A bearing groove curvature radius high-precision measurement method and system for small-arc large-radius working conditions

Abstract

This invention discloses a high-precision measurement method and system for the curvature radius of bearing raceways under small arc and large radius operating conditions. After acquiring contour data at a fixed location, the method automatically identifies the horizontal reference sections on both sides of the raceway through local slope analysis, accurately locating the raceway apex. Subsequently, a 60° core region, symmetrically truncated around the apex to minimize the impact of end effects, is extracted. Based on this, a layered filtering strategy of "MAD coarse sieve + 3σ fine sieve" is employed to eliminate flying spots and burrs caused by specular reflection at the algorithmic level. Finally, Levenberg-Marquardt geometric circle fitting is performed on the high-purity data, fundamentally solving the fitting accuracy bottleneck under small arc and large radius operating conditions. To fundamentally ensure measurement accuracy, the levelness of the turntable is adjusted using a level to eliminate tilt errors; an adaptive iterative method of "rotation search—curvature extreme value feedback" is used to eliminate the eccentricity error between the measured bearing and the turntable.

Description

Technical Field

[0001] This invention belongs to the field of precision instrument manufacturing and testing measurement technology, and relates to a method and system for measuring the geometric parameters of aerospace bearings, specifically a method and system for measuring the radius of curvature of the inner ring groove of a deep groove ball bearing based on line structured light sensing technology. Background Technology

[0002] In the field of high-end bearing manufacturing, the radius of curvature of the inner ring of a deep groove ball bearing is one of the core geometric parameters that determines its load-bearing capacity, rotational accuracy, and service life. High-reliability applications such as aerospace bearings and precision machine tool spindle bearings place increasingly stringent requirements on the accuracy of the groove curvature radius measurement and its consistency throughout its entire lifecycle.

[0003] Non-contact measurement methods based on line structured light sensing have gradually become the mainstream technology for detecting the geometric parameters of bearing raceways due to their advantages such as high efficiency, non-destructive nature, and online deployment. The basic measurement principle is as follows: a line structured light sensor projects laser stripes onto the surface of the bearing raceway; a camera captures images of the deformed light stripes; the pixel offset is converted into a two-dimensional coordinate point cloud of the profile cross-section using the principle of triangulation; and then the radius of curvature of the raceway is obtained through a circle fitting method.

[0004] Existing methods for measuring the radius of curvature of bearing raceways have the following drawbacks:

[0005] 1. Specular reflection interference leading to data contamination: The surface of the bearing raceway has extremely high smoothness and metallic luster, making it prone to specular reflection and stray light when illuminated by structured line light. This results in a large number of flying points and outliers in the acquired contour data. Existing solutions mainly include: spraying developer (damaging the workpiece surface and adding processes) and adding polarizers (complex operation and limited effectiveness), but all of these solutions have limitations to varying degrees.

[0006] 2. Bowl-shaped groove structure leads to fitting error: The bearing groove has a bowl-shaped (concave arc surface) structure, with a distinct curvature transition region between the bottom and edge. Data in the edge region is significantly affected by machining marks, surface roughness, and edge chamfering, and cannot accurately reflect the curvature characteristics of the groove bottom. However, existing methods typically perform overall fitting of the extracted complete groove point cloud without specifically selecting the data regions involved in the fitting, resulting in low-quality data from the edge transition region introducing a non-negligible fitting error.

[0007] 3. The "small arc, large radius" condition limits fitting accuracy and stability: The central angle corresponding to the arc of the bearing raceway is typically only 60°~120°, while the radius of curvature is much larger than the arc length, which is a typical "small arc, large radius" fitting problem. Under this condition, even small amounts of data noise can cause drastic drift in the fitted center position, resulting in significant deviations in the radius calculation results. The least squares algebraic fitting method commonly used in existing technologies has poor numerical stability and difficulty in guaranteeing accuracy in such ill-conditioned problems.

[0008] 4. The “global averaging” measurement mode loses spatial distribution information: The existing measurement method fits all the data collected by rotating the turntable once, and finally outputs only a single radius of curvature value. This “continuous rotation-global fitting” mode has three inherent defects: (1) When there are grinding vibration marks, grinding wheel wear or clamping deformation caused by different circumferential angles in the groove, the single average radius cannot reflect the true quality distribution; (2) When there are flying points or outliers in the contour data at a certain angle, the whole measurement result will be contaminated by the overall fitting, resulting in an unreliable radius value in the final output; (3) It is impossible to locate the specific angle position where the curvature abnormality or processing defect occurs, which is not conducive to process traceability and targeted adjustment of processing parameters.

[0009] Therefore, there is an urgent need for a bearing raceway curvature radius measurement method that can systematically address the above problems and achieve high-precision, high-robustness, and spatially resolvable automated measurement without increasing hardware complexity. Summary of the Invention

[0010] To overcome the technical defects of existing linear structured bearing groove curvature radius measurement methods, such as low measurement accuracy, poor robustness, and lack of point information caused by specular reflection interference, bowl-shaped groove end effect, ill-conditioned fitting of small arc and large radius, and global averaging measurement mode, this invention provides a high-precision measurement method and system for bearing groove curvature radius under small arc and large radius conditions.

[0011] The objective of this invention is achieved through the following technical solution:

[0012] A high-precision measurement method for the curvature radius of bearing raceways under small arc and large radius operating conditions includes the following steps:

[0013] Step 1, System Assembly and Tilt Calibration:

[0014] The bearing to be tested is fixed on a high-precision air-bearing turntable. A level is placed on the line structured light sensor. The level reading is observed. The direction of the line structured light strip is made parallel to the rotation axis of the turntable by adjusting the tooling screws of the line structured light sensor.

[0015] Step 2, Eccentricity Pre-calibration:

[0016] Step 2-1: Adjust the relative position of the line structured light sensor and the bearing under test in the horizontal direction, collect one frame of contour data at each position and execute the complete algorithm flow from Step 4 to Step 13 to obtain the initial value of the radius of curvature corresponding to each position.

[0017] Step 2-2: Find the maximum value in the radius of curvature sequence and its corresponding position. At this point, it is determined that the eccentricity has been eliminated.

[0018] Step 3: Select fixed points for contour data collection.

[0019] Step 3-1: Set the measurement points according to the measurement requirements. After the turntable steps to the first predetermined angle and stops, the line structured light sensor collects a frame of channel cross-sectional contour data to obtain a two-dimensional coordinate point sequence.

[0020] Step 3-2: The turntable steps to the next angle, stops, and collects data until all predetermined angles have been collected.

[0021] Step 4, Data Preprocessing:

[0022] For the collected contour sequence, null points in the original contour data are removed to obtain a valid contour point sequence;

[0023] Step 5: Automatic identification of horizontal reference segment:

[0024] Step 5-1: For each point in the effective contour point sequence, perform a first-order linear fit in its local neighborhood, and use the fitting slope as the local slope of that point.

[0025] Step 5-2: Mark points whose local slope absolute value is less than a preset threshold as horizontal candidate points;

[0026] Step 5-3: Perform connected component analysis on the horizontal candidate points to identify all continuous horizontal segments, and select the two segments with the longest lengths, which are denoted as the left horizontal segment and the right horizontal segment, respectively.

[0027] Step 5-4: Perform edge erosion treatment on both ends of the left and right horizontal segments to eliminate interference from the transition area between the horizontal and arc segments;

[0028] Step 6: Initial selection of the arc region:

[0029] Using the rightmost index of the left horizontal segment and the leftmost index of the right horizontal segment as boundaries, shrink inward by a preset number of points respectively, and extract the contour data between the two boundaries as the initial arc region;

[0030] Step 7: Iterative elimination of flying points based on the MAD criterion:

[0031] For the data in the initial arc region, perform the following iterative operations: fit the current valid data using a quadratic polynomial, calculate the fitting residual for each data point, and use a threshold of several times the median absolute deviation (MAD) of the residuals to remove data points that exceed the threshold.

[0032] Step 8: Conformal smoothing filter:

[0033] Step 8-1: Perform median filtering on the Y-coordinate sequence of the arc data after removing flying points to eliminate isolated spikes;

[0034] Step 8-2: Perform Savitzky-Golay smoothing filtering to suppress residual random noise;

[0035] Step 9: Algebraic Least Squares Coarse Fitting and Channel Type Determination:

[0036] Step 9-1: Search for the minimum and maximum Y-coordinate positions in the middle region of the smoothed arc data. If the minimum value is closer to the center, it is determined to be... Type (opening upwards), otherwise it is judged as Type (opening downwards);

[0037] Step 9-2: Perform algebraic least squares circle fitting on all smoothed arc data: Let , , , , Construct and solve the following system of linear equations:

[0038]

[0039] according to and Obtain the coarse fitted circle center and coarse fitting radius ;

[0040] Step 10: Precise positioning of the channel apex:

[0041] Step 10-1: Fit the reference line using the data of the left and right horizontal reference segments. Calculate the unit vector of the normal direction of the line. ;

[0042] Step 10-2: Coarsely fit the center of the circle Starting from the normal direction, two intersection points are obtained by intersecting the rough fitted circle. Based on the channel type determined in step 9, the intersection point that is consistent with the actual channel arc surface direction is selected as the theoretical vertex.

[0043] Step 10-3: Find the data point in the circular arc data that is closest to the theoretical vertex as the actual channel vertex. ;

[0044] Step 11: Symmetrical cutout of the 60° core region:

[0045] Calculate the channel vertex relative to the coarse-fitted circle center azimuth ,by Centered on, extract All arc data points within the range are used as the 60° core area;

[0046] Step 12, 3σ criterion for secondary purification:

[0047] Step 12-1: Perform algebraic least squares circle fitting again on the 60° core region data to obtain the temporary center and temporary radius;

[0048] Step 12-2: Calculate the radial residual from each data point to the temporary circle, and use 3 times the standard deviation as the threshold to remove data points that exceed the threshold, thus completing the secondary filtering of local micro-burrs.

[0049] Step 13, Levenberg-Marquardt geometric circle fitting:

[0050] Step 13-1: For the data after secondary purification, use the coarse fitting result from step 9. Using the initial values, the Levenberg-Marquardt algorithm is used for geometric circle fitting, minimizing the sum of squared geometric distances from all data points to the fitted circle. ;

[0051] Step 13-2: During the iteration of the Levenberg-Marquardt algorithm, the damping factor is adaptively adjusted according to the change of the objective function. The iteration terminates when the parameter change is less than the preset threshold or the maximum number of iterations is reached.

[0052] Step 13-3: Output the center of the finely fitted circle and radius of curvature .

[0053] A high-precision measuring device for bearing raceway curvature radius that implements the above method includes a line structured light sensor, a high-precision air-bearing turntable, a horizontal automatic displacement stage, an electric chuck, a vertical manual displacement stage, a horizontal manual displacement stage, an industrial control computer, and a level, wherein:

[0054] The line structured light sensor is used to project structured light onto the bearing raceway surface and collect contour data;

[0055] The high-precision air-bearing turntable is used to support and rotate the bearing being tested.

[0056] The horizontal automatic displacement stage is used to adjust the relative working distance between the line structured light sensor and the bearing under test to accommodate bearings with different inner ring diameters.

[0057] The electric gripper is used to automatically clamp the bearing being tested;

[0058] The vertical manual displacement stage is used to adjust the vertical distance between the line structured light sensor and the bearing under test to accommodate bearings at different heights.

[0059] The horizontal manual displacement stage is used to adjust the relative position of the axes between the line structured light sensor and the bearing under test to eliminate eccentricity.

[0060] The industrial control computer is used to realize the human-computer interaction of the hardware system communication and functions, control the structured light sensor to collect data at a specified location, and execute the data processing algorithms of steps 4 to 13.

[0061] The level is used to assist in adjusting the horizontal attitude of the system.

[0062] Compared with the prior art, the present invention has the following advantages:

[0063] 1. Suppressing specular reflection at the algorithm level and avoiding hardware dependence: This invention uses a hierarchical filtering strategy consisting of "MAD criterion iterative elimination" and "3σ criterion secondary purification" to achieve automatic identification and elimination of flying points and outliers caused by specular reflection at the software level. No additional hardware is required, the workpiece surface is not damaged, and the system cost and operation complexity are reduced.

[0064] 2. The 60° symmetrical truncation strategy effectively eliminates end effects: This invention achieves automated symmetrical truncation of the 60° core region by automatically identifying the horizontal reference segment and accurately locating the channel apex. This strategy directly eliminates interference from the transition region at the edge of the bowl-shaped channel, ensuring that the data used for fitting best reflects the true curvature characteristics of the channel bottom.

[0065] 3. Geometric fitting fundamentally solves the accuracy problem of small arcs with large radii: Based on thorough data purification, this invention employs the Levenberg-Marquardt algorithm for geometric circle fitting, directly minimizing the geometric distance rather than the algebraic distance. Compared to algebraic least squares, geometric fitting exhibits higher numerical stability and fitting accuracy when dealing with ill-conditioned problems such as small arcs with large radii.

[0066] 4. Conformal filtering strategy ensures the stability of the arc structure: In the smoothing filtering, this invention strictly keeps the original sampled value of the X coordinate unchanged and only filters the Y coordinate, ensuring that the arc length projection of the channel section and the horizontal position of the vertex are not changed, providing a distortion-free coordinate reference for subsequent geometric fitting.

[0067] 5. Independent measurement at specific points leads to more accurate result evaluation: This invention independently acquires, filters, and fits contour data at designated angular positions. When a cross-section has a fly-through point, the independent processing mechanism ensures that this anomaly will not contaminate the measurement results of other cross-sections through overall fitting, significantly improving system robustness. More importantly, this mode can accurately identify curvature anomaly intervals on the circumference, providing data support for grinding vibration diagnosis, grinding wheel wear monitoring, and clamping deformation evaluation.

[0068] 6. Systematic error pre-calibration ensures measurement benchmark: Before measurement, this invention adjusts the tilt and curvature extreme values ​​using a level to correct eccentricity, forming a complete installation error elimination scheme. Eccentricity calibration can be completed using only the data feedback from the measurement system itself, without the need for additional centering instruments, resulting in a high degree of automation.

[0069] 7. Full-process automation and batch processing capabilities: From horizontal benchmark identification, arc interception, flying point rejection, vertex positioning, 60° region interception to high-precision fitting, the entire process requires no manual intervention and is suitable for online inspection scenarios on production lines. Attached Figure Description

[0070] Figure 1 The diagram shows the structure of a high-precision measuring device for bearing raceway curvature radius based on hierarchical data purification and geometric fitting. 1-Line structured light sensor, 2-Electric gripper, 3-Vertical manual displacement stage, 4-Horizontal manual displacement stage, 5-High-precision air-bearing turntable, 6-Horizontal automatic displacement stage.

[0071] Figure 2 A flowchart of a high-precision measurement method for bearing raceway curvature radius based on hierarchical data purification and geometric fitting;

[0072] Figure 3 To demonstrate the processing results, (a) shows the processing effect on regular data points, and (b) shows the processing effect on data points with strong noise. Detailed Implementation

[0073] The technical solution of the present invention will be further described below with reference to the accompanying drawings, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention that do not depart from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention.

[0074] This invention provides a high-precision measurement method for the curvature radius of bearing raceways based on hierarchical data purification and geometric fitting. The core idea is as follows: After acquiring contour data at a fixed location, firstly, local slope analysis is used to automatically identify the horizontal reference sections on both sides of the raceway, accurately locating the raceway apex; then, a 60° core area with minimal end-effects is symmetrically extracted centered on the apex; based on this, a hierarchical filtering strategy of "MAD coarse sieve + 3σ fine sieve" is adopted to eliminate flying spots and burrs caused by specular reflection at the algorithmic level; finally, Levenberg-Marquardt geometric circle fitting is performed on the high-purity data, fundamentally solving the fitting accuracy bottleneck under small arc and large radius conditions. To fundamentally ensure measurement accuracy, this invention also sets up a systematic error pre-calibration step before data acquisition: adjusting the turntable's levelness with a level to eliminate tilt errors; and eliminating the eccentricity error between the measured bearing and the turntable through an adaptive iterative method of "rotation search—curvature extreme value feedback." Figure 2 As shown, the specific steps are as follows:

[0075] Step 1, System Assembly and Tilt Calibration:

[0076] The bearing under test is fixed on a high-precision air-bearing turntable using electric jaws. A level is placed on the line structured light sensor, and the level reading is observed. The direction of the line structured light stripe is made parallel to the rotation axis of the turntable by adjusting the tooling screws of the line structured light sensor, ensuring that the contour data collected subsequently corresponds to the actual axial section of the channel.

[0077] Step 2, Eccentricity Pre-calibration:

[0078] Adjust the relative positions of the sensor and the device under test in the horizontal direction, acquire one frame of contour data at each position, and execute the complete algorithm flow from steps 4 to 13 to obtain the initial value of the radius of curvature corresponding to each position; find the maximum value in the radius of curvature sequence and its corresponding position, at which point it is determined that the eccentricity has been eliminated. The principle is that when the sensor sweeps across the true lowest point of the channel, the fitted radius of curvature is the largest.

[0079] Step 3: Select fixed points for contour data collection.

[0080] According to the measurement requirements, the measurement points are set. After the turntable steps to the first predetermined angle and stops, the line structured light sensor collects a frame of channel cross-sectional contour data to obtain a two-dimensional coordinate point sequence. Then the turntable steps to the next angle, stops and collects data until all predetermined angles are collected.

[0081] Steps 4 through 13 are performed independently for each frame of contour data acquired in step 3, without interference between them:

[0082] Step 4, Data Preprocessing:

[0083] For the collected contour sequence, null points in the original contour data are removed to obtain a valid contour point sequence.

[0084] Step 5: Automatic identification of horizontal reference segment:

[0085] For each point in the effective contour point sequence, a first-order linear fit is performed in its local neighborhood, and the fitted slope is used as the local slope of that point. Points with local slope absolute values ​​less than a preset threshold are marked as horizontal candidate points. Connectivity analysis is performed on the horizontal candidate points to identify all continuous horizontal segments. The two segments with the longest lengths are selected and denoted as the left and right horizontal segments, respectively. Edge erosion processing is performed on the two ends of the left and right horizontal segments to eliminate interference from the transition region between the horizontal segment and the arc segment.

[0086] Step 6: Initial selection of the arc region:

[0087] Using the rightmost index of the left horizontal segment and the leftmost index of the right horizontal segment as boundaries, shrink inward by a preset number of points, and extract the contour data between the two boundaries as the initial arc region.

[0088] Step 7: Iterative elimination of flying points based on the MAD criterion:

[0089] For the data in the initial circular arc region, the following iterative operations are performed: A quadratic polynomial is used to fit the current valid data, the fitting residual for each data point is calculated, and data points exceeding the median absolute deviation (MAD) of the residuals are discarded as a threshold. This step effectively eliminates flypoints caused by specular reflections, etc.

[0090] Step 8: Conformal smoothing filter:

[0091] Median filtering was applied to the Y-coordinate sequence of the arc data after removing flypoints to eliminate isolated spikes; subsequently, Savitzky-Golay smoothing filtering was performed to suppress residual random noise. The X-coordinate sequence was kept strictly unchanged from the original sampled values ​​to ensure that the arc length topology of the channel cross-section was not altered, providing a distortion-free coordinate reference for subsequent geometric fitting.

[0092] Step 9: Algebraic Least Squares Coarse Fitting and Channel Type Determination:

[0093] Search for the minimum and maximum Y-coordinate values ​​in the middle region of the smoothed circular arc data. If the minimum value is closer to the center, it is considered... Type (opening upwards), otherwise it is judged as Type (opening downwards). Perform algebraic least-squares circle fitting on all smoothed arc data: Let , , , , Construct and solve the following system of linear equations:

[0094]

[0095] according to and Obtain the coarse fitted circle center and coarse fitting radius ;

[0096] Step 10: Precise positioning of the channel apex:

[0097] Fit a reference straight line using data from the left and right horizontal reference segments. Calculate the unit vector of the normal direction of the line. Using a coarse fit to the center of the circle Starting from the point where the curve intersects the coarsely fitted circle along the normal direction, two intersection points are obtained. Based on the channel type determined in step nine, the intersection point aligned with the actual channel arc surface direction is selected as the theoretical vertex. The data point closest to this theoretical vertex is then found in the arc data and used as the actual channel vertex. .

[0098] Step 11: Symmetrical cutout of the 60° core region:

[0099] Calculate the channel vertex relative to the coarse-fitted circle center azimuth ,by Centered on, extract All arc data points within the range are used as the 60° core region. This region covers the most regular part of the channel bottom, excluding interference from the bowl-shaped edge transition zone.

[0100] Step 12, 3σ criterion for secondary purification:

[0101] Algebraic least squares circle fitting is performed again on the 60° core region data to obtain a temporary circle center and a temporary radius. The radial residual from each data point to this temporary circle is calculated, and data points exceeding the threshold of 3 times the standard deviation are removed, thus completing the secondary filtering of local micro-burrs.

[0102] Step 13, Levenberg-Marquardt geometric circle fitting:

[0103] For the data after secondary purification, the coarse fitting result from step 9 is used. Using the initial values, the Levenberg-Marquardt algorithm is used for geometric circle fitting. This algorithm directly minimizes the sum of the squared geometric distances from all data points to the fitted circle. Instead of algebraic distance, the damping factor is adaptively adjusted based on changes in the objective function during iteration. The iteration terminates when the parameter change is less than a preset threshold or the maximum number of iterations is reached. The output is the center of the precisely fitted circle. and radius of curvature .

[0104] The preferred values ​​of the core parameters for each step of this invention are as follows: local slope window width: 31 points; horizontal segment judgment slope threshold: 0.2; number of horizontal segment edge erosion points: 5 points; number of horizontal segment inward contraction points: 15 points; MAD iteration count: 3 times; MAD elimination threshold coefficient: 10 times the median absolute deviation; median filter window width: 5 points; Savitzky-Golay filter order: 3rd order, window width 21 points; initial value of LM algorithm damping factor: 0.01, amplification factor 10, reduction factor 0.1; LM algorithm convergence threshold: 1×10 -4 The maximum number of iterations for the LM algorithm is 80.

[0105] The present invention also provides a high-precision measuring device for the bearing raceway curvature radius that implements the above method, such as... Figure 1 As shown, the device includes a line structured light sensor 1, a high-precision air-bearing turntable 5, a horizontal automatic displacement stage 6, an electric chuck 2, a vertical manual displacement stage 3, a horizontal manual displacement stage 4, an industrial control computer, and a level. Specifically: the line structured light sensor 1 projects structured light onto the bearing raceway surface and collects contour data; the high-precision air-bearing turntable 5 carries and rotates the bearing under test; the horizontal automatic displacement stage 6 adjusts the relative working distance between the line structured light sensor 1 and the bearing under test to accommodate bearings with different inner ring diameters; the electric chuck 2 automatically clamps the bearing under test; the vertical manual displacement stage 3 adjusts the vertical distance between the line structured light sensor 1 and the bearing under test to accommodate bearings at different heights; the horizontal manual displacement stage 4 adjusts the relative position of the axes between the line structured light sensor 1 and the bearing under test to eliminate eccentricity; the industrial control computer enables communication and human-computer interaction of the hardware system, controls the line structured light sensor to collect data at designated locations, and executes the data processing algorithm of this invention; the level assists in adjusting the horizontal attitude of the system.

[0106] Example:

[0107] This embodiment uses a deep groove ball bearing as an example to describe in detail the complete measurement process of the present invention:

[0108] Step 1: Workpiece installation and attitude adjustment:

[0109] The bearing under test is fixed to a high-precision air-bearing turntable using electric jaws. A level is placed on the line structured light sensor, and the level reading is observed. By adjusting the tooling screws of the line structured light sensor, the level reading is made better than 1° to ensure that the direction of the line structured light stripe is parallel to the rotation axis of the turntable.

[0110] Step 2: Adjusting the relative position between the line structured light sensor and the bearing under test:

[0111] By using a horizontal automatic displacement stage and a hand-cranked vertical manual displacement stage, the position of the line structured light sensor is adjusted so that the laser stripe plane coincides with the axial section of the bearing groove being measured, ensuring that the light stripe is clearly imaged in the bottom area of ​​the groove.

[0112] Step 3, Eccentricity Pre-calibration:

[0113] Adjust the relative horizontal positions of the line structured light sensor and the bearing under test. Acquire one frame of contour data at each position and execute the complete algorithm flow from steps 4 to 14 to obtain the initial value of the radius of curvature corresponding to each position. Find the maximum value in the radius of curvature sequence and its corresponding position; at this point, it is determined that the eccentricity has been eliminated.

[0114] Step 4: Contour Data Acquisition

[0115] Once the turntable rotates to the target measurement angle, the line structured light sensor acquires a frame of channel cross-sectional contour data, obtaining a two-dimensional coordinate sequence of a total of 1443 contour points. .

[0116] Step 5, Data Preprocessing:

[0117] Remove null points from the original contour data to obtain a valid contour point sequence.

[0118] Step 6: Automatic identification of horizontal reference segment:

[0119] For each point in the effective contour point sequence, a first-order linear fit is performed within its local 31-point window to obtain the local slope. Points with an absolute local slope value less than 0.2 are marked as horizontal candidate points. Connectivity analysis is performed on the horizontal candidate points to identify all continuous horizontal segments. The two segments with the longest lengths are selected and denoted as the left and right horizontal segments, totaling 254 points. Five points are removed from each end of the left and right horizontal segments to eliminate interference from transition regions.

[0120] Step 7: Initial selection of the arc region:

[0121] Using the rightmost index of the left horizontal segment and the leftmost index of the right horizontal segment as boundaries, the area is contracted inward by 15 points on each side, and the contour data between the two boundaries is extracted as the initial arc region. In this embodiment, the initial arc region contains approximately 809 data points.

[0122] Step 8: Iteratively eliminate flying points using the MAD criterion:

[0123] The initial circular arc region is subjected to three MAD iterations for data removal. In each iteration, a quadratic polynomial is used to fit the current valid data, and the fitting residuals at each point are calculated. Data points exceeding the median absolute deviation of the residuals are removed as a threshold.

[0124] Step 9: Conformal smoothing filter:

[0125] The Y-coordinate sequence of the arc data after removing flypoints is subjected to median filtering with a window size of 5 to eliminate isolated spikes; then, a third-order 21-point Savitzky-Golay smoothing filter is applied to suppress residual random noise. The X-coordinate sequence remains unchanged from the original sampled values.

[0126] Step 10: Algebraic Least Squares Coarse Fitting and Channel Type Determination:

[0127] The maximum and minimum Y-coordinate values ​​are searched within a 40% range of the middle of the smoothed circular arc data. In this embodiment, the minimum value is closer to the center, thus determining the channel type as... The shape is open upwards. Algebraic least squares circle fitting is performed on all smoothed arc data points to obtain the coarse fitting circle center coordinates as (-3744.35, 88944.76) μm and the coarse fitting radius as 6342.82 μm.

[0128] Step 11, Precise positioning of the channel apex:

[0129] A baseline straight line is fitted using the data from the left and right horizontal segments. The unit vector of the normal direction of this line is calculated. Starting from the center of the coarse fitted circle, the line intersects the coarse fitted circle along the normal direction, obtaining two intersection points. Based on the channel type... The intersection point below is selected as the theoretical vertex. The data point closest to this theoretical vertex is found in the arc data and used as the actual channel vertex (-3753, 82603) μm.

[0130] Step 12, Symmetrical cutout of the 60° core region:

[0131] Calculate the azimuth angle of the channel vertex relative to the center of the coarse-fitted circle. -90.08°, with Centered on, extract All arc data points within the range are used to obtain a 60° core region. In this embodiment, It contains 504 data points.

[0132] Step 13, 3σ criterion for secondary purification:

[0133] right Algebraic least squares circle fitting was performed again, and the radial residuals of each data point were calculated. Using three times the standard deviation as a threshold, 29 local micro-spurs were removed to obtain the final fitted dataset. .

[0134] Step 14, Levenberg-Marquardt geometric circle fitting:

[0135] Using the coarse fitting results (-3744.3549, 88944.7640, 6342.82) μm as initial values, the following was performed: Perform LM geometric circle fitting. The algorithm outputs the finely fitted circle center (-3743.02, 88924.45) μm and the radius of curvature 6323.10 μm.

Claims

1. A high-precision measurement method for the radius of curvature of bearing raceways under small arc and large radius operating conditions, characterized in that... The method includes the following steps: Step 1: System assembly and tilt calibration; Step 2, Eccentricity Pre-calibration; Step 3: Select fixed points to collect contour data; Step 4, Data Preprocessing: For the collected contour sequence, null points in the original contour data are removed to obtain a valid contour point sequence; Step 5: Automatic identification of horizontal reference segment: Step 5-1: For each point in the effective contour point sequence, perform a first-order linear fit in its local neighborhood, and use the fitting slope as the local slope of that point. Step 5-2: Mark points whose local slope absolute value is less than a preset threshold as horizontal candidate points; Step 5-3: Perform connected component analysis on the horizontal candidate points to identify all continuous horizontal segments, and select the two segments with the longest lengths, which are denoted as the left horizontal segment and the right horizontal segment, respectively. Step 5-4: Perform edge erosion treatment on both ends of the left and right horizontal segments to eliminate interference from the transition area between the horizontal and arc segments; Step 6: Initial selection of the arc region: Using the rightmost index of the left horizontal segment and the leftmost index of the right horizontal segment as boundaries, shrink inward by a preset number of points respectively, and extract the contour data between the two boundaries as the initial arc region; Step 7: Iterative elimination of flying points based on the MAD criterion: For the data in the initial arc region, perform the following iterative operations: fit the current valid data using a quadratic polynomial, calculate the fitting residual for each data point, and use a threshold of several times the absolute deviation of the median residual to remove data points that exceed the threshold. Step 8: Conformal smoothing filter: Step 8-1: Perform median filtering on the Y-coordinate sequence of the arc data after removing flying points to eliminate isolated spikes; Step 8-2: Perform Savitzky-Golay smoothing filtering to suppress residual random noise; Step 9: Algebraic Least Squares Coarse Fitting and Channel Type Determination: Step 9-1: Search for the minimum and maximum Y-coordinate positions in the middle region of the smoothed arc data. If the minimum value is closer to the center, it is determined to be... Type, otherwise judged as type; Step 9-2: Perform algebraic least squares circle fitting on all smoothed arc data: Let , , , , Construct and solve the following system of linear equations: according to and Obtain the coarse fitted circle center and coarse fitting radius ; Step 10: Precise positioning of the channel apex: Step 10-1: Fit the reference line using the data of the left and right horizontal reference segments. Calculate the unit vector of the normal direction of the line. ; Step 10-2: Coarsely fit the center of the circle Starting from the normal direction, two intersection points are obtained by intersecting the rough fitted circle. Based on the channel type determined in step 9, the intersection point that is consistent with the actual channel arc surface direction is selected as the theoretical vertex. Step 10-3: Find the data point in the circular arc data that is closest to the theoretical vertex as the actual channel vertex. ; Step 11: Symmetrical cutout of the 60° core region: Calculate the channel vertex relative to the coarse-fitted circle center azimuth ,by Centered on, extract All arc data points within the range are used as the 60° core area; Step 12, 3σ criterion for secondary purification: Step 12-1: Perform algebraic least squares circle fitting again on the 60° core region data to obtain the temporary center and temporary radius; Step 12-2: Calculate the radial residual from each data point to the temporary circle, and use 3 times the standard deviation as the threshold to remove data points that exceed the threshold, thus completing the secondary filtering of local micro-burrs. Step 13, Levenberg-Marquardt geometric circle fitting: Step 13-1: For the data after secondary purification, use the coarse fitting result from step 9. Using the initial values, the Levenberg-Marquardt algorithm is used for geometric circle fitting, minimizing the sum of squared geometric distances from all data points to the fitted circle. ; Step 13-2: During the iteration of the Levenberg-Marquardt algorithm, the damping factor is adaptively adjusted according to the change of the objective function. The iteration terminates when the parameter change is less than the preset threshold or the maximum number of iterations is reached. Step 13-3: Output the center of the finely fitted circle and radius of curvature .

2. The high-precision measurement method for bearing raceway curvature radius under small arc and large radius working conditions as described in claim 1, characterized in that... The specific steps of step 1 are as follows: The bearing to be tested is fixed on a high-precision air-bearing turntable. A level is placed on the line structured light sensor, and the level reading is observed. The direction of the line structured light stripe is made parallel to the rotation axis of the turntable by adjusting the tooling screws of the line structured light sensor.

3. The high-precision measurement method for bearing raceway curvature radius under small arc and large radius working conditions as described in claim 1, characterized in that... The specific steps of step 2 are as follows: Step 2-1: Adjust the relative position of the line structured light sensor and the bearing under test in the horizontal direction, collect one frame of contour data at each position and execute the complete algorithm flow from Step 4 to Step 13 to obtain the initial value of the radius of curvature corresponding to each position. Step 2-2: Find the maximum value in the radius of curvature sequence and its corresponding position. At this point, it is determined that the eccentricity has been eliminated.

4. The high-precision measurement method for bearing raceway curvature radius under small arc and large radius working conditions as described in claim 1, characterized in that... The specific steps of step 3 are as follows: Step 3-1: Set the measurement points according to the measurement requirements. After the turntable steps to the first predetermined angle and stops, the line structured light sensor collects a frame of channel cross-sectional contour data to obtain a two-dimensional coordinate point sequence. Step 3-2: The turntable steps to the next angle, stops, and collects data until all predetermined angles have been collected.

5. The high-precision measurement method for bearing raceway curvature radius under small arc and large radius working conditions as described in claim 1, characterized in that... The .

6. A high-precision measuring device for bearing raceway curvature radius implementing the method of any one of claims 1-5, characterized in that... The device includes a line structured light sensor, a high-precision air-bearing turntable, a horizontal automatic displacement stage, an electric gripper, a vertical manual displacement stage, a horizontal manual displacement stage, an industrial control computer, and a level, wherein: The line structured light sensor is used to project structured light onto the bearing raceway surface and collect contour data; The high-precision air-bearing turntable is used to support and rotate the bearing being tested. The horizontal automatic displacement stage is used to adjust the relative working distance between the line structured light sensor and the bearing under test to accommodate bearings with different inner ring diameters. The electric gripper is used to automatically clamp the bearing being tested; The vertical manual displacement stage is used to adjust the vertical distance between the line structured light sensor and the bearing under test to accommodate bearings at different heights. The horizontal manual displacement stage is used to adjust the relative position of the axes between the line structured light sensor and the bearing under test to eliminate eccentricity. The industrial control computer is used to realize the human-computer interaction of the hardware system communication and functions, control the structured light sensor to collect data at a specified location, and execute the data processing algorithms of steps 4 to 13. The level is used to assist in adjusting the horizontal attitude of the system.

Images