Non-contact wheel profile detection device and method

Abstract

The application provides a non-contact wheel profile detection device and method, and belongs to the field of wheel profile detection.The device comprises a rack unit, a point laser measurement unit, a positioning and fixing unit and a data interaction unit.The rack unit is used to drive the point laser measurement unit to continuously measure after fixing a reference position;the point laser measurement unit is used to continuously measure wheel section point data;the positioning and fixing unit is used to fix the reference position by taking an edge arc as the reference position;and the data interaction unit is used to store the wheel section point data and input the continuously measured wheel section point data into a PC end.The application effectively solves the problem that the traditional detection device cannot continuously sample, improves work efficiency, and the data processing method provided by the application can effectively remove wheel data abnormal points, realize smooth processing of the wheel data, perform targeted segmented processing on the profile curve, improve calculation efficiency, and finally obtain a complete wheel profile curve, thereby providing data support for wheel health management.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of wheel profile detection, and particularly relates to a non-contact wheel profile detection device and method. BACKGROUND

[0002] Wheelset is the part of locomotive vehicle which contacts with steel rail, and has the functions of bearing the whole load and impact of the vehicle, guiding the vehicle to run, and generating traction force or braking force with the adhesion of steel rail. The complex wheel-rail contact environment can cause the wheel to produce wear and peeling problems. Although the actual contact area of the profile is limited, the force it bears is extremely complex, and various diseases are prone to occur, so it is particularly important to establish the wheel health management data.

[0003] In the traditional wheel health management system, the profile detection method is manual operation of a measuring ruler. The maintenance personnel judge the wheel profile state through the measuring ruler data to determine whether the wheel needs to be refinished. This detection method has low work efficiency, and may be misjudged due to subjective factors, and cannot provide reliable data support for the wheel health management system.

[0004] To solve the above problems, people have been seeking an effective method. SUMMARY

[0005] In view of the above shortcomings in the prior art, the present application provides a non-contact wheel profile detection device and method. The present application effectively solves the problem that the traditional detection device cannot continuously sample, improves the work efficiency, and the data processing method proposed by the present application can effectively remove the abnormal points of the wheel data to realize the smoothing processing of the wheel data, and the profile curve is processed in segments, the calculation efficiency is improved, and finally the complete wheel profile curve is obtained to provide data support for the wheel health management.

[0006] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows: a non-contact wheel profile detection device, comprising a rack unit, a point laser measurement unit, a positioning and fixing unit, and a data interaction unit;

[0007] The rack unit is used to drive the point laser measurement unit to continuously measure after fixing the reference position;

[0008] The point laser measurement unit is used to continuously measure the wheel cross-section point data;

[0009] The positioning and fixing unit is used to fix the reference position by taking the edge arc as the reference position;

[0010] The data interaction unit is used to store the wheel cross-section point data, and input the continuously measured wheel cross-section point data to the PC end.

[0011] The beneficial effects of the present application are: the present application uses the laser displacement sensor to identify the wheel cross-section point data after the positioning and fixing unit aligns the rear wheel, realizes the continuous identification function of the wheel cross-section point data by combining the sensor moving part, transmits the wheel cross-section point data to the PC end, processes the data by using the algorithm, obtains the wheel contour cross-section data, then compares the wheel cross-section data with the design model data, judges the wheel state through the data difference, and provides data support for the wheel health management.

[0012] Further, the rack unit comprises a rack, an upper shell located above the rack and connected with the rack, a front end head of the shell, a touch display screen and a rear end head of the shell, a lower shell located below the rack and connected with the rack, a bottom shell connected with the lower shell, and a support frame connected with the bottom shell and the lower shell; the rack is connected with the point laser measurement unit and the data interaction unit respectively, and the support frame is connected with the positioning and fixing unit.

[0013] The beneficial effects of the above further scheme are that the rack body provides support for the detection unit and the positioning unit in the present application.

[0014] Further, the point laser measurement unit comprises a linear guide rail and a motor support seat mounted on the rack respectively, a driving motor mounted on the motor support seat, a synchronous pulley connected with the driving motor, a synchronous belt idler connected with the driving motor in meshing connection, an idler mounting seat connected with the synchronous belt idler, a mounting seat connected with the linear guide rail, a synchronous belt pressing plate connected with the mounting seat, and a laser displacement sensor mounted on the mounting seat; the synchronous belt pressing plate is connected with the synchronous belt and the mounting seat; the idler mounting seat is mounted on the rack, and the synchronous pulley is connected with the driving motor.

[0015] The beneficial effects of the above further scheme are that the present application realizes the continuous scanning function of the cross-section by driving the point laser displacement sensor with the driving motor, so that the wheel cross-section detection data is more comprehensive and the data reliability is high.

[0016] Further, the positioning and fixing unit comprises a positioning rod and a magnet assembly connected with the support frame respectively; the positioning rod and the magnet assembly are locked and mounted on the support frame by screws.

[0017] The beneficial effects of the above further scheme are that when artificial detection is simulated, the detection of the traditional tool is aligned, and the magnet assembly in the present application can fix the device to realize hands-off detection, thereby reducing the work intensity during artificial detection.

[0018] The present application also provides a non-contact wheel contour detection method, comprising the following steps:

[0019] S1, using the edge arc as the reference position, aligning the wheel rim outer arc part with the positioning rod, and attaching the magnet assembly to the inner side plane of the wheel and adsorbing it;

[0020] S2, driving the synchronous pulley by the driving motor, driving the synchronous belt engaged with the synchronous pulley to move, so that the mounting base of the laser displacement sensor is also displaced, and the laser displacement sensor is used to continuously measure the wheel cross-section point data;

[0021] S3, using the data interaction unit to input the continuously measured wheel cross-section point data to the PC end;

[0022] S4, in the PC end, using the moving average method to detect and remove the jump points, and according to the wheel profile characteristics, the removed wheel cross-section point data is segmented and processed to obtain the first segment of the wheel profile data, the second segment of the wheel profile data and the third segment of the wheel profile data;

[0023] S5, the first segment of the wheel profile data and the second segment of the wheel profile data are processed by the moving average method, and the third segment of the wheel profile data is processed by the quartile range method;

[0024] S6, using the locally weighted regression to process the first segment of the wheel profile data and the second segment of the wheel profile data processed by the moving average method;

[0025] S7, using the weighted linear least square method to fit the first segment of the wheel profile data, the second segment of the wheel profile data and the third segment of the wheel profile data processed to obtain a complete fitting profile curve;

[0026] S8, comparing the complete fitting profile curve obtained with the profile curve output by the standard design model to obtain the wheel surface wear data, and completing the detection of the non-contact wheel profile.

[0027] The beneficial effects of the present application are: the present application is based on the moving average method (Moving Average, MA), the quartile range method (Interquartile Rang, IQR), the locally weighted regression (Locally Weighted Regression, LWR) and the weighted linear least square fitting data processing method, which can effectively remove the abnormal points of the wheel data and realize the smoothing processing of the wheel data, and the profile curve is processed segmentally, the calculation efficiency is improved, and finally the complete wheel profile curve is obtained, the present application effectively solves the problem that the traditional detection device cannot continuously sample, improves the work efficiency, and provides data support for the health management of the wheel.

[0028] Further, the S4 comprises the following steps:

[0029] In the PC end, using the moving average method to detect the jump points, and using the large window for the initial processing of the jump points.

[0030] The wheel cross-section point data is divided into three sections at the highest point and the position 30mm before the nominal rolling circle, wherein the first section is recorded as the first section of the wheel profile data before the highest point, the second section is recorded as the second section of the wheel profile data from the highest point to the position 30mm before the nominal rolling circle, and the third section is recorded as the third section of the wheel profile data after the position 30mm before the nominal rolling circle.

[0031] Further, the third section of the wheel profile data is smoothed by using the interquartile range method, including the following steps:

[0032] The third section of the wheel profile data is sorted from small to large, divided into three division points of four equal parts, and the quartiles Q1 and Q3 are calculated, wherein Q1 represents the lower quartile and Q3 represents the upper quartile.

[0033] The interquartile range IQR is calculated by using the difference between the quartiles Q1 and Q3.

[0034] Based on the interquartile range IQR, the normal value range of the third section of the wheel profile data is defined, wherein the lower limit is , the upper limit is , and the range other than the upper and lower limits is defined as an abnormal value.

[0035] The abnormal data is cleaned, and the smoothing of the third section of the wheel profile data is completed.

[0036] The beneficial effects of the above further scheme are that the abnormal data is cleaned, the smoothing of the third section of the wheel profile data is completed, the abnormal data is effectively removed, and the finally obtained data has high reliability and is closer to the true value.

[0037] Further, the S6 includes the following steps:

[0038] A Gaussian kernel function is used to assign a corresponding weight to each wheel data sample point in the window neighborhood, wherein , , n represents the total number of wheel data sample points in the window, i represents the i-th wheel data sample; i

[0039] For each prediction point , a weight diagonal matrix is constructed.

[0040] In the field of the prediction point , a polynomial function is used to fit the wheel data.

[0041] Based on the wheel data fitting result and the constructed weight diagonal matrix, a coefficient vector matrix is solved by a weighted linear least square method​ ;

[0042] based on the obtained coefficient vector matrix and the prediction point , the prediction value ;

[0043] based on the prediction value obtained for each prediction point , the processing of the first section of wheel profile data and the second section of wheel profile data is completed.

[0044] Further, the expression of the prediction value is as follows:

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] wherein, represents the prediction value, represents the constant coefficient matrix, represents the d th order term coefficient matrix, represents the th power of the prediction point d , X represents the wheel data sample polynomial base function matrix, represents the transpose of , represents the weight diagonal matrix, represents the column vector of , consisting of the sample points , represents the th power of d , represents the n th wheel data sample point, represents the corresponding Gaussian kernel function when i = n , represents the Gaussian kernel function, represents the bandwidth parameter.

[0051] The beneficial effect of the further scheme is that the method can efficiently solve the smoothing problem of nonlinear and interference-containing wheel profile data, especially for the case that the local wheel profile data changes greatly and there is great interference in the local position. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 The method flowchart of the present application.

[0053] Figure 2 The working schematic diagram of the detector in the present application.

[0054] Figure 3 The schematic diagram of the three-dimensional structure of the detector in the present application.

[0055] Figure 4 The schematic diagram of the internal structure of the detector in the present application.

[0056] Figure 5 The schematic diagram of the rack unit structure of the present application.

[0057] Figure 6 The schematic diagram of the point laser measurement unit structure of the present application.

[0058] Figure 7 The schematic diagram of the positioning and fixing unit structure of the present application.

[0059] 1-rack unit, 101-rack, 102-upper shell, 103-touch display screen, 104-outer shell front end, 105-outer shell rear end, 106-support frame, 107-lower shell, 108-bottom shell, 2-point laser measurement unit, 201-driving motor, 202-motor support seat, 203-synchronous pulley, 204-linear guide rail, 205-mounting seat, 206-laser displacement sensor, 207-synchronous belt pressing plate, 208-synchronous belt idler, 209-idler mounting seat, 210-synchronous belt, 3-positioning and fixing unit, 301-positioning rod, 302-magnet assembly, 4-data interaction unit. DETAILED DESCRIPTION

[0060] The specific embodiments of the present application are described below to facilitate the understanding of the present application by those skilled in the art, but it should be clear that the present application is not limited to the scope of the specific embodiments, and for those skilled in the art, it is obvious that various changes are within the spirit and scope of the present application defined and determined by the appended claims, and all the inventions utilizing the concept of the present application are within the scope of protection.

[0061] Example 1

[0062] As Figure 3 , Figure 4 , Figure 5, Figure 6 and Figure 7 As shown, the present invention provides a non-contact vehicle contour detection device, including a frame unit 1, a point laser measurement unit 2, a positioning and fixing unit 3, and a data interaction unit 4;

[0063] The frame unit 1 is used to drive the point laser measurement unit 2 to perform continuous measurement after the reference position is fixed;

[0064] Point laser measurement unit 2 is used to continuously measure point data of wheel cross-section;

[0065] Positioning and fixing unit 3 is used to fix the reference position with the edge arc as the reference position;

[0066] Data interaction unit 4 is used to store wheel cross-section point data and input continuously measured wheel cross-section point data to the PC.

[0067] In this embodiment, as Figures 3 to 5 As shown, Figure 4 The diagram shows the internal structure of the detector. The frame unit 1 includes a frame 101, an upper outer shell 102 located above and connected to the frame 101, a front end head 104 of the outer shell, a touch screen display 103 and a rear end head 105 of the outer shell, a lower outer shell 107 located below and connected to the frame 101, a bottom outer shell 108 connected to the lower outer shell 107, and a support frame 106 connected to the bottom outer shell 108 and the lower outer shell 107. The frame 101 is connected to the point laser measurement unit 2 and the data interaction unit 4, and the support frame 106 is connected to the positioning and fixing unit 3.

[0068] In this embodiment, as Figure 3 and Figure 6 As shown, Figure 3 As shown in the internal structure diagram, the point laser measurement unit 2 includes a linear guide rail 204 and a motor support 202 mounted on the frame 101, a drive motor 201 mounted on the motor support 202, a synchronous pulley 203 connected to the drive motor 201, a synchronous belt idler pulley 208 meshing with the drive motor 201, an idler pulley mounting seat 209 connected to the synchronous belt idler pulley 208, a mounting seat 205 connected to the linear guide rail 204, a synchronous belt pressure plate 207 connected to the mounting seat 205, and a laser displacement sensor 206 mounted on the mounting seat 205. The synchronous belt pressure plate 207 connects the synchronous belt 210 and the mounting seat 205. The idler pulley mounting seat 209 is mounted on the frame 101, and the synchronous pulley 203 is connected to the drive motor 201.

[0069] In this embodiment, as Figure 7As shown, the positioning and fixing unit 3 includes a positioning rod 301 and a magnet assembly 302 fixedly connected with the support frame 106, and the positioning rod 301 and the magnet assembly 302 are fixedly installed on the support frame 106 by screws.

[0070] In this embodiment, the device comprises a position fixing function, a continuous detection function of the wheel cross section, and a data transmission function.

[0071] In this embodiment, the rack unit 1 comprises a rack 101 fixedly connected with an upper shell 102, a shell front end 104, a shell rear end 105, and a support frame 106, and a lower shell 107 and a bottom shell 108 fixedly connected with the support frame 106. A rack internal data interaction unit 4 is fixedly connected with the rack 101.

[0072] In this embodiment, the point laser measurement unit 2 comprises a driving motor 201 fixedly installed on a motor support seat 202, a front section of the driving motor 201 fixedly connected with a synchronous belt pulley 203, the synchronous belt pulley 203 and a synchronous belt idler 208 meshingly connected with a synchronous belt 210, the synchronous belt idler 208 fixedly connected with an idler mounting seat 209, a laser displacement sensor 206 fixedly installed on a sensor mounting seat 205, the mounting seat 205 fixedly connected with the synchronous belt 210 and a synchronous belt pressing plate 207, the mounting seat 205 further connected with a linear guide rail 204, and the linear guide rail 204, the motor support seat 202, and the idler mounting seat 209 fixedly installed on the rack 101.

[0073] In this embodiment, the positioning and fixing unit 3 comprises a positioning rod 301 and a magnet assembly 302 fixedly connected with the support frame 106.

[0074] In this embodiment, after the positioning and fixing unit 3 is aligned with the rear wheel, the driving motor 201 drives the driving wheel to rotate, drives the synchronous belt 210 to move horizontally, and drives the sensor mounting seat 205 fixedly connected with the synchronous belt 210 to move horizontally, so as to drive the laser displacement sensor 206 to move horizontally, and the laser displacement sensor 206 to recognize point data. The device combines the laser displacement sensor 206 to realize the function of continuous recognition of wheel cross section point data, transmits the wheel cross section point data to a PC, processes the data by using an algorithm, obtains wheel profile cross section data (a complete fitting profile curve), compares the wheel profile data with standard design model wheel profile data, judges the wheel state through the wheel data difference, provides data support for wheel health management, that is, obtains wheel surface wear data through the data difference, and further judges the wheel service state.

[0075] In this embodiment, the wheel profile model is a wheel profile model obtained after algorithm processing, and the standard wheel profile design model is a standard existing model, such as a standard design model in TB / T 2817-2018.

[0076] The working principle of the present application is as follows:

[0077] In the traditional manual detection, the detection ruler (such as the wheel tread profile detector in Figure 2 , the wheel tread profile detector is set with the outer circular arc of the wheel rim as the reference edge, and the inner side of the wheel is aligned with the reference surface for detection. As shown in Figure 2 , the device is used as the reference in this way, the positioning rod 301 assembly in the positioning and fixing unit 3 is aligned with the outer circular arc of the wheel rim, the magnet assembly 302 is magnetically attracted to the side plane of the wheel to fix the device at the measurement position, the laser displacement sensor 206 is started, the wheel cross-section point data on the path of the moving unit is continuously measured, the internal data interaction unit 4 of the device transmits the measurement data to the PC end, and the algorithm is used to jump point processing and curve smoothing of the abnormal values and data fluctuations existing during sensor measurement, then the profile curve is segmented and processed, the integrity of the data is retained, and finally the wheel profile curve is obtained. Then compare the wheel profile data with the design data, judge the wheel state through the data difference, and provide data support for the wheel health management system.

[0078] Example 2

[0079] As shown in Figure 1 , the present application provides a non-contact wheel profile detection method, and the implementation method is as follows:

[0080] S1, using the edge circular arc as the reference position, using the positioning rod 301 to align the outer circular arc part of the wheel rim, and using the magnet assembly 302 to stick flat to the inner side plane of the wheel and be attracted;

[0081] S2, using the driving motor 201 to drive the synchronous pulley 203, the synchronous pulley 203 drives the meshing synchronous belt 210 to move, so that the mounting seat 205 of the laser displacement sensor 206 also displaces, and the laser displacement sensor 206 is used to continuously measure the wheel cross-section point data;

[0082] S3, using the data interaction unit 4 to input the continuously measured wheel cross-section point data to the PC end;

[0083] S4, in the PC end, using the moving average method to detect and remove the jump points, and according to the wheel profile characteristics, the removed wheel cross-section point data is segmented and processed to obtain the first segment wheel profile data, the second segment wheel profile data and the third segment wheel profile data, and the implementation method is as follows:

[0084] In the PC end, using the moving average method to detect the jump points, and using the large window to preliminarily process the higher jump points.

[0085] The wheel cross-section point data is divided into three sections with the highest point and the position 30 mm before the nominal rolling circle, wherein the first section is recorded as the first section of the wheel profile data, the second section is recorded as the second section of the wheel profile data from the highest point to the position 30 mm before the nominal rolling circle, and the third section is recorded as the third section of the wheel profile data after the position 30 mm before the nominal rolling circle.

[0086] S5, the first section of the wheel profile data and the second section of the wheel profile data are processed by moving average method, and the third section of the wheel profile data is processed by using quartile range method; wherein the third section of the wheel profile data is smoothed by using quartile range method, and the implementation method is as follows:

[0087] The third section of the wheel profile data is sorted from small to large, divided into three division points of four equal parts, and the quartiles Q1 and Q3 are calculated, wherein Q1 represents the lower quartile, and Q3 represents the upper quartile;

[0088] The difference between the quartiles Q1 and Q3 is used to calculate the interquartile range IQR;

[0089] Based on the interquartile range IQR, the normal value range of the third section of the wheel profile data is defined, wherein the lower limit is , the upper limit is , and the range except the upper and lower limits is defined as an abnormal value;

[0090] The abnormal data is cleaned, and the smoothing processing of the third section of the wheel profile data is completed;

[0091] S6, the first section of the wheel profile data and the second section of the wheel profile data processed by moving average method are processed by using local weighted regression; the implementation method is as follows:

[0092] A Gaussian kernel function is used to predict the point window Each wheel data sample point in the neighborhood is given a corresponding weight, wherein , n represents the total number of wheel data samples in the window, i represents the first i wheel data sample; the smoothing process is a process of processing each data point, and the prediction point is the point that needs to be smoothed at present; the wheel data sample point is all points in the neighborhood of the point that needs to be smoothed;

[0093] For each prediction point , a weight diagonal matrix is constructed;

[0094] In the field of the prediction point , a polynomial function is used for wheel data fitting;

[0095] Based on the wheel data fitting results and the constructed weight diagonal matrix, the coefficient vector matrix is solved by weighted linear least square method ;

[0096] Based on the solved coefficient vector matrix and the prediction point , the predicted value is obtained ;

[0097] Based on the predicted value obtained for each prediction point , the processing of the first section of wheel profile data and the second section of wheel profile data is completed ;

[0098] S7, using weighted linear least square method to fit the processed first section of wheel profile data, second section of wheel profile data and third section of wheel profile data, and obtain the complete fitting profile curve

[0099] S8, comparing the obtained complete fitting profile curve with the profile curve output by the standard design model, obtaining the wheel surface wear data, and completing the detection of non-contact wheel profile.

[0100] In this embodiment, the data processing methods of moving average method (Moving Average, MA), interquartile range method (Interquartile Rang, IQR), locally weighted regression (Locally Weighted Regression, LWR) and weighted linear least square fitting are used to effectively solve the abnormal data measured by the sensor, smooth the measured data, and keep the integrity of the data in the profile curve segmentation processing. The above methods have the characteristics of good final data effectiveness, fast data processing speed and strong anti-noise ability.

[0101] In this embodiment, two positioning rods 301 are used to position the tangent position of the wheel rim outer circle arc, and the magnet assembly 302 is tightly attached to the inner side of the wheel to be flat and adsorbed. This structure has the advantages of fast installation, simple operation and accurate positioning. It effectively reduces the work intensity of manual operation and avoids the data anomaly caused by manual positioning error.

[0102] In this embodiment, the detection device uses the edge arc as the reference position, and uses two positioning rods 301 to align the outer arc part of the wheel rim, and then the magnet assembly 302 is attached to the inner side of the wheel and is attracted, so as to realize the fixation of the device in the working position. The driving motor 201 drives the synchronous belt wheel 203 to drive the synchronous belt 210 to displace, and the laser displacement sensor 206 mounting seat 205 installed at the synchronous belt 210 also moves, continuously measuring the wheel section point data, and the measurement wheel section point data is transmitted to the data interaction unit 4 through the connection line, and the data interaction unit 4 transmits the measurement wheel section point data to the PC end, and processes the sensor data by Moving Average (MA) Interquartile Rang (IQR) Locally Weighted Regression (LWR) and weighted linear least square fitting method. The specific steps of the method are as follows:

[0103] Step one, first, the moving average method is used to detect the jump point, and the wheel section point data is processed by large window iteration, the large window is set for processing high jump points, the window is set to 50, and the threshold is 10mm. Since there are obvious jump points before and after the highest point, in order to improve the calculation rate, combined with the actual measurement of the laser displacement sensor 206, the wheel section point data is divided into three parts: the highest point before the first segment wheel profile data, the highest point to the nominal rolling circle 30mm position, and the third wheel data after the nominal rolling circle 30mm position.

[0104] Step two, the first segment wheel profile data and the second segment wheel profile data are further processed, the window is set to 5, the threshold is 0.5, and the small range change jitter is processed.

[0105] Since the third segment wheel profile data is relatively stable, the interquartile range method is used to smooth the third segment wheel profile data, and the specific steps are as follows:

[0106] 1), sort the third segment wheel profile data from small to large, divide the four equal parts of the split point, and calculate the quartiles: Q1 (lower quartile): the value at the 25% position, Q3 (upper quartile): at the 75% position;

[0107] 2), calculate the interquartile range (IQR) IQR=Q3-Q1;

[0108] 3), define the normal value range, wherein the lower limit is: , and the upper limit is: , and the remaining range is defined as an abnormal value;

[0109] 4) Data cleaning for abnormal data;

[0110] Step three, further remove and smooth the abnormal wheel profile data by using local weighted regression. In order to improve the calculation efficiency, only the first wheel profile data and the second wheel profile data are processed by local weighted regression. As a non-parametric learning method, the core idea of local weighted regression is to assign different weights to the local data points near each prediction point, and fit the local model by weighted linear least squares method, so as to capture the local characteristics of the data. For each prediction point Construct a special local approximation model. The specific implementation effect in the profile curve is as follows:

[0111] 1) Use Gaussian kernel function: Assign corresponding weights to each sample point in the neighborhood of the prediction point Window Each sample point in the neighborhood , where, denotes the bandwidth parameter, which is used to control the speed of weight decay (bandwidth parameter), denotes the Gaussian kernel function, denotes the n wheel data sample point, , n denotes the total number of wheel data samples in the window, i denotes the i wheel data sample;

[0112] 2) Construct weight matrix

[0113] For each prediction point , construct the weight diagonal matrix ;

[0114] 3) Local polynomial fitting

[0115] In the neighborhood of the prediction point , use polynomial function for data fitting, that is, for each in the neighborhood,

[0116] ;

[0117] where, denotes the polynomial degree (here, in order to stabilize the data fitting, take );

[0118] Design sample polynomial basis function matrix ;

[0119] Coefficient vector matrix ;

[0120] 4) Solve the coefficients by weighted linear least squares method: minimize the weighted error sum of squares in the local neighborhood: ;

[0121] Take the derivative and let it be zero, get ;

[0122] 5) Data prediction: substitute each prediction point into the fitting polynomial to get the predicted value ;

[0123] where, represents the predicted value, represents the constant coefficient matrix, represents the d order coefficient matrix, represents the power of d , X represents the wheel data sample polynomial basis function matrix, represents the transpose of , represents the weight diagonal matrix, represents the column vector of , , , represents the power of d , represents the n th wheel data sample point, represents the corresponding Gaussian kernel function when i n , represents the Gaussian kernel function, represents the bandwidth parameter.

[0124] Repeat the above steps to get the predicted value corresponding to each prediction point , and get the fitting profile curve of the first and second wheel profile data.

[0125] Step four, for the processed three-segment wheel profile data, perform global data fitting processing by weighted linear least squares method to ensure the continuity and smoothness of the overall data.

[0126] After data processing, compare the obtained wheel profile model with the design model, judge the wheel state through model difference, and provide data support for the wheel health management system.​​

Claims

1. A non-contact vehicle profile detection method, which is applied to a non-contact vehicle profile detection device, the non-contact vehicle profile detection device comprising a frame unit (1), a point laser measurement unit (2), a positioning and fixing unit (3), and a data interaction unit (4); the frame unit (1) is used to drive the point laser measurement unit (2) to perform continuous measurement after fixing the reference position; the point laser measurement unit (2) is used to continuously measure the point data of the wheel cross section; the positioning and fixing unit (3) is used to fix the reference position with the edge arc as the reference position; The data interaction unit (4) is used to store wheel cross-section point data and input continuously measured wheel cross-section point data to the PC terminal; the point laser measurement unit (2) includes a linear guide rail (204) mounted on the frame (101), a drive motor (201), a synchronous pulley (203) connected to the drive motor (201), and a laser displacement sensor (206) mounted on the mounting base (205); the positioning and fixing unit (3) includes a positioning rod (301) and a magnet assembly (302) respectively connected to the support frame (106); characterized in that, The non-contact vehicle profile detection method includes the following steps: S1. Using the edge arc as a reference position, use the positioning rod (301) to align with the outer arc of the wheel rim, and then attach the magnet assembly (302) to the inner plane of the wheel and attract it. S2. Drive the synchronous pulley (203) with the drive motor (201). The synchronous pulley (203) drives the meshing synchronous belt (210) to move, so that the mounting base (205) of the laser displacement sensor (206) also moves accordingly. The laser displacement sensor (206) continuously measures the data of the wheel section point. S3. Use the data interaction unit (4) to input the continuously measured wheel section point data to the PC terminal; S4. On the PC, the moving average method is used to detect and remove jump points, and the removed wheel section point data is segmented according to the vehicle profile features to obtain the first segment of vehicle profile data, the second segment of vehicle profile data, and the third segment of vehicle profile data. S5. The first segment of vehicle profile data and the second segment of vehicle profile data are processed using the moving average method, and the third segment of vehicle profile data is processed using the interquartile range method. S6. Use local weighted regression to process the first segment of vehicle profile data and the second segment of vehicle profile data after processing by the moving average method. S7. Use the weighted linear least squares method to fit the processed first segment of vehicle profile data, the second segment of vehicle profile data, and the third segment of vehicle profile data to obtain a complete fitted profile curve. S8. Compare the obtained complete fitted profile curve with the profile curve output by the standard design model to obtain wheel surface wear data and complete the detection of the non-contact vehicle profile.

2. The non-contact vehicle profile detection method according to claim 1, characterized in that, S4 includes the following steps: On the PC, the moving average method is used to detect jump points, and the wheel section point data is processed by iterative processing with large and small windows. The large window is set to initially process jump points. The wheel section data is divided into three segments based on the highest point and the position 30mm in front of the nominal rolling circle. The segment before the highest point is recorded as the first segment of the vehicle profile data, the segment from the highest point to the position 30mm in front of the nominal rolling circle is recorded as the second segment of the vehicle profile data, and the segment after the position 30mm in front of the nominal rolling circle is recorded as the third segment of the vehicle profile data.

3. The non-contact vehicle profile detection method according to claim 1, characterized in that, The process of processing the third segment of the vehicle profile data using the interquartile range method includes the following steps: The third segment of the vehicle profile data is sorted from smallest to largest, divided into three dividing points of four equal parts, and the quartiles Q1 and Q3 are calculated, where Q1 represents the lower quartile and Q3 represents the upper quartile. Calculate the interquartile range (IQR) using the difference between quartiles Q1 and Q3; Based on the interquartile range (IQR), the normal value range of the third segment of vehicle profile data is defined, where the lower bound is... The upper boundary is Values ​​outside the upper and lower bounds are defined as outliers; The abnormal data was cleaned, and the smoothing process of the third segment of vehicle profile data was completed.

4. The non-contact vehicle profile detection method according to claim 1, characterized in that, S6 includes the following steps: Using a Gaussian kernel function, the predicted points are... window Each wheel data sample point in the neighborhood Assign corresponding weights, where , n This indicates the total number of wheel data samples within the window. i Indicates the first i One wheel data sample; For each prediction point Construct a weight diagonal matrix; At the prediction point In the field of [the study], polynomial functions are used for wheel data fitting; Based on the wheel data fitting results and the constructed weight diagonal matrix, the coefficient vector matrix is ​​solved using the weighted linear least squares method. ; Based on the coefficient vector matrix obtained by the solution and prediction points , to obtain the predicted value ; Based on each prediction point The predicted values ​​obtained The processing of the first segment of vehicle profile data and the second segment of vehicle profile data is completed.

5. The non-contact vehicle profile detection method according to claim 4, characterized in that, The predicted value The expression is as follows: ； ； ； ； ； in, Indicates the predicted value. Represents a matrix with constant coefficients. express d The coefficient matrix of the second term Indicates the prediction point of d Power of 1 X Represents the polynomial basis function matrix of the wheel data samples. express transpose, This represents the weight diagonal matrix. Represents sample points Composition column vectors, , express of d Power of 1 Indicates the first n Each wheel data sample point, express i = n The corresponding Gaussian kernel function at time , Represents the Gaussian kernel function. This represents the bandwidth parameter.

6. The non-contact vehicle profile detection method according to claim 1, characterized in that, The frame unit (1) includes a frame (101), an upper shell (102) located above the frame (101) and connected to the frame (101), a front end head (104) of the shell, a touch screen display (103) and a rear end head (105) of the shell, a lower shell (107) located below the frame (101) and connected to the frame (101), a bottom shell (108) connected to the lower shell (107), and a support frame (106) connected to the bottom shell (108) and the lower shell (107); the frame (101) is connected to the point laser measurement unit (2) and the data interaction unit (4) respectively, and the support frame (106) is connected to the positioning and fixing unit (3); The point laser measurement unit (2) further includes a motor support base (202), a synchronous belt idler pulley (208) meshing with the drive motor (201), an idler pulley mounting base (209) connected to the synchronous belt idler pulley (208), a mounting base (205) connected to the linear guide rail (204), and a synchronous belt pressure plate (207) connected to the mounting base (205). The synchronous belt pressure plate (207) connects the synchronous belt (210) and the mounting base (205). The idler pulley mounting base (209) is mounted on the frame (101), the synchronous belt pulley (203) is connected to the drive motor (201), and the drive motor (201) is mounted on the motor support base (202). The positioning rod (301) and the magnet assembly (302) are mounted to the support frame (106) by screws.

Images