Method for early defect detection of rolling bearings

Abstract

This application provides a method for early defect detection of rolling bearings, comprising: dividing the circumference of the rolling bearing into multiple angular windows and selecting a test angular window; fixing the outer ring of the bearing and driving the inner ring of the bearing to rotate; taking two adjacent rings as a set of detection cycles, applying a base load in the first cycle and acquiring a first response signal within the test angular window, and superimposing a pulse load on the base load in the second cycle and acquiring a second response signal within the test angular window; the amplitude of the pulse load is 3% to 25% of the base load; performing a differential operation on the second response signal and the first response signal to obtain a response increment; and determining whether an early defect exists at the circumferential position corresponding to the test angular window based on the response increment. The early defect detection method for rolling bearings of this application can reliably identify early defects and their corresponding locations.

Description

Technical Field

[0001] This application relates to the field of bearing fault detection technology, and in particular to a method for detecting early defects in rolling bearings. Background Technology

[0002] Existing methods for detecting rolling bearing faults primarily rely on the response signals generated during bearing operation for condition identification. A common approach involves collecting signals such as vibration, acceleration, sound, current, temperature, or acoustic emission during bearing rotation, and then combining these signals with time-domain analysis, frequency-domain analysis, envelope analysis, order analysis, or correlation statistical analysis to extract fault features and determine whether the bearing exhibits any abnormalities.

[0003] The basic principle of this type of method is that when defects exist in the inner ring, outer ring, rolling elements, or cage of a bearing, the rolling elements passing over the defect location will cause a change in the contact state and generate a corresponding impact or fluctuation response. Under certain conditions, this response will exhibit periodic characteristics and show characteristic frequencies or abnormal changes related to bearing failure in the vibration spectrum, envelope spectrum, or other characteristic parameters. Based on this principle, when the defect has developed to the point where it can stably generate a significant impact response, existing methods can usually achieve good detection results. For example, when the defect manifests as obvious spalling, deep indentations, large crack propagation areas, or significant rolling element surface damage, the fault characteristics can often be easily extracted from the acquired signals.

[0004] However, existing methods are often ineffective in detecting early, minute defects, leading to frequent missed detections. This is because early defects are typically small in size and haven't yet formed significant geometric abrupt changes or persistent impact boundaries, resulting in weak response changes when the rolling element passes over the defect area. Under conditions of low load, low speed, good lubrication, or high structural damping, these weak responses are further weakened by the lubricating film, bearing housing, mounting structure, and transmission path. In such cases, the acquired signals are often dominated by normal rolling contact signals, motor running disturbances, clamping vibrations, and environmental background noise, with the abnormal components corresponding to defects accounting for a very low proportion. This results in unremarkable characteristic frequencies and insignificant changes in characteristic parameters, making reliable identification of early defects difficult.

[0005] In addition, another type of existing method involves synchronous sampling, angle alignment, multi-turn averaging, order tracking, or correlation analysis of bearing multi-turn operation signals to improve the stability of fault characteristics. The basic idea of ​​this type of method is to utilize the periodicity of the fault response during the repeated rotation of the bearing to align and statistically process the multi-turn signals, thereby reducing random noise and improving the identifiability of repeatable features. This type of method is well-suited for fault signals that can reliably repeat over multiple turns of operation.

[0006] However, the methods described above still rely on the fact that defects can continuously generate repeatable anomalies under natural operating conditions. If the response caused by the early defects themselves is weak, multi-loop averaging or statistical processing, while reducing random noise, may also weaken local transient anomalies, smoothing out or masking weak defect features. Especially when the defect is in its early stages, the response is unstable, and only slight local disturbances are observed when passing through the defect location, simply relying on passive response acquisition and statistical analysis during natural operation often makes it difficult to stably distinguish the defect location from the overall background response.

[0007] Therefore, existing methods for detecting early defects in rolling bearings generally suffer from the following shortcomings: First, the detection relies primarily on the bearing generating abnormal responses under natural operating conditions, while early, minute defects typically fail to generate sufficiently strong and stable abnormal signals. Second, even with multi-turn synchronization, averaging, or statistical analysis, the main improvement is limited to enhancing the visibility of existing repetitive signals, offering limited improvement for the already weak and unstable responses to early defects. Especially in scenarios with small defect sizes, intact lubricating films, low loads, stable rotational speeds, high background noise, or complex transmission paths, existing methods often struggle to reliably identify early defects and their corresponding locations. Summary of the Invention

[0008] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a method for early defect detection of rolling bearings, which can reliably identify early defects and their corresponding locations.

[0009] In a first aspect, this application provides a method for early defect detection of a rolling bearing, the rolling bearing comprising an inner ring and an outer ring, the method comprising:

[0010] The circumference of the rolling bearing is divided into multiple angle windows, and one of the angle windows is selected as the angle window to be measured.

[0011] Fix the outer ring of the bearing and drive the inner ring of the bearing to rotate;

[0012] Two adjacent cycles are considered as a detection cycle. In the first cycle, a base load is applied and a first response signal is acquired within the measured angle window. In the second cycle, a pulse load is superimposed on the base load within the measured angle window and a second response signal is acquired. The amplitude of the pulse load is 3% to 25% of the base load.

[0013] Perform a differential operation on the second response signal and the first response signal to obtain the response increment;

[0014] The response increment is used to determine whether there are early defects at the circumferential position corresponding to the angle window to be tested.

[0015] In one implementation of the first aspect, the duration of the pulse load is no greater than 80% of the rotation time corresponding to the angle window to be measured.

[0016] In one implementation of the first aspect, the angle window to be measured is a single angle window, or multiple angle windows selected in a preset order.

[0017] In one implementation of the first aspect, the base load and the pulse load are applied to the outer ring of the bearing and are applied radially toward the center of the bearing.

[0018] In one implementation of the first aspect, prior to the step of performing the detection cycle, the early defect detection method for rolling bearings further includes the following steps:

[0019] The controller sends a pulse loading command to the pulse loading mechanism and records the time of command transmission.

[0020] The pulse loading mechanism applies the pulse load to the outer ring of the bearing according to the pulse loading command;

[0021] The actual loading moment when the pulse load is applied to the outer ring of the bearing is obtained by a force sensor fixed to the outer ring of the bearing;

[0022] The difference between the loading time and the instruction sending time is calculated to obtain the execution delay time of the pulse loading mechanism;

[0023] The angular position and angular velocity of the inner ring of the bearing are obtained by an encoder, which is mounted on a spindle connected to the inner ring of the bearing.

[0024] Calculate the angular advance amount based on the execution delay time and the angular velocity;

[0025] During the detection cycle, the early defect detection method for the rolling bearing further includes the following steps:

[0026] Determine the starting angle position corresponding to the angle window to be measured;

[0027] The advance angle position for sending the pulse loading command is determined based on the starting angle position and the angle advance amount.

[0028] When the angular position measured by the encoder reaches the advance angular position, the controller sends the pulse loading command to the pulse loading mechanism so that the pulse load falls within the measured angle window.

[0029] In one implementation of the first aspect, the controller applies the base load to the outer ring of the bearing via a cylinder, and the pulse loading mechanism is a voice coil motor connected between the cylinder and the outer ring of the bearing.

[0030] In one implementation of the first aspect, the push rod end of the voice coil motor is connected to the outer ring of the bearing in sequence through a steel upper pressure plate, a polyurethane gasket, and a steel arc-shaped pressure head. The steel arc-shaped pressure head is provided with an arc-shaped concave surface that mates with the outer circumferential surface of the outer ring of the bearing.

[0031] In one implementation of the first aspect, the step of performing a differential operation on the second response signal and the first response signal to obtain the response increment includes:

[0032] From the first response signal, feature parameters used to characterize the signal strength are extracted to obtain the first feature value;

[0033] The second feature value is obtained by extracting the feature parameters used to characterize the signal strength from the second response signal.

[0034] The difference between the second feature value and the first feature value is obtained;

[0035] The ratio of the feature difference to the first feature value is used as the response increment.

[0036] In one implementation of the first aspect, the characteristic parameters are the peak value, root mean square value, signal energy, bandwidth energy, kurtosis, or pulse count of the signal.

[0037] In one implementation of the first aspect, determining whether an early defect exists at the circumferential position corresponding to the measured angle window based on the response increment includes:

[0038] Multiple test cycles are performed on normal bearings to obtain the response increments of multiple normal bearings, and the average value of the response increments of multiple normal bearings is taken to obtain the reference response increment.

[0039] The response increments of the same angle window under test in multiple consecutive detection cycles are statistically analyzed to obtain multiple measured response increments. The average measured response increment is obtained by averaging the multiple measured response increments.

[0040] The ratio of the average measured response increment to the reference response increment is calculated to obtain the response anomaly intensity;

[0041] The detection period in which the measured response increment is greater than the reference response increment is defined as the abnormal period, and the proportion of the abnormal period in multiple detection periods is calculated to obtain the response anomaly density.

[0042] Obtain the peak time of the second response signal within the abnormal period, and determine the abnormal angle position based on the angular position of the bearing inner ring corresponding to the peak time;

[0043] Calculate the standard deviation of multiple abnormal angle positions to obtain the abnormal angle dispersion;

[0044] The angle instability is obtained by calculating the ratio of the abnormal angle dispersion to the width of the angle window to be measured.

[0045] Determine the angle stability based on the aforementioned angle instability;

[0046] The defect confidence level of the test angle window is obtained by weighted summation of the response anomaly intensity, the response anomaly density, and the angle stability.

[0047] When the confidence level of the defect is greater than or equal to the preset confidence threshold, it is determined that there is an early defect at the circumferential position corresponding to the angle window to be tested.

[0048] The technical solution of this application has the following beneficial effects:

[0049] This application does not continuously apply additional excitation to the entire bearing circumference. Instead, it uses encoder angle information to briefly trigger pulse loads within a preset angle window, giving the pulse action a clear angular correspondence. By establishing a correspondence between the load application time and the bearing circumferential position, this application enables the detection process to unfold around a specific test angle window, thereby improving the targeting of local defect areas and facilitating defect location identification, rather than simply obtaining an overall anomaly judgment result.

[0050] The criterion for judgment in this application is not the absolute response magnitude after pulse loading, but rather the response difference between the base load cycle and the pulse load cycle within the same angular window. This before-and-after comparison method allows the test results to better reflect the sensitivity of the measured location to short-term load disturbances, rather than being solely influenced by overall vibration levels, background noise levels, or sensor installation conditions. For normal areas, the response change before and after the pulse is usually small; for areas with early defects, the pulse is more likely to induce more obvious local contact anomalies, resulting in a more prominent response increment. Therefore, this application helps to transform early, weak defects that are not easily apparent under natural operating conditions into more easily identifiable differential anomalies, improving the ability to detect early defects.

[0051] This application uses the same angular window in two adjacent revolutions for comparison, so that the basic load response and the pulse load response are under operating conditions as close as possible. Because the interval between the two tests is short, the changes in boundary conditions such as speed, load, temperature rise, lubrication state, and clamping state are small. Therefore, the influence of operating condition drift on the comparison results can be reduced, and the obtained differential results can more accurately reflect the changes in local contact state, thereby improving the stability and comparability of the test results.

[0052] This application does not make defect judgments based on a single anomaly, but rather combines a multi-group adjacent circle re-confirmation mechanism for comprehensive judgment. The defect location is only output when the same measured angle window or corresponding position shows stable anomalies in multiple consecutive sets of detections. This method effectively reduces misjudgments caused by random noise, accidental impacts, and fluctuations in single sampling, improving the reliability of early defect identification results.

[0053] This application also incorporates a pulse interference rejection mechanism to distinguish and eliminate non-defect anomalies caused by the pulse loading mechanism itself, the structural transmission path, or transient interference, thereby avoiding misidentification of loading disturbances as bearing defects. Therefore, this application not only enhances the response increment in the early defect region but also suppresses the influence of non-defect background responses on the detection results, improving the accuracy of defect identification. Attached Figure Description

[0054] Figure 1 The flowchart shown is a method for early defect detection of rolling bearings according to an embodiment of this application.

[0055] Figure 2 Displayed as Figure 1 The detailed flowchart of step S400.

[0056] Figure 3 The diagram shows a comparison of the response of the defective bearing under basic load and pulse load.

[0057] Figure 4 The graph shows a comparison of the response of a normal bearing under basic load and pulse load.

[0058] Figure 5 The graph shows a comparison of the normalized differential responses of normal and defective bearings. Detailed Implementation

[0059] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0060] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the illustrations only show the components related to this application and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the shape, quantity and proportion of each component can be adaptively adjusted, and the layout of the components may also be more complex.

[0061] Please refer to Figures 1 to 5 This application provides a method for early defect detection of rolling bearings. This method is based on comparative detection of the same angular window between two adjacent bearings. A basic response is collected in the previous bearing, and a short-duration pulse load is applied to the same angular window in the subsequent bearing, and the disturbance response is collected. The difference in response between the two adjacent bearings at the same angular window is then used to determine whether there is a localized area abnormally sensitive to short-duration load disturbances, thereby achieving early detection of defects in the rolling bearing. This method is applicable to detection scenarios where the outer ring is fixed and the inner ring rotates. It can enhance the localized abnormal response through controlled disturbances when early indentations, shallow spalling, microcracks, localized scratches, or material discontinuities have not yet formed obvious impact characteristics under natural operating conditions. Furthermore, the comparison between adjacent bearings weakens the effects of background noise, fixture-transmitted disturbances, and changes in operating boundary conditions.

[0062] Please refer to Figure 1 In one embodiment, this application provides a method for early defect detection of a rolling bearing, the rolling bearing including an inner ring and an outer ring, the method comprising:

[0063] Step S100: Divide the circumference of the rolling bearing into multiple angle windows, and select one of the angle windows as the angle window to be measured.

[0064] In step S100, the angle window refers to the angle interval obtained by dividing the bearing circumference according to the spindle rotation angle. Each angle window corresponds to a detection segment on the bearing circumference. This angle window can be established by encoder angle pulses or formed by the acquisition and calculation unit after assigning angle domains to the sampled data based on the spindle angle. Let the width of each angle window be... If the unit is degrees, then the total number of windows is... It can be represented as:

[0065]

[0066] in,

[0067] This represents the total number of angle windows obtained by dividing the bearing circumference;

[0068] This indicates the angular width of a single angle window.

[0069] In one embodiment, Desirable to For example, when Pick At that time, a week can be divided into 36 angle windows. A window at a specific angle can be defined as:

[0070]

[0071] in,

[0072] Indicates the first The angle range corresponding to each angle window;

[0073] Indicates the angle window number, and .

[0074] Using an angle window instead of a fixed time window for detection ensures that the same detection segment in the basic load circle and the pulse load circle remains angularly aligned even when there are slight fluctuations in the spindle speed. This allows subsequent differential results to more effectively reflect the response changes after local disturbances.

[0075] In one embodiment, the corresponding value per spindle revolution can be... The angle range is divided into 36 equal angle windows, and the angle width of each angle window is... Among them, the first angle window can correspond to to The second angle window can correspond to to The 18th angle window can correspond to to The 19th angle window can correspond to to And so on, until the 36th angle window corresponds to to .

[0076] If the 18th angle window is selected as the angle to be measured, it means that under the principal axis angle reference system, for to This week, the inspection is carried out in the interval. Within the basic load cycle, the acquisition spindle rotates to... to The first response signal within the interval; within the pulse load loop, when the spindle rotates to the trigger angle position corresponding to the measured angle window, a pulse load is applied, and data is collected. to The second response signal within the interval; then the response of the measured angle window in two adjacent revolutions is compared to obtain the corresponding response increment.

[0077] In this application, the angle window is preferably established based on the angle of the spindle encoder that rotates synchronously with the inner ring of the bearing. The circumferential position corresponding to the angle window to be measured refers to the circumferential detection position under the encoder angle reference system. For inner ring defects, the circumferential detection position can correspond one-to-one with the actual defect position of the inner ring raceway; for outer ring defects and rolling element defects, the component to which they belong can be further identified by combining the spatial stability or migration law of the anomaly in consecutive multiple revolutions.

[0078] In one embodiment, the angle window to be tested is a single angle window, or multiple angle windows selected in a preset order. When the angle window to be tested is a single angle window, the single window can be repeatedly tested to improve the accuracy of confirming suspected abnormal positions; when the angle window to be tested is multiple angle windows selected in a preset order, adjacent two-circle comparison tests can be performed on multiple windows in sequence to achieve window-by-window scanning and rapid screening of the entire circumferential position.

[0079] Step S200: Fix the outer ring of the bearing and drive the inner ring of the bearing to rotate.

[0080] In step S200, fixing the outer ring of the bearing and driving the inner ring to rotate means restricting the spatial position of the outer ring of the bearing through a positioning fixture, and driving the inner ring to rotate through a spindle connected to the inner ring. Preferably, a low-speed pre-rotation is performed before formal testing to confirm the absence of jamming, rubbing, or clamping abnormalities. The bearing positioning fixture can also be equipped with a positioning detection element and a clamping force detection element. If the bearing is not installed in place, or the clamping force is not within the set range, the system can prevent entry into the formal testing process. By maintaining the stable position of the outer ring and allowing the inner ring to rotate with the spindle, a correspondence between the spindle angular position and the bearing rotation position can be established under the encoder angular reference. This facilitates the subsequent unification of pulse triggering, response acquisition, and abnormal position judgment into the same angular coordinate system.

[0081] Step S300: Two adjacent cycles are taken as a set of detection cycles. In the first cycle, a base load is applied and a first response signal is acquired within the angle window to be measured. In the second cycle, a pulse load is superimposed on the base load within the angle window to be measured and a second response signal is acquired. The amplitude of the pulse load is 3% to 25% of the base load.

[0082] In step S300, two adjacent cycles are considered as a detection cycle, which means that the first cycle is... The ring, as the basic load ring, will be the adjacent ring. The first two loops together form the pulse load loop. Group detection cycle. The group detection cycle can be expressed as:

[0083]

[0084] in,

[0085] Indicates the first Group testing cycle;

[0086] Indicates the first The basic load circle in the group;

[0087] Indicates the first The pulse load loop in the group;

[0088] This indicates the testing group number.

[0089] Using two adjacent cycles for comparison, rather than comparing across multiple cycles, ensures that the base load cycle and the pulse load cycle are in as consistent a state as possible in terms of temperature rise, lubrication, spindle speed, clamping, and sensor contact. This reduces errors introduced by slow changes in operating conditions. In other words, the response difference between two adjacent cycles is more likely to originate from the pulse load itself, rather than from time drift or environmental changes.

[0090] In step S300, the base load refers to the steady-state load continuously applied to the bearing during the testing process. Its function is to maintain stable contact between the rolling elements and the raceway, rather than subjecting the bearing to fatigue loading or destructive testing. The pulse load refers to a short-term disturbance load superimposed on the base load. Its function is to subject the local contact area corresponding to the target angle window to a controlled short-term load disturbance, thereby enhancing the response increment caused by sudden changes in contact stiffness, micro-slippage, micro-collision, or stress concentration at the edge of the local defect. The first response signal refers to the response signal collected within the measured angle window in the base load loop; the second response signal refers to the response signal collected within the same measured angle window in the pulse load loop. The response signal can be a single signal or response information composed of one or more of vibration signals and acoustic emission signals.

[0091] In step S300, let the foundation load be... The pulse load is Then the two satisfy:

[0092]

[0093] in,

[0094] Indicates the base load;

[0095] This indicates a pulse load superimposed on the base load.

[0096] In one embodiment, when the base load When the pulse load is 100 N, The amplitude can range from 3 N to 25 N, for example, 10 N, 15 N, or 20 N. Limiting the pulse load amplitude to 3% to 25% of the base load ensures that the defective area produces a measurable response increment under short-term load disturbances, while avoiding excessive pulse loads that could lead to a significant increase in the impact on the loading mechanism itself, or an excessive increase in the overall response of the normal area, fixture structure, and transmission system. By limiting the ratio of the pulse load to the base load, a balance can be achieved between enhancing the local abnormal response and controlling background disturbances.

[0097] Step S400: Perform a differential operation on the second response signal and the first response signal to obtain the response increment.

[0098] In this application, the response increment is used as a general term to represent the change in response of the same measured angle window in the pulse load loop relative to the response in the base load loop; when the change is represented by the difference between the second eigenvalue and the first eigenvalue, it is called differential response; when the change is further normalized using the response of the base load loop, it is called normalized differential response.

[0099] In this application, the local defect area and the normal contact area of ​​the rolling bearing exhibit different response increment characteristics under short-term additional load, which is beneficial for detecting early defects.

[0100] For the normal region where the raceway surface is continuous and the rolling element contact state is stable, under the base load, the rolling element mainly produces a smooth rolling contact response when passing through this region. Even if a pulse load is superimposed within the same measured angle window in the later revolution of an adjacent detection cycle, the response in this normal region usually only shows a limited increment with increasing load. The change is mostly reflected in a gradual increase in amplitude, without forming a significant sudden increase in transient impact components or abnormal high-frequency acoustic components. In other words, the normal region has relatively low sensitivity to short-time pulse loads, and the differential response between two consecutive revolutions usually remains at a low level.

[0101] Conversely, when early indentations, shallow spalling, microcracks, localized scratches, or material discontinuities exist on the inner ring raceway, outer ring raceway, or rolling element surface of a bearing, the geometric continuity, contact stiffness, and load transfer state at these local locations will change. When the rolling element passes through such areas, abrupt changes in contact stiffness, localized stress concentration, microslippage, micro-impact, or short-term elastic recovery may occur at the defect edge, indentation transition zone, or spalling boundary. Under natural operating conditions, due to the small defect size, low base load, relatively intact lubrication film, or high structural damping, these anomalies often only correspond to low-amplitude local transient responses and are easily masked by normal rolling noise, motor disturbances, fixture vibrations, and background noise. Therefore, it is difficult to directly identify stability through single-turn absolute response.

[0102] Based on this, this application does not rely on defects to spontaneously generate sufficiently obvious anomalies during natural operation. Instead, it establishes a before-and-after comparison within the same measured angle window in two adjacent cycles. In the first cycle, a first response signal is acquired under the base load for that angle window, characterizing the instantaneous base response at that position under the current operating state. In the second cycle, a short-duration pulse load is superimposed on the base load within the same angle window, and a second response signal is acquired, characterizing the response change at that position under controlled short-duration disturbances. Since the two samplings correspond to the same measured angle window and occur in two adjacent cycles, the spindle speed, base load, temperature rise, lubrication status, clamping status, and sensor contact status typically do not change significantly. Therefore, the difference between the two responses mainly reflects the sensitivity of that local position to the short-duration pulse load.

[0103] Furthermore, if the measured angle window corresponds to a normal contact area, the increment of the second response signal relative to the first response signal is usually small. If the measured angle window corresponds to an area with local early defects, the short-time pulse load will further amplify the sudden changes in contact force, local collisions, and high-frequency transient responses at the defect edge, making the response increment of this window in the pulse load cycle significantly greater than the response in the basic load cycle. Therefore, by calculating the differential response of the same measured angle window in two adjacent cycles, early weak defects that are not easily apparent under natural operating conditions can be transformed into more easily identifiable response increment anomalies.

[0104] In one embodiment, the first in the basic load circle A window at an angle The first feature value extracted within the group detection period can be denoted as: The second eigenvalue corresponding to the pulse load loop is denoted as , of which The angle window at the first The differential response in the group detection cycle can be expressed as:

[0105]

[0106] in,

[0107] Indicates the first The first in the group testing cycle Differential response of each angle window;

[0108] Indicates the first The first in the group testing cycle The first characteristic value corresponding to the first response signal of each angle window in the basic load circle;

[0109] Indicates the first The first in the group testing cycle The second characteristic value corresponding to the second response signal of each angle window in the pulse load loop.

[0110] When there are significant differences in the fundamental response amplitudes, or when it is desirable to further reduce the impact of absolute amplitude differences between different bearings, a normalized differential response form can also be used:

[0111]

[0112] in,

[0113] Indicates the first The first in the group testing cycle Normalized differential response of each angle window;

[0114] Indicates the first The first in the group testing cycle The second characteristic value of a angular window in the pulse load loop;

[0115] Indicates the first The first in the group testing cycle The first eigenvalue of a angular window in the basic load circle;

[0116] Represents the stability constant, used to avoid The denominator is abnormal when the value is too small.

[0117] After adopting the above differential method, the judgment criterion is no longer the absolute response magnitude in a certain revolution after pulse loading, but the response change of the same measured angle window between the basic load revolution and the pulse load revolution. This can reduce the influence of different bearings, different clamping conditions, different background noise levels, and different sensor installation conditions on the absolute response value, so that the detection results more centrally reflect whether the measured angle window has abnormal sensitivity to short-term load disturbances.

[0118] Furthermore, this application uses two adjacent cycles as a detection period, which further suppresses the influence of boundary condition drift. If there are too many cycles between the basic response and the pulse response, factors such as speed fluctuations, temperature rise changes, changes in lubrication film condition, and changes in the force transmission path of the fixture may all enter the differential results, thereby weakening the correlation between the differential value and local defects. Using two adjacent cycles for detection allows the two samplings to be under the most similar operating conditions possible, thus making the differential results more reflective of the influence of the pulse load itself on the local contact state.

[0119] Step S500: Determine whether there are early defects at the circumferential position corresponding to the angle window to be tested based on the response increment.

[0120] Furthermore, this application does not use a single differential anomaly as the final basis. Instead, it requires that an abnormal response increment continuously appearing in the corresponding angle window to be tested within a preset number of consecutive testing cycles before determining that an early defect exists at the corresponding circumferential position. The reason for this setting is that if an anomaly only appears sporadically at the moment of pulse triggering, but cannot stably correspond to the same angle window to be tested or its adjacent windows in subsequent testing groups, then the anomaly is more likely to originate from the pulse loading mechanism itself, the impact transmitted by the fixture, or random noise, and is not suitable as a local defect output of the bearing. Conversely, if a certain angle window to be tested shows stable differential enhancement in multiple consecutive adjacent testing cycles, and the enhancement result corresponds stably to the angle window to be tested, it indicates that the position has a continuous abnormal sensitivity to short-term load disturbances, and an early defect can be determined based on this.

[0121] In one embodiment, A positive number much smaller than the normal response amplitude can be used, such as 0.001, 0.0001, or the minimum stable compensation value determined according to the quantization resolution of the sampling system. Using the normalized difference form, the difference result reflects the enhancement ratio of the pulsed response relative to the basic response, rather than just the absolute increment, which helps improve the comparability between different batches of bearings.

[0122] In the above embodiments, by using the same angular window of the bearing itself in two adjacent rings as the before and after comparison object, the bearing does not rely on defects to generate obvious abnormalities on their own under natural operating conditions. Instead, it actively enhances the response increment of the local defect area by using controlled short-term disturbances and weakens the common background response by differential, so that early weak defects can be more stably displayed from the background of the whole ring.

[0123] In one embodiment, the duration of the pulse load is no more than 80% of the rotation time corresponding to the angle window to be measured.

[0124] Let the rise time of the pulse load be The holding time is The descent time is The total duration of the pulse load It can be represented as:

[0125]

[0126] in,

[0127] Indicates the rise time of the pulse load;

[0128] Indicates the holding time of the pulse load;

[0129] Indicates the fall time of the pulse load;

[0130] This indicates the total duration of the pulse load.

[0131] Let the spindle speed be The unit is r / min; the width of the window for the angle to be measured is... The unit is degrees; then the window transit time required for the principal axis to pass through the measured angle window is... It can be represented as:

[0132]

[0133] in,

[0134] This indicates the window transit time required for the main axis to pass through the measured angle window;

[0135] Indicates the spindle speed;

[0136] This indicates the angular width of the window to be measured.

[0137] By setting It can effectively reduce window assignment aliasing caused by pulse tailing, so that subsequent difference calculations can more accurately correspond to the window of the angle to be measured, without incorrectly including the response of adjacent windows in the current window.

[0138] In one embodiment, the base load and pulse load are applied to the outer ring of the bearing and are applied radially toward the center of the bearing.

[0139] In this embodiment, both the base load and the pulse load are input to the rolling contact pair via radial loading applied to the outer ring of the bearing. Using a radial loading method towards the bearing center allows for a more direct change in the contact load between the rolling elements and the raceways of the inner and outer rings, thus making it easier to identify contact defects in the inner ring raceway, outer ring raceway, and rolling elements. In some embodiments, the base load and pulse load can also be applied first to the bearing housing, a loading block, or a loading boss rigidly connected to the outer ring, and then the load is transferred to the outer ring by this structure. In specific engineering implementations, the application to the outer ring can include direct application to the outer ring or indirect application to the outer ring through a force-transmitting component rigidly coupled to the outer ring, as long as a radial load input to the outer ring is ultimately achieved.

[0140] Using radial loading helps to directly correlate the controlled pulse disturbance with the normal contact state of the rolling contact area, thereby more sensitively stimulating local contact stiffness abrupt changes and transient impact responses at the defect edge, and improving the ability to detect early contact anomalies of the raceway and rolling elements.

[0141] In one embodiment, the early defect detection method for rolling bearings further includes the following steps before performing the detection cycle steps:

[0142] Step S610: Send a pulse loading command to the pulse loading mechanism through the controller and record the time of command sending.

[0143] In step S620, the pulse loading mechanism applies a pulse load to the outer ring of the bearing according to the pulse loading command.

[0144] Step S630: The loading moment when the pulse load actually acts on the outer ring of the bearing is obtained by a force sensor fixed to the outer ring of the bearing.

[0145] Step S640: Calculate the difference between the loading time and the command sending time to obtain the execution delay time of the pulse loading mechanism.

[0146] Step S650: Obtain the angular position and angular velocity of the bearing inner ring through an encoder, which is mounted on the spindle connected to the bearing inner ring.

[0147] Step S660: Calculate the angular advance based on the execution delay time and angular velocity.

[0148] Step S670: Within the detection cycle, determine the starting angle position corresponding to the angle window to be measured.

[0149] Step S680: Determine the advance angle position for sending the pulse loading command based on the starting angle position and the angle advance amount.

[0150] Step S690: When the angular position measured by the encoder reaches the advance angular position, the controller sends a pulse loading command to the pulse loading mechanism so that the pulse load falls into the window of the angle to be measured.

[0151] In steps S610 to S640, the execution delay time refers to the time difference between the controller issuing the pulse loading command and the pulse load actually starting to act on the bearing. This time difference includes not only the control command transmission time, but may also include the electromagnetic response time of the actuator, the time for eliminating mechanical transmission backlash, the contact establishment time of the loading head, and the detection delay of the force sensor, etc. Let the execution delay time be... Then we have:

[0152]

[0153] in,

[0154] Indicates the execution delay time;

[0155] This indicates the actual loading moment when the pulse load is applied to the bearing;

[0156] This indicates the time when the controller sends the pulse loading command.

[0157] In one embodiment, the force sensor can detect an increase in the applied force to... The moment when the load is detected, or the moment when a significant deviation of the load from the base load platform is detected, is taken as the actual moment of application of the pulse load. By using the actual loading moment detected by the force sensor, rather than simply using the controller's command moment, the true response timing of the actuator can be reflected more accurately.

[0158] In steps S650 and S660, the angular velocity of the spindle is assumed to be... Then the angle advance It can be represented as:

[0159]

[0160] in,

[0161] Indicates the lead time of the angle;

[0162] This indicates the angular velocity of the main shaft, and the unit can be degrees per second.

[0163] Indicates the execution delay time.

[0164] When the angular velocity is expressed in r / min Sometimes, it can also be written as:

[0165]

[0166] in,

[0167] Indicates the lead time of the angle;

[0168] Indicates the spindle speed;

[0169] Indicates the execution delay time.

[0170] In step S670, the starting angle position refers to the reference angle position selected to ensure that the pulse load falls within the measured angle window. In one embodiment, the first... Each window of the angle to be measured is defined as... Then the starting angle position corresponding to this window can be represented as:

[0171]

[0172] in,

[0173] Indicates the first The starting angle position corresponding to each angle window to be measured;

[0174] Indicates the window number of the angle to be measured;

[0175] Indicates the width of the angled window.

[0176] In step S680, the advance angle position of sending the pulse loading command is... It can be represented as:

[0177]

[0178] in,

[0179] Indicates the advance angle position;

[0180] Indicates the starting angle position;

[0181] Indicates the lead time of the angle.

[0182] By pre-measuring the execution delay time and calculating the angular advance based on the spindle speed, the pulse load can be ensured to actually fall into the target window when the spindle reaches it, rather than being delayed to an adjacent window due to actuator lag. This method helps improve the accuracy of the pulse application angle and reduces the impact of angular offset on the differential judgment results.

[0183] In one embodiment, the 18th angle window is selected as the angle window to be measured, and the angle window to be measured corresponds to the principal axis angle reference system. to Range. To ensure that the pulse load actually acts within the 18th angular window, the execution delay time of the pulse actuator can be pre-calibrated. The execution delay time can be understood as the time interval between the control unit issuing the trigger command and the pulse actuator actually applying the pulse load to the bearing.

[0184] In one embodiment, the calibrated execution delay time is 5 ms. During the detection process, if the spindle speed is 600 r / min, the spindle rotates 10 revolutions per second, corresponding to an angular velocity of... per second. Based on the execution delay time and angular velocity, the angular advance can be calculated as follows: In this case, to ensure that the pulse load actually falls within the 18th angle window... to Within the range, the spindle can be rotated to A trigger command is issued at the specified time. Thus, after considering the execution delay, the pulse load can actually act on... to Interval.

[0185] Within the basic load cycle, the acquisition spindle rotates to... to The first response signal within the interval; within the pulse load loop, a trigger command is issued at the compensated trigger angle position, causing the pulse load to act on the [object] after the execution delay. to Within the interval, a second response signal is acquired within that interval. Subsequently, based on the first and second response signals, the corresponding response feature values ​​are extracted, and the response increment of the 18th angular window is calculated.

[0186] If, in multiple consecutive detection cycles, the compensated pulse load applied to the 18th angular window can consistently make the response increment within the 18th angular window greater than the preset increment threshold, then the advance compensation can be considered effective, and it can be determined that there is an early defect at the detection position corresponding to the 18th angular window.

[0187] In practical applications, if the spindle speed changes, the acquisition and calculation unit can dynamically update the angular advance based on the spindle angular velocity measured in real time by the encoder, and adjust the trigger angle position accordingly to improve the accuracy of the pulse load falling within the target angle window. If necessary, the execution delay time can also be iteratively corrected by combining historical trigger errors to further improve the compensation accuracy.

[0188] In one embodiment, if there are slight fluctuations in the spindle speed, the acquisition and calculation unit can avoid truncating windows according to a fixed time length. Instead, it can use an angle resampling method to assign the sampled data. That is, it first binds the spindle angle to each sampling point according to the encoder signal, and then assigns the sampling point to the corresponding angle window. In this way, even if there are small changes in speed, it can ensure that the same angle window in the basic load circle and the pulse load circle is aligned in an angular sense. At the same time, if there is a deviation between the controller's command time and the actual loading time detected by the force sensor, the acquisition and calculation unit can also correct the angle assignment of the response data based on the deviation, thereby avoiding the incorrect allocation of pulse responses that should belong to the measured window to adjacent windows.

[0189] By adopting the loading control and angle compensation method of this embodiment, a stable and reproducible correspondence can be established between pulse triggering and angle window. This not only facilitates high-precision detection in a single window, but also provides a timing basis for multi-window polling triggering and rapid full-circle screening.

[0190] In one embodiment, the controller applies a basic load to the outer ring of the bearing via a cylinder, and the pulse loading mechanism is a voice coil motor connected between the cylinder and the outer ring of the bearing.

[0191] In this embodiment, the base load is provided by a cylinder, which continuously outputs a relatively stable thrust to maintain stable contact load on the bearing during the testing process. The pulse load is provided by a voice coil motor, which is connected in series between the cylinder and the outer ring of the bearing. This voice coil motor is used to superimpose a rapid, short-stroke, controllable short-duration pulse disturbance onto the base load provided by the cylinder. This combination of cylinder and voice coil motor balances the stability of the base load with the rapid response of the pulse loading. The cylinder is suitable for establishing the base contact conditions, while the voice coil motor has advantages such as fast response, high control precision, and ease of generating controllable waveforms in the rising, holding, and falling segments, making it more suitable for implementing short-duration pulse loading within a specified angular window.

[0192] In one embodiment, the controller can send control signals to a proportional valve or solenoid valve to adjust the cylinder output, stabilizing the base load at a set value. Simultaneously, the controller can also output pulse current commands to the voice coil motor driver to control the voice coil motor to output a pulse load of a set amplitude and waveform at a target time. By separating the base load from the pulse load generation mechanism, the interference of pulse control on the stability of the base load can be reduced, improving detection repeatability.

[0193] In one embodiment, the push rod end of the voice coil motor is connected to the outer ring of the bearing in sequence through a steel upper pressure plate, a polyurethane gasket, and a steel arc-shaped pressure head. The steel arc-shaped pressure head has an arc-shaped concave surface that matches the outer circumferential surface of the bearing outer ring.

[0194] In this embodiment, a steel upper pressure plate is used to transmit the axial output of the push rod and provide structural rigidity support. A polyurethane gasket is used to buffer localized hard contact, reduce additional impacts introduced by high-frequency rigid collisions, and improve the uniformity of contact force between the pressure head and the outer ring. A steel arc-shaped pressure head is used to establish surface or near-surface contact with the outer circumferential surface of the bearing outer ring. The arc-shaped concave surface refers to the fact that the force-bearing contact surface of the pressure head is machined into an inwardly concave arc shape that matches the curvature of the outer circumference of the bearing outer ring, thereby avoiding localized pressure damage, load concentration, or unstable force application caused by point or line contact. By adopting a series structure of the upper pressure plate, polyurethane gasket, and arc-shaped pressure head, the mechanical impact of the loading mechanism itself can be suppressed while ensuring load transmission accuracy, making the pulses applied to the bearing outer ring smoother, more controllable, and more repeatable.

[0195] This structure helps reduce false abnormal responses caused by rigid collisions of the loading head, improves the distinguishability between pulsed loads and real rolling contact anomalies, and thus improves the reliability of early defect development and detection.

[0196] Please refer to Figure 2 In one embodiment, a differential operation is performed on the second response signal and the first response signal to obtain the response increment, including:

[0197] Step S410: Extract feature parameters used to characterize the signal strength from the first response signal to obtain the first feature value.

[0198] Step S420: Extract feature parameters used to characterize the signal strength from the second response signal to obtain the second feature value.

[0199] Step S430: Subtract the second feature value from the first feature value to obtain the feature difference value.

[0200] Step S440: The ratio of the feature difference to the first feature value is used as the response increment.

[0201] In this embodiment, feature extraction refers to calculating numerical features that characterize the local contact state from the original waveform within the measured angle window, rather than directly using the entire original waveform as the final judgment criterion. The original waveform is mainly used to check for conditions such as sensor saturation, direct impact from the loading head, single-point spikes, or abnormal long-term attenuation; only data that passes the validity check enters the feature extraction process. By first screening the original waveform for validity and then extracting the window feature values, both anti-interference capability and judgment stability can be considered simultaneously.

[0202] In one embodiment, for the first The first angle window and the first The original response waveform in the base load cycle of the group detection cycle can be denoted as: The original response waveform in the pulse load loop can be denoted as: .in, Indicates the first ring of the foundation load. The original waveform of the first response signal within each angle window; Indicates the first pulse load cycle The original waveform of the second response signal within each angle window; This indicates the local sampling time within that angle window.

[0203] When checking the validity of the original waveform, one or more of the following criteria can be used: determine whether the response channel exhibits clipping saturation; determine whether isolated spikes caused by non-contact interference occur; determine whether the structural response decays to below the background threshold within a preset decay time after pulse triggering. If the above validity requirements are not met, the corresponding window data can be marked as interference data and will not be included in the defect determination.

[0204] In one embodiment, to avoid misidentifying non-defect anomalies caused by the pulse loading mechanism itself, the instant of contact between the loading head and the bearing outer ring, the force transmission path of the fixture, and structural residual vibration as local bearing defects, an interference rejection step can be set after the validity check of the original waveform and before feature extraction. This interference rejection step may include: establishing a loading interference reference template under normal conditions, identifying significant interference components that occur synchronously with the loading moment, and subtracting, shielding, or rejecting detection groups with significant synchronous interference. Through the above processing, the mechanism impact and structural residual vibration that occur synchronously with the loading action can be identified and suppressed from the second response signal, making the response entering the differential judgment more reflective of the abnormal sensitivity of the rolling contact area itself to short-term load disturbances. This avoids misidentifying non-defect anomalies caused by the pulse loading mechanism as local bearing defects, improving the accuracy and stability of early defect detection results.

[0205] In one embodiment, the characteristic parameters are the signal peak value, root mean square value, signal energy, bandwidth energy, kurtosis, or pulse count.

[0206] In one embodiment, the peak value can represent the maximum absolute value of the signal amplitude within the window; the root mean square value can represent the overall energy level of the signal within the window; the signal energy can represent the sum of squares of the discrete sampled values ​​within the window; the bandwidth energy can represent the sum of the spectral energy of the signal within a preset frequency range; the kurtosis can represent the sharpness of the signal impulse within the window; and the pulse count can represent the number of pulse events exceeding a preset amplitude threshold. If the discrete sampled sequence within the window is... The number of sampling points is Then the root mean square value and signal energy They can be represented as:

[0207]

[0208]

[0209] in,

[0210] Represents the root mean square value;

[0211] Indicates signal energy;

[0212] Indicates the first The signal amplitude at each sampling point;

[0213] This indicates the number of sampling points within the window.

[0214] In one embodiment, the vibration signal may be extracted from one or more of peak value, root mean square value, kurtosis and envelope energy; the acoustic emission signal may be extracted from one or more of acoustic emission energy, pulse count or band energy.

[0215] In one embodiment, the feature parameters extracted from the first response signal are assumed to be... Let the feature parameters extracted from the second response signal be... Then the feature difference can be expressed as:

[0216]

[0217] in,

[0218] Indicates the first The first in the group testing cycle Feature differences of each angle window;

[0219] Indicates the first The first in the group testing cycle The first feature value of each angle window;

[0220] Indicates the first The first in the group testing cycle The second feature value of each angle window.

[0221] Accordingly, when the ratio of the feature difference to the first feature value is used as the response increment, the response increment can be expressed as:

[0222]

[0223] in,

[0224] Indicates the first The first in the group testing cycle The response increment of each angle window;

[0225] Indicates the first The first in the group testing cycle Feature differences of each angle window;

[0226] Indicates the first The first in the group testing cycle The first feature value of each angle window.

[0227] In practical applications, to avoid abnormal denominators caused by excessively small first eigenvalues, normalization with a stable constant can also be used.

[0228] In one embodiment, determining whether there is an early defect at the circumferential position corresponding to the measured angle window based on the response increment includes:

[0229] Step S510: Perform multiple sets of detection cycles on normal bearings to obtain the response increments of multiple normal bearings, and take the average value of the response increments of multiple normal bearings to obtain the reference response increment.

[0230] Step S520: Statistically analyze the response increments of the same angle window under test in multiple consecutive detection cycles to obtain multiple measured response increments, and take the average value of the multiple measured response increments to obtain the average measured response increment.

[0231] Step S530: Calculate the ratio of the average measured response increment to the reference response increment to obtain the response anomaly intensity.

[0232] Step S540: The detection period in which the measured response increment is greater than the reference response increment is determined as the abnormal period, and the proportion of the abnormal period in multiple detection periods is calculated to obtain the response anomaly density.

[0233] Step S550: Obtain the peak time of the second response signal within the abnormal period, and determine the abnormal angle position based on the angular position of the bearing inner ring corresponding to the peak time.

[0234] Step S560: Calculate the standard deviation of multiple abnormal angle positions to obtain the abnormal angle dispersion.

[0235] Step S570: Calculate the ratio of the abnormal angle dispersion to the width of the window of the angle to be measured to obtain the angle instability.

[0236] Step S580: Determine the angle stability based on the angle instability.

[0237] Step S590: The response anomaly intensity, response anomaly density, and angle stability are weighted and summed to obtain the defect confidence of the angle window to be tested.

[0238] Step S600: When the defect confidence level is greater than or equal to the preset confidence threshold, it is determined that there is an early defect at the circumferential position corresponding to the test angle window.

[0239] In one embodiment, let the first The average normalized differential response of each angle window across multiple detections is: The average normalized differential response of the suspected window across multiple detections is: The reference response increment is Then the first The intensity of the response anomaly in each angle window can be expressed as:

[0240]

[0241] in,

[0242] Indicates the first The abnormal intensity of the response of each angle window;

[0243] This represents the average normalized differential response of the suspected window across multiple detections;

[0244] Indicates the reference response increment;

[0245] This represents the stability constant.

[0246] The intensity of the response anomaly is used to characterize the degree of enhancement of a suspected anomaly window relative to a normal window. If... A larger value indicates that the corresponding window exhibits a significantly enhanced response under impulse perturbation compared to the normal background level, and is more likely to correspond to a real local anomaly location.

[0247] In one embodiment, let the first The standard deviation of the abnormal angles corresponding to the angle windows in multiple detections is: The width of the window for the angle to be measured is Then the angular stability can be expressed as:

[0248]

[0249] in,

[0250] Indicates the first Angular stability of each angle window;

[0251] This represents the standard deviation of the angle corresponding to the abnormal window across multiple detection sets.

[0252] Indicates the width of the angled window.

[0253] when The closer the value is to 1, the more stably the anomaly appears near the same angular window. In other words, the anomaly has better repeatability in terms of angle, which is more consistent with the stable development characteristics of local defects.

[0254] In one embodiment, let the first The average normalized difference response in multiple detections across multiple angle windows is The differential judgment threshold is The number of detection groups exceeding the threshold in this window is The total number of test groups is Angular stability is Then the first The confidence level of defects in a given angle window can be expressed as:

[0255]

[0256] And satisfy:

[0257]

[0258] in,

[0259] Indicates the first Confidence level of defects in each angle window;

[0260] This represents the average normalized differential response across multiple detections within the window.

[0261] Indicates the differential judgment threshold;

[0262] This indicates the percentage of times the window repeatedly exceeds the threshold.

[0263] Indicates angular stability;

[0264] , , This represents the weighting coefficient.

[0265] By weighting and summing the response anomaly intensity, recurrence rate, and angular stability, a more comprehensive defect judgment index can be formed. When the aforementioned defect confidence level reaches or exceeds a preset confidence threshold, it can be determined that an early defect exists at the circumferential position corresponding to the tested angle window.

[0266] In one embodiment, the energy within a short time window at the moment of micropulse triggering is . The response energy within the delay time window after the micropulse is Then the first The post-pulse response retention coefficient for each angle window can be expressed as:

[0267]

[0268] in,

[0269] Indicates the first Micro-pulse response retention coefficient for each angle window;

[0270] This represents the energy within a short time window at the moment of micropulse triggering;

[0271] This represents the response energy within the delay time window after the micropulse.

[0272] This represents the stability constant.

[0273] like If the value is too small, it indicates that the anomaly is mainly concentrated in the loading instant, possibly due to a loading head impact; if... A consistently high value across multiple tests suggests a more likely sustained response caused by localized defects in the bearing. This coefficient can serve as an auxiliary criterion to differentiate between instantaneous loading impact and responses to localized defects.

[0274] In one embodiment, let the first The response anomaly intensity of each angle window is Then its defect severity score can be expressed as:

[0275]

[0276] And satisfy:

[0277]

[0278] in,

[0279] Indicates the first Defect severity score for each angle window;

[0280] This indicates the degree to which the differential response exceeds the threshold;

[0281] Indicates the proportion of repetitions;

[0282] Indicates the intensity of the abnormal response;

[0283] , , This represents the weighting coefficient.

[0284] Defect severity scores can be used to sort multiple anomaly windows, thus providing a reference for subsequent retesting, disassembly, or maintenance priority determination.

[0285] In one embodiment, if the 18th angle window is selected as the angle window to be measured, then the angle window to be measured corresponds to the principal axis angle reference system. to The interval. Within a detection cycle, there may be adjacent base load cycles and pulse load cycles. The base load cycle is used to acquire the reference response under the action of the base load, and the pulse load cycle is used to acquire the enhanced response under the combined action of the base load and the additional pulse load.

[0286] In the basic load cycle, when the encoder detects that the spindle has rotated to the position corresponding to the 18th angle window... to When the angle interval is defined, the acquisition and calculation unit acquires the sensor output signal within that angle interval and determines the response signal within that interval as the first response signal corresponding to the 18th angle window. Optionally, the first response signal may be... to A segment of original time-domain vibration signal acquired within the interval can also be window response data obtained after truncation, filtering, or integration of the original signal within the interval.

[0287] Within the pulse load loop, the control unit issues a trigger command based on a predetermined trigger angle position, causing the additional pulse load, after considering lead compensation, to actually act on the 18th angle window. to The interval. When the encoder detects the spindle rotating to the desired position again... to When the interval is defined, the acquisition and calculation unit acquires the sensor output signal within the angular interval and determines the response signal within the interval as the second response signal corresponding to the 18th angular window.

[0288] In this embodiment, both the first response signal and the second response signal originate from the same angle window under the same principal axis angle reference frame, that is, both originate from the 18th angle window. to Therefore, the response increment calculated based on the first and second response signals can be determined as the response increment corresponding to the 18th angular window. In other words, the response increment corresponding to the measured angular window means that the response increment is obtained by comparing two sets of response signals of the same measured angular window in the basic load circle and the pulse load circle, rather than by comparing response signals between different angular windows.

[0289] In one embodiment, feature values ​​characterizing the window response intensity can be extracted from the first response signal and the second response signal, respectively, and the response increment can be calculated based on the difference between the two, their ratio, or other quantities reflecting the degree of response change. For example, the response increment can be calculated as follows: to The peak value, root mean square value, envelope peak value, energy value, or frequency band integral value of the vibration signal within the interval are used as window characteristic values. If the characteristic value of the 18th angle window in the base load loop is A1, and the characteristic value of the 18th angle window in the pulse load loop is A2, then A2 minus A1 can be used to determine the differential response corresponding to the 18th angle window; or, A2 minus A1 and then divided by A1 can be used to determine the normalized differential response corresponding to the 18th angle window.

[0290] For example, in one embodiment, the envelope energy characteristic value of the 18th angular window in the basic load loop is 0.82, and the envelope energy characteristic value in the pulse load loop is 1.37. Then, the response increment corresponding to the measured angular window can be recorded as:

[0291]

[0292] If a normalized form is used, it can be written as:

[0293]

[0294] Since both of the above feature values ​​are taken from the 18th angular window to Therefore, the obtained 0.55 or the corresponding normalized result should be understood as the response increment corresponding to the 18th angular window.

[0295] In one embodiment, the acquisition and calculation unit can perform the above process for each of the 36 angle windows, namely: extracting the first response signal corresponding to each angle window in the basic load circle, extracting the second response signal corresponding to the same angle window in the pulse load circle, and calculating the response increment of each angle window respectively, thereby forming a one-to-one correspondence between the response increment and the angle window. At this time, if the response increment of the 18th angle window is continuously greater than the preset increment threshold and is significantly higher than the response increment of other angle windows, it can be considered that the abnormal response is concentrated at the position corresponding to the 18th angle window.

[0296] Furthermore, if, in multiple consecutive detection cycles, the 18th angle window consistently exhibits a large response increment based on the same angle division method, the same window extraction rule, and the same corresponding calculation method, then it can be considered that there is a stable correspondence between the response increment and the angle window to be tested, and based on this, it can be determined that there is an early defect at the detection position corresponding to the angle window to be tested.

[0297] In one embodiment, the method of this application is not limited to a single-window triggering method. In a full-circle rapid screening scenario, multiple angle windows can be selected in a preset order and pulse triggering can be performed on each window. In this case, at least one of the following conditions is preferably met between two adjacent pulse triggers: the vibration response or acoustic emission response after the previous pulse has decayed to below a set background threshold; the angular interval between adjacent trigger windows is not less than a preset window interval; and the time interval between two pulses is not less than the system structure response decay time. By setting the window interval and decay waiting conditions, response aliasing between adjacent pulses can be reduced, improving the accuracy of window assignment during multi-window polling detection.

[0298] In one embodiment, the characteristic period of the corresponding component can be calculated based on the bearing geometry parameters, and the measured anomaly interval can be matched with the characteristic period. Let the measured anomaly interval be... The corresponding component's theoretical characteristic period is The relative error It can be represented as:

[0299]

[0300] in,

[0301] This represents the relative error between the measured anomaly interval and the theoretical characteristic period;

[0302] Indicates the interval between measured anomalies;

[0303] This indicates the theoretical characteristic period of the corresponding component.

[0304] In one embodiment, when When the rate is no more than 10% to 15%, the anomaly can be considered to match the motion pattern of the component. By combining differential enhancement, repetition, and motion cycle matching, the detection results can go beyond simply determining whether an anomaly is present, and can further provide information on suspected defect locations.

[0305] In summary, this application divides the bearing circumference into multiple angular windows, and collects the basic response and pulse disturbance response for the same measured angular window in two adjacent circumferences. Differential enhancement is used to reveal local anomaly locations. Combined with pulse amplitude constraints, pulse duration constraints, execution delay compensation, window-level feature extraction, continuous multi-set repetition confirmation, and window stability judgment, a feasible early defect detection scheme for rolling bearings is constructed. This scheme maintains high sensitivity to early, minor defects while effectively suppressing the risk of misjudgment caused by overall load increase, natural noise, and interference from loading mechanisms, demonstrating good engineering feasibility and detection reliability.

[0306] Normal bearings and bearings with early defects were inspected separately. The entire revolution was divided into multiple angular windows according to the encoder angle, and the basic load revolution response was obtained for each angular window. and pulse load loop response Then calculate the corresponding normalized difference response. Among them, the basic load cycle response This represents the response feature value extracted within the corresponding angle window when no pulse is applied; pulse load loop response. This represents the response feature value extracted after a short-time trigger pulse within the corresponding angular window, and is the normalized differential response. Used to characterize the response increment of the same angular window between the basic load cycle and the pulse load cycle.

[0307] Figure 3 The results show a comparison of the response of early defective bearings at various angular windows throughout the entire revolution, where the blue curve represents the response of the base load revolution. The orange curve represents the pulse load loop response. .Depend on Figure 3 As can be seen, within most angled windows, Compared to The changes were relatively small, but near a specific window, the pulsed load loop response was significantly higher than the basic load loop response, indicating that this local location generated a stronger response increment under pulse action, suggesting the possible presence of early defects at this location. In other words, the defective area may only exhibit a weak response under natural operating conditions, but after applying a short-duration pulse within the corresponding angular window, the change in its local contact state is further amplified, thus forming a significant abnormal peak value within that window.

[0308] Figure 4 The results show a comparison of the response of a normal bearing under various angular windows throughout the entire revolution, where the blue curve represents the response of the base load revolution. The orange curve represents the pulse load loop response. .Depend on Figure 4 It can be seen that the normal bearing in each angular window and The differences between the bearings were small and relatively uniform, with no abnormally prominent peaks in any local window. This indicates that for areas without defects or with uniform contact, the response increment caused by the pulse is usually relatively stable and does not form obvious abrupt changes in individual angular windows. Therefore, observing only the absolute response level over the entire revolution, both normal and defective bearings may exhibit some fluctuations; however, by comparing local responses within angular windows, it can be found that defective bearings show more significant differences in specific windows.

[0309] Figure 5 The normalized differential response of a normal bearing and an early defective bearing is shown. The comparison results show that one curve corresponds to the differential result of a normal bearing, and the other curve corresponds to the differential result of a defective bearing. Figure 5 It can be seen that a normal bearing, within the entire revolution range... The overall response remained at a low and stable level; however, defective bearings exhibited significantly higher peaks at certain specific angular windows compared to other windows. This result indicates that using the same angular window for differential comparison between the base load cycle and the pulsed load cycle effectively suppresses the influence of the overall background response and natural operational fluctuations, and highlights the sensitive response of early defect areas to pulsed loads. Compared to directly determining anomalies based on absolute response amplitude, the normalized differential response is more effective in separating defect areas from the overall background.

[0310] In one embodiment, a 6205 deep groove ball bearing is selected as the test object. The outer ring is fixed, and the inner ring is driven by a servo motor to rotate at 600 r / min. The base load is set to 100 N, the peak pulse load is set to 15 N, the rise time is 1.5 ms, the hold time is 4 ms, the fall time is 1.5 ms, and the total pulse duration is 7 ms. The encoder resolution is 3600 pulses per revolution. The vibration sensor is installed radially on the outer wall of the bearing housing, and the sampling frequency is 100 kHz. The entire circle is divided into 36 angular windows, each window... Five consecutive sets of tests are performed, with two adjacent sets constituting one set. The defect location is output when the normalized differential response is greater than 0.30 for five consecutive sets and the abnormal window number fluctuates by no more than one window.

[0311] Man-made rolling track A shallow indentation defect with a width of 0.2 mm and a depth of 0.03 mm was prepared nearby. The test results showed that windows W18 and W19 showed a differential response exceeding the threshold.

[0312] Table 1 provides exemplary detection data near certain angle windows. Table 1 lists the basic load cycle responses of normal and defective bearings under different angle windows. Pulse load loop response and normalized difference response .

[0313] Table 1

[0314]

[0315] As shown in Table 1, the pulse load response is as follows within each angular window corresponding to a normal bearing. Relative to the basic load cycle response It only shows a small change, normalized difference response It also remained at a low level, and the changes between adjacent windows were relatively gradual. For example, in a window... to Inside, normal bearings The values ​​were all within a similar range, with no obvious sudden increase.

[0316] Conversely, for defective bearings, within a portion of the non-defective window, Compared to The changes were also relatively small, corresponding to Similar to a normal bearing; however, in the window near the defect location, the pulsed load loop response... A significant increase occurs, causing the normalized difference response to... Significantly increased. For example, in the window. and Defective bearing Significantly higher than the corresponding ,and The temperature is significantly higher than that of normal bearings and adjacent ordinary windows, indicating that this location is more sensitive to pulse disturbances, and the corresponding region is more likely to have early local anomalies. Meanwhile, in the window... , and At non-protruding window locations, the differential results of defective bearings are close to those of normal bearings, indicating that the anomalies identified by this method have obvious angular selectivity and are not caused by overall lifting of the entire bearing.

[0317] Therefore, this embodiment does not simply compare the absolute amplitude of a single signal cycle, but rather extracts the incremental features of a local area under short-term additional load by analyzing the response difference between the base load cycle and the pulse load cycle within the same angular window. For normal areas, this incremental feature is usually small; however, for areas with early defects, this incremental feature is significantly enhanced, thereby improving the ability to identify early defects and locate angles.

[0318] The protection scope of the early defect detection method for rolling bearings in this application is not limited to the execution order of the steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the protection scope of this application.

[0319] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0320] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for early defect detection of a rolling bearing, the rolling bearing comprising an inner ring and an outer ring, characterized in that, The early defect detection method for rolling bearings includes: The circumference of the rolling bearing is divided into multiple angle windows, and one of the angle windows is selected as the angle window to be measured. Fix the outer ring of the bearing and drive the inner ring of the bearing to rotate; Two adjacent cycles are considered as a detection cycle. In the first cycle, a base load is applied and a first response signal is acquired within the measured angle window. In the second cycle, a pulse load is superimposed on the base load within the measured angle window and a second response signal is acquired. The amplitude of the pulse load is 3% to 25% of the base load. Perform a differential operation on the second response signal and the first response signal to obtain the response increment; The response increment is used to determine whether there are early defects at the circumferential position corresponding to the angle window to be tested.

2. The method for early defect detection of rolling bearings according to claim 1, characterized in that, The duration of the pulse load is no greater than 80% of the rotation time corresponding to the angle window to be measured.

3. The method for early defect detection of rolling bearings according to claim 1, characterized in that, The angle to be measured can be a single angle window or multiple angle windows selected in a preset order.

4. The method for early defect detection of rolling bearings according to claim 1, characterized in that, The base load and the pulse load are applied to the outer ring of the bearing and are applied radially toward the center of the bearing.

5. The method for early defect detection of rolling bearings according to claim 1, characterized in that, Prior to performing the steps of the aforementioned inspection cycle, the early defect detection method for rolling bearings further includes the following steps: The controller sends a pulse loading command to the pulse loading mechanism and records the time of command transmission. The pulse loading mechanism applies the pulse load to the outer ring of the bearing according to the pulse loading command; The actual loading moment when the pulse load is applied to the outer ring of the bearing is obtained by a force sensor fixed to the outer ring of the bearing; The difference between the loading time and the instruction sending time is calculated to obtain the execution delay time of the pulse loading mechanism; The angular position and angular velocity of the inner ring of the bearing are obtained by an encoder, which is mounted on a spindle connected to the inner ring of the bearing. Calculate the angular advance amount based on the execution delay time and the angular velocity; During the detection cycle, the early defect detection method for the rolling bearing further includes the following steps: Determine the starting angle position corresponding to the angle window to be measured; The advance angle position for sending the pulse loading command is determined based on the starting angle position and the angle advance amount. When the angular position measured by the encoder reaches the advance angular position, the controller sends the pulse loading command to the pulse loading mechanism so that the pulse load falls within the measured angle window.

6. The method for early defect detection of rolling bearings according to claim 4, characterized in that, The controller applies the basic load to the outer ring of the bearing via a cylinder, and the pulse loading mechanism is a voice coil motor connected between the cylinder and the outer ring of the bearing.

7. The method for early defect detection of rolling bearings according to claim 6, characterized in that, The push rod end of the voice coil motor is connected to the outer ring of the bearing in sequence through a steel upper pressure plate, a polyurethane gasket, and a steel arc-shaped pressure head. The steel arc-shaped pressure head has an arc-shaped concave surface that matches the outer circumferential surface of the bearing outer ring.

8. The method for early defect detection of rolling bearings according to claim 1, characterized in that, The step of performing a differential operation on the second response signal and the first response signal to obtain the response increment includes: From the first response signal, feature parameters used to characterize the signal strength are extracted to obtain the first feature value; The second characteristic value is obtained by extracting the characteristic parameters used to characterize the signal strength from the second response signal. The difference between the second feature value and the first feature value is obtained; The ratio of the feature difference to the first feature value is used as the response increment.

9. The method for early defect detection of rolling bearings according to claim 8, characterized in that, The characteristic parameters are the peak value, root mean square value, signal energy, bandwidth energy, kurtosis, or pulse count of the signal.

10. The method for early defect detection of rolling bearings according to claim 8, characterized in that, The step of determining whether there is an early defect at the circumferential position corresponding to the measured angle window based on the response increment includes: Multiple test cycles are performed on normal bearings to obtain the response increments of multiple normal bearings, and the average value of the response increments of multiple normal bearings is taken to obtain the reference response increment. The response increments of the same angle window under test in multiple consecutive detection cycles are statistically analyzed to obtain multiple measured response increments. The average measured response increment is obtained by averaging the multiple measured response increments. The ratio of the average measured response increment to the reference response increment is calculated to obtain the response anomaly intensity; The detection period in which the measured response increment is greater than the reference response increment is defined as the abnormal period, and the proportion of the abnormal period in multiple detection periods is calculated to obtain the response anomaly density. Obtain the peak time of the second response signal within the abnormal period, and determine the abnormal angle position based on the angular position of the bearing inner ring corresponding to the peak time; Calculate the standard deviation of multiple abnormal angle positions to obtain the abnormal angle dispersion; The angle instability is obtained by calculating the ratio of the abnormal angle dispersion to the width of the angle window to be measured. Determine the angle stability based on the aforementioned angle instability; The defect confidence level of the test angle window is obtained by weighted summation of the response anomaly intensity, the response anomaly density, and the angle stability. When the confidence level of the defect is greater than or equal to the preset confidence threshold, it is determined that there is an early defect at the circumferential position corresponding to the angle window to be tested.

Images