current status monitoring window, reaches 3, the bearing status will be determined as a warning level.
[0056] The processing result is as follows Figure 2 As shown. Figure 2 (a) indicates the window fault deviation. Changes, Figure 2 (b) indicates the number of confirmed warnings. Changes, Figure 2(c) Indicates the change in bearing condition level. The window fault deviation of the 17681st condition monitoring window is 6.3747, which does not exceed the window fault deviation threshold; its warning confirmation quantity is 1, and the bearing condition remains at the normal level. The window fault deviation of the 17686th condition monitoring window is 6.9725, and the warning confirmation quantity is 0.9980, but the number of confirmations for the same operation condition has not yet reached the preset confirmation quantity, so the bearing condition is at the alert level. The window fault deviation of the 17700th condition monitoring window is 11.7199, and the warning confirmation quantity is 1. The number of confirmations for the same operation condition has reached 3, so a warning level is output.
[0057] Depend on Figure 2 As can be seen, under the parameter settings of this embodiment, a warning level will not be directly output if only the number of warning confirmations reaches the corresponding threshold, or if only the window fault deviation of a single status monitoring window exceeds the corresponding threshold. A warning level will be output only after both the window fault deviation and the number of warning confirmations reach the corresponding threshold, and the number of confirmations for the same operation status reaches the preset number.
[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A bearing fault early warning data processing method based on operational state constraints, characterized in that, Includes the following steps: S1: Acquire the monitoring time series data of the bearing to be monitored and the equipment operating status data, divide the monitoring time series data into status monitoring windows, determine the operating status of each status monitoring window according to the equipment operating status data, determine the segment for comparison according to the operating status, and extract candidate indicators from each status monitoring window to form a candidate indicator set; S2: Construct the candidate index segment difference matrix, the fault-sensitive segment pair matrix, and the operation disturbance segment pair matrix; Specifically, the following steps are included: S2.1: For any candidate index Calculate fragments With fragments Segment difference statistics The expression is as follows: in, and Each represents a segment and fragments , The statistical method for representing the difference between two segment index sequences is based on the sum of the values of each segment pair. Composition of candidate indicator fragment difference matrix The expression is as follows: Candidate indicators were determined based on the statistical differences in normal reference segments under the same operating conditions. Difference threshold ; S2.2: Establish a fault-sensitive segment pair matrix based on preset bearing fault-sensitive reference information. ; The fault-sensitive segment pair matrix is used to characterize segments that exhibit changes in bearing fault sensitivity under the same operating conditions. Let the first segment be... Bearing-like fault-sensitive reference information in fragments With fragments The difference between them is ,but: in, and Each represents a segment and fragments The corresponding job status Determine the threshold for fault-sensitive segments; S2.3: Establish a work disturbance segment pair matrix based on the equipment operating status data. ; The operation disturbance segment pair matrix is used to characterize the segment pair differences caused by differences in equipment operating conditions or operation process disturbances. Let the first segment pair be... The equipment operating status data in the fragment With fragments The difference in response between them is ,but: in, Determine the threshold for the operation disturbance segment; S3: Based on the segment difference statistics corresponding to each segment pair in the candidate index segment difference matrix, combined with the fault-sensitive segment pair matrix and the operation disturbance segment pair matrix, determine the fault-sensitive matching and operation disturbance matching in the candidate index differences, and construct the effective sensitive segment matrix of the candidate index. S4: Based on the effective sensitive segment matrix of each candidate index, perform a constrained coverage search on the fault sensitive segment pair matrix to obtain the bearing fault early warning index group; S5: During online monitoring, for the current status monitoring window, calculate the window fault deviation and warning confirmation amount based on the deviation of each indicator in the bearing fault warning indicator group from the normal operating reference value under the same operating condition. S6: Based on the window fault deviation, early warning confirmation amount, and the operation status to which the status monitoring window belongs, confirm the operation status and output the bearing status level.
2. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, In step S1, the equipment operating status data includes state variables characterizing the equipment operating conditions. These state variables include speed, torque, radial force, load, current level, displacement, control status, or braking status. When the work cycle identifier can be obtained, the equipment operating status data also includes the work cycle identifier. The equipment operating status data is used to determine the work status and to construct a work disturbance segment pair matrix. State variables that directly characterize the equipment operating conditions are not used as candidate indicators in the coverage search.
3. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, S3 specifically includes the following steps: S3.1: Constructing candidate indices Fault-sensitive matching matrix The expression is as follows: in, Indicate candidate indicators In fragment pairs The difference in the fragments reaches a threshold, and the fragment pair belongs to a fault-sensitive fragment pair. This is a statistic of segment differences. For fault-sensitive segment pairs matrix, , The index number of the segment; S3.2: Constructing candidate indices The operation perturbation matching matrix The expression is as follows: in, Indicate candidate indicators In fragment pairs The difference in the data reaches the threshold, and the segment pair belongs to the operational disturbance segment pair. For the perturbation fragment pair matrix, The difference threshold; S3.3: For and For fragment pairs, calculate the fragment pair confirmation count. The expression is as follows: in, Indicates the type of bearing fault-sensitive reference information. This is the amount of information used to confirm the location of the characteristic frequency band. This is the confirmed amount for the periodic impact interval. This refers to the number of confirmations that occur repeatedly under the same work status or in different work cycles. The relative enhancement confirmation value for the reference measuring point. , , and Non-negative weights It is a positive number; according to Constructing the overlap difference confirmation matrix The expression is as follows: in, To confirm the threshold; right and When the segments are compared and the differences are confirmed, And the number of characteristic frequency band locations confirmed Or periodic impact interval confirmation quantity If one of the conditions is met, the confirmation quantity occurs repeatedly in the same work status or different work cycles. Or the relative enhancement confirmation value of the reference measurement point When one of them satisfies the corresponding condition, it makes And retain the candidate index differences of the segment pair as effective sensitive differences. If the confirmation conditions are not met, make ; The confirmation conditions include: at least one of the characteristic frequency band location confirmation quantity or the periodic impact interval confirmation quantity satisfies the corresponding condition, and at least one of the repeated occurrence confirmation quantity or the relative enhancement confirmation quantity of the reference measurement point satisfies the corresponding condition. S3.4: Construct candidate indices according to the following formula Effective sensitive segment matrix The expression is as follows: in, This represents element-wise multiplication. The operation perturbation matching matrix, This indicates an element-wise logical OR operation. This indicates that the segment will be... Differences in candidate indicators are retained as valid sensitive differences.
4. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, S4 specifically includes the following steps: S4.1: Will Fragment pairs as the set to be covered and will Initialize to an empty set; S4.2: For candidate indicators that were not selected Calculate its set to be covered New coverage And calculate the proportion of unconfirmed overlap of operational disturbances and the proportion of confirmed overlap differences. and effective discrimination ; Calculate its set to be covered New coverage The expression is as follows: in, , The index number of the segment. For effective sensitive segment matrix; The formula for calculating the proportion of unconfirmed overlap of work disturbances is as follows: The formula for calculating the overlap difference confirmation ratio is as follows: The formula for calculating effective discrimination is as follows: in, Indicate candidate indicators The proportion of fault-sensitive matching that overlaps with operational disturbances and fails to pass confirmation. Indicate candidate indicators The proportion of overlapping operational disturbances that has been confirmed. Indicate candidate indicators Compared to the ability to effectively distinguish normal fluctuations in the same working state. Representing fragments With fragments This is a normal operating reference fragment pair. It is a positive number; when the candidate index There is no satisfying When the fragment pairs are equal, the effective discrimination of the candidate indicator is improved. Furthermore, this candidate metric is not included in subsequent coverage searches; S4.3: When satisfying , and Among the candidate indicators, select The largest candidate indicator was added to the bearing failure early warning indicator group. and update the set to be covered. ,in, This indicates the threshold for the proportion of overlapping operational disturbances that have not been confirmed. Indicates the effective discrimination threshold; S4.4: Remove already covered fragment pairs from the set to be covered. Remove from the middle and calculate the coverage ratio. : in, Bearing Fault Early Warning Indicator Group The corresponding effective sensitive segment matrix is composed of Effective sensitive segment matrix of each candidate indicator The element-wise logical OR operation yields the following expression: Indicates fragment pairs Bearing Failure Early Warning Indicator Group At least one candidate indicator should be used as an effective sensitive difference coverage. It is a positive number; when The number of selected indicators has reached the preset upper limit. Or there is no such condition. , and When candidate indicators are found, stop the coverage search and obtain the bearing fault early warning indicator group. ,in, This indicates the coverage ratio threshold.
5. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, In S5, the calculation window fault deviation amount The expression is: in, As candidate indicators, This is a bearing fault early warning indicator group. For status monitoring window Internal candidate indicators The deviation from the reference value under the same operating conditions. Candidate indicators The weights; The amount of early warning confirmation The calculation expression is: in, This serves as a valid sensitive confirmation flag for the current window. It is a positive number.
6. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, In step S6, the specific steps of confirming the same operating status and outputting the bearing status level are as follows: Select the window that matches the current status monitoring window from the historical status monitoring window. Windows belonging to the same job status are grouped into a set of confirmation windows with the same status. When a work cycle identifier exists, the set of confirmation windows with the same status shall preferentially include windows from different work cycles; The current status monitoring window meets the requirements. and Furthermore, after including the current status monitoring window, when the number of windows satisfying the above conditions in the current status monitoring window and the set of windows with the same status confirmation reaches a preset confirmation number m, an early warning level is output, wherein, For window fault deviation, For the confirmation of early warning, As the warning threshold, The threshold for early warning confirmation; The current status monitoring window meets the requirements. If the conditions for issuing a warning level are not met, an attention level will be issued; the current status monitoring window meets the requirements. When the time is right, output the normal level.
7. The bearing fault early warning data processing method based on operational state constraints according to claim 1, characterized in that, The difference threshold Warning threshold and warning confirmation threshold The effective discrimination threshold is determined based on normal operating reference data under the same working conditions. The preset threshold; the effective discrimination. The candidate index is determined based on the standardized difference between the segment difference statistics of the candidate index on the effective sensitive segment pairs and the segment difference statistics of the candidate index on the reference segment pairs operating normally in the same state.
8. The bearing fault early warning data processing method based on operational state constraints according to claim 3, characterized in that, When the reference measuring point is not set, the reference measuring point is used to relatively increase the confirmation value. and set the corresponding weights The overlap difference confirmation matrix is determined based on the characteristic frequency band location confirmation amount, the periodic impact interval confirmation amount, and the recurrence confirmation amount. When setting reference measurement points, the relative enhancement confirmation amount of the reference measurement point is determined based on the degree of response enhancement of the bearing measurement point relative to the reference measurement point. .
9. A bearing fault early warning data processing method based on operational state constraints according to claim 6, characterized in that, The preset confirmation quantity m is the number of confirmations required to output the warning level; when there is a work cycle identifier, the count is made according to the number of work cycles with the same work status and meeting the warning conditions in the current status monitoring window and different historical work cycles; when there is no work cycle identifier, the count is made according to the number of status monitoring windows that meet the warning conditions in the current status monitoring window and the set of confirmation windows with the same status.
