Small-diameter ball head grinding wheel in-place precise dressing method and system based on multi-source information fusion

Abstract

The invention belongs to the related technical field of grinding wheel dressing, and discloses a small-diameter ball head grinding wheel in-place precise dressing method and system based on multi-source information fusion. The method comprises the following steps: S1, rough finishing: (a) obtaining the coordinate of each point on a to-be-finished ball-end grinding wheel, calculating the rough finishing amount, and finishing the to-be-finished ball-end grinding wheel according to the rough finishing amount; (b) the step (a) is repeated until the difference value between the maximum distance and the minimum distance of the small-diameter ball head grinding wheel to be finished meets the rough finishing preset threshold value requirement; s2, fine finishing: (a) acquiring slice contours of the to-be-finished ball-end grinding wheel at different latitudes; and the finishing time of the contour of each slice layer under different latitudes is calculated, and fine finishing is carried out according to the finishing time until the characteristic parameters of each slice layer of the to-be-finished ball-end grinding wheel meet the preset requirements of fine finishing. By means of the method, precise and efficient finishing of the ball-end grinding wheel is achieved, the service life of the grinding wheel is prolonged, production cost is saved, and machining quality and efficiency are guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field related to grinding wheel dressing, and more specifically, relates to an on-site precision dressing method and system for a small-diameter ball-end grinding wheel based on multi-source information fusion. Background Art

[0002] Curved surface grinding is one of the key technologies in the field of precision manufacturing, and is widely used in machinery manufacturing, aerospace, optical manufacturing and other fields. Due to the complex structural characteristics of curved surface components, traditional grinding wheels are prone to interference with workpieces during grinding, resulting in reduced processing efficiency and even damage to the workpiece. Since small-diameter ball-end grinding wheels are small in size and highly adaptable, their use is essential for maintaining the surface accuracy and surface quality of curved surface workpieces. However, as the grinding process proceeds, the small-diameter ball-end grinding wheel will gradually wear and become blunt, making it difficult to maintain effective grinding capabilities. Therefore, efficient and high-precision dressing of the small-diameter ball-end grinding wheel is a necessary step to ensure grinding quality and accuracy.

[0003] Mechanical dressing is still the main application method for in-situ dressing of small-diameter ball-end grinding wheels due to its high dressing efficiency and ease of implementation. Mechanical dressing achieves the purpose of dressing by grinding the grinding wheel against a grinding block or a dressing wheel to remove material. However, most of the existing high-precision grinding wheel dressing equipment is an offline dressing device, which requires the grinding wheel to be removed from the machine tool and installed on a special dressing device for dressing, thereby introducing clamping errors and requiring the purchase of special equipment, which incurs additional costs.

[0004] The current grinding wheel dressing process is mainly analyzed by collecting a single signal, which has certain limitations. Especially in the dressing of small-diameter ball-end grinding wheels, due to their small size, large curvature, and different processing midline speeds, traditional single monitoring methods are difficult to accurately capture their dressing status, affecting the dressing accuracy and efficiency. In summary, the current technical solutions for the dressing of small-diameter ball-end grinding wheels still have a lot of room for improvement. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides an in-situ precision dressing method and system for a small-diameter ball-end grinding wheel based on multi-source information fusion, which solves the problems of low precision and efficiency in the grinding wheel dressing process.

[0006] To achieve the above object, according to one aspect of the present invention, a method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion is provided, the method comprising:

[0007] S1 Rough trimming

[0008] (a) obtaining the coordinates of each point on the ball-end grinding wheel to be dressed, calculating the distance from each point to the center of the ball, obtaining the maximum distance and the minimum distance from the point on the surface of the grinding wheel to be dressed to the center of the ball, setting a rough dressing amount according to the maximum distance and the minimum distance, and dressing the ball-end grinding wheel to be dressed according to the rough dressing amount;

[0009] (b) repeating step (a) until the difference between the maximum distance and the minimum distance of the ball-end grinding wheel to be dressed meets the preset requirement for rough dressing;

[0010] S2 Fine Finishing

[0011] The equator of the ball-end grinding wheel to be dressed is set, and the slice contours of the ball-end grinding wheel to be dressed at different latitudes are obtained; the characteristic parameters of each slice layer contour at different latitudes are calculated, and the dressing time of each slice layer is calculated using the characteristic parameters and the preset processing parameters of the ball-end grinding wheel to be dressed, and the ball-end grinding wheel to be dressed is finely dressed according to the dressing time and the preset processing parameters of the ball-end grinding wheel to be dressed, until the characteristic parameters of each slice layer of the ball-end grinding wheel to be dressed meet the preset requirements for fine dressing.

[0012] Further preferably, in step S1, the coordinates of each point on the ball-end grinding wheel to be dressed are obtained according to the following steps:

[0013] The ball-end grinding wheel to be dressed rotates around the Y axis and moves up and down along the Y axis at the same time, and the ball-end grinding wheel to be dressed is scanned to obtain sampling points on each slice layer perpendicular to the Y direction, and the coordinates of each sampling point in the XOZ plane are obtained;

[0014] According to the speed and frequency of the up and down movement of the ball-end grinding wheel to be dressed along the Y axis, the coordinates of each sampling point direction are calculated to obtain the three-dimensional coordinates of each sampling point, that is, the coordinates of each point on the ball-end grinding wheel to be dressed.

[0015] Further preferably, the setting of the rough trimming amount according to the maximum distance and the minimum distance is performed according to the following formula:

[0016] b=c·(r max -r min )

[0017] Where b is the rough trimming amount, c is the rough trimming coefficient, and r max is the maximum distance from the point on the grinding wheel surface to be dressed to the center of the sphere, r min It is the minimum distance from the point on the grinding wheel surface to be dressed to the center of the sphere.

[0018] Further preferably, in step S2, the equator is an outer contour where a plane passing through the center of the ball-end grinding wheel to be dressed and perpendicular to the grinding wheel rod intersects with the ball-end grinding wheel to be dressed.

[0019] Further preferably, the slicing profile of the ball-end grinding wheel to be dressed at different latitudes is obtained according to the following steps: the ball-end grinding wheel to be dressed rotates while swinging in the YOZ plane, the surface of the ball-end grinding wheel to be dressed is scanned and photographed, and the slicing profiles at different latitudes are obtained after image processing.

[0020] Further preferably, the characteristic parameters include contour characteristic parameters, surface characteristic parameters and vibration characteristic parameters.

[0021] Further preferably, the characteristic parameter is calculated as follows:

[0022] The calculation formula of the contour feature parameters is as follows:

[0023] Arithmetic mean roughness:

[0024] RMS roughness:

[0025] Maximum height of the profile: Rz = max(z i )-min(z i ),

[0026] Average spacing of profile peaks: Rs = average(T i ),

[0027] Skewness of the profile:

[0028] The calculation formula of the surface characteristic parameters is as follows:

[0029] Standard Deviation:

[0030] Maximum connected domain area: S max =max(S i ),

[0031] Connected domain skewness:

[0032] Among them, z i is the height of each sampling point, z mean is the average height of the sampling points in a single cycle, T i is the duration of each cycle detected by the cycle segmentation algorithm, n is the total number of cycle sampling points, S mean is the average area of ​​the white connected domain, x c and c are the horizontal and vertical coordinates of the centroid of all white connected domains, x m and m are the horizontal and vertical coordinates of the image center, n w is the number of white connected domains in a single slice layer, Si is the area of ​​the i-th white connected region in a single slice layer;

[0033] The vibration characteristic parameters include: the main frequency of acoustic emission, the average value and fluctuation amplitude of the tangential and normal directions of the force signal, and the average value, fluctuation amplitude and main frequency of the tangential and normal directions of the acceleration signal.

[0034] Further preferably, the total number of white connected regions in the single slice layer and the area of ​​each white connected region are calculated using a seed algorithm, and the steps of the seed algorithm are as follows:

[0035] (1) Acquire a surface topography image of a single slice layer, perform image processing on the image to obtain a binary image; search for white connected areas in the binary image in the following manner:

[0036] (2) Determine whether the current pixel i is a black point or a white point;

[0037] If it is a black dot, i=i+1, determine whether i is less than n, if so, return to step (2); otherwise, end;

[0038] If it is a white spot, follow the steps below;

[0039] (a)n w =n w +1, establish a seed queue, and put the current pixel point i as the seed pixel point into the seed queue,

[0040] (b) searching for four-connected pixels adjacent to the seed pixel and determining whether the adjacent pixels are white or black;

[0041] (c) If the adjacent pixel is a white point, add the adjacent pixel to the seed queue and mark the pixel i as visited. i =s i +1; place the white adjacent node in the seed queue as the seed pixel and return to step (b), i=i+1, determine whether i is less than n, if yes, return to step (2), otherwise, end;

[0042] If the adjacent node is a black dot or a visited node, i=i+1, and determine whether i is less than n. If so, return to step (2), otherwise, end.

[0043] Further preferably, the dressing time of each slice layer is calculated by using the characteristic parameters and the preset processing parameters of the ball-end grinding wheel to be dressed according to the following formula:

[0044] t i =∑α j ·x j +∑β j ·yj +ε

[0045] Among them, t i is the prediction time for each slice trimming, x j is each feature parameter, α j is the multiple regression coefficient of each feature parameter, y j is the processing parameter, β j is the multiple regression coefficient of each processing parameter, and ε is the error term.

[0046] According to another aspect of the present invention, a system for dressing using the above-mentioned in-situ precision dressing method of a small-diameter ball-end grinding wheel based on multi-source information fusion is provided, the system comprising a rough dressing module, a fine dressing module, a visual reconstruction module and a signal measurement module, wherein:

[0047] The visual reconstruction module is used to obtain the coordinates of each point on the ball-end grinding wheel to be dressed during the rough dressing process and calculate the rough dressing amount, and to obtain the slice profile of the ball-end grinding wheel to be dressed at different latitudes during the fine dressing process;

[0048] The signal measurement module is used to calculate the characteristic parameters of each slice layer profile during the fine trimming process, and calculate the trimming time of each slice layer using the characteristic parameters;

[0049] The rough dressing module is used to perform rough dressing on the ball-end grinding wheel to be dressed according to the dressing amount calculated by the visual reconstruction module;

[0050] The fine dressing module is used to perform fine dressing on the ball-end grinding wheel to be dressed according to the dressing time calculated by the signal measuring module.

[0051] In general, the above technical solutions conceived by the present invention have the following beneficial effects compared with the prior art:

[0052] 1. The present invention combines the rough dressing and fine dressing methods to firstly perform rough dressing of the surface defects of the small-diameter ball-end grinding wheel with the grinding block, so that the contour accuracy of the grinding wheel converges quickly to the required range, and then performs fine dressing of the cup-type grinding wheel. The dressing process is measured and predicted by the multi-source information fusion method of the vibration signal and the visual signal, thereby greatly improving the dressing efficiency of the small-diameter ball-end grinding wheel.

[0053] 2. In the fine dressing process, the present invention dresses the small-diameter ball-end grinding wheel layer by layer through a layered slicing method, and realizes real-time monitoring and intelligent decision-making of the grinding wheel dressing process by integrating information from different information sources, such as visual information and vibration signals, so as to achieve the purposes of dynamically adjusting the dressing time, optimizing the dressing strategy, predicting the dressing results, etc., and provide guidance for the dressing process, improve the dressing accuracy and efficiency, and use a multi-source information fusion method combined with visual and vibration parameters to predict and guide the dressing process, which greatly improves the dressing accuracy.

[0054] 3. The seed algorithm used in the present invention searches for all white connected areas by searching the surrounding neighborhoods from the seed point pixels, and continuously expands the seed point pixel range as the search proceeds. It can realize the search and counting of white connected areas in the entire image with a smaller number of searches, and has the advantages of easy programming and high search efficiency.

[0055] 4. The device of the present invention can be freely moved and assembled, can adapt to the working environment of different occasions, and can realize the in-situ dressing of small-diameter ball-end grinding wheels, and has strong adaptability and versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of an in-situ precision dressing method of a small-diameter ball-end grinding wheel based on multi-source information fusion according to an embodiment of the present invention constructed in accordance with a preferred embodiment of the present invention;

[0057] Figure 2 It is a schematic diagram of a ball grinding wheel swing dressing method according to an embodiment of the present invention constructed in accordance with a preferred embodiment of the present invention;

[0058] Figure 3 1 is a schematic diagram of the rough dressing amount calculation of an embodiment of the present invention constructed according to the preferred embodiment of the present invention, wherein (a) is a schematic diagram of the grinding wheel Y-axis movement position 1, (b) is a schematic diagram of the grinding wheel Y-axis movement position 2, and (c) is a schematic diagram of the dressing amount calculation;

[0059] Figure 4 1 is a layered slice contour extraction diagram of an embodiment of the present invention constructed according to a preferred embodiment of the present invention, wherein (a) is a schematic diagram of a grinding wheel swing position 1, (b) is a schematic diagram of a grinding wheel swing position 2, (c) is a schematic diagram of a grinding wheel swing position 3, (d) is a schematic diagram of layered slicing, and (e) is a schematic diagram of a slicing contour;

[0060] Figure 5 is a schematic diagram of surface morphology processing of an embodiment of the present invention constructed according to a preferred embodiment of the present invention, wherein (a) is the optical morphology of the grinding wheel surface Figure 1 , (b) is the corresponding grinding wheel surface morphology binary image, (c) is the grinding wheel surface optical morphology Figure 2 , (d) is the corresponding binary image of the grinding wheel surface morphology;

[0061] Figure 6 It is a structural schematic diagram of in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion according to an embodiment of the present invention constructed in accordance with a preferred embodiment of the present invention;

[0062] Figure 7 It is a schematic diagram of the structure of the rough trimming and fine trimming modules of an embodiment of the present invention constructed according to the preferred embodiment of the present invention, wherein (a) is a schematic diagram of the structure of the rough trimming module, and (b) is a schematic diagram of the structure of the fine trimming module.

[0063] Throughout the drawings, the same reference numerals are used to denote the same elements or structures, wherein:

[0064] 1-dressing module, 2-fine dressing module, 3-visual reconstruction module, 4-signal measurement module, 5-control computer, 6-ball head grinding wheel to be dressed;

[0065] 11-precision leveling mechanism, 12-metal grinding block, 13-grinding block base;

[0066] 21-cup type dressing wheel, 22-precision servo motor, 23-motor base, 24-motor control system;

[0067] 31-in-situ camera, 32-in-situ camera bracket, 33-infrared camera, 34-infrared camera bracket, 35-laser displacement sensor, 36 sensor data acquisition card, 37-sensor bracket;

[0068] 41 - dynamometer, 42 - dynamometer acquisition card, 43 - acoustic emission sensor, 44 - acoustic emission acquisition card, 45 - acceleration sensor, 46 - acceleration acquisition card. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0070] like Figure 6 and 7 As shown, an in-situ precision dressing device for a small-diameter ball-end grinding wheel based on multi-source information fusion is composed of a rough dressing module 1, a fine dressing module 2, a visual reconstruction module 3 and a signal measurement module 4.

[0071] The rough trimming module 1 comprises a precision leveling mechanism 11, a metal grinding block 12 and a grinding block base 13. The metal grinding block 12 is mounted on the grinding block base 13, and the grinding block base 13 is mounted on the precision leveling mechanism 11, so that the surface of the grinding block 12 can be leveled.

[0072] The finishing die 2 includes a cup-shaped finishing wheel 21, a precision servo motor 22, a motor base 23 and a motor control system 24. The cup-shaped finishing wheel 21 is mounted on the precision servo motor 22, the precision servo motor 22 is mounted on the motor base 23, and the control system 24 is connected to the precision servo motor 22 to control the finishing speed and direction.

[0073] The visual reconstruction module 3 includes an in-situ camera 31, an in-situ camera bracket 32, an infrared camera 33, an infrared camera bracket 34, a laser displacement sensor 35, a sensor data acquisition card 36, a sensor bracket 37 and a control computer 5. The in-situ camera 31 is installed on the in-situ camera bracket 32, and can shoot the surface morphology of the small-diameter ball-end grinding wheel 6. The in-situ camera bracket 32 ​​is installed on the machine tool workbench, and the camera posture can be adjusted. The in-situ camera 31 is connected to the control computer 5. The infrared camera 33 is installed on the infrared camera bracket 34, and the temperature during dressing can be monitored. The infrared camera bracket 34 is installed on the machine tool workbench, and the camera posture can be adjusted. The infrared camera 34 is connected to the control computer 5. The laser displacement sensor 35 is installed on the sensor bracket 37, and the surface profile of the small-diameter ball-end grinding wheel 6 can be measured by layering and slicing and a three-dimensional model can be established. The sensor bracket 37 is installed on the machine tool workbench, and the sensor posture can be adjusted. The laser displacement sensor 35 is connected to the sensor acquisition card 36, and the sensor acquisition card 36 is connected to the control computer 5.

[0074] The signal measurement module 4 includes a dynamometer 41, a dynamometer acquisition card 42, an acoustic emission sensor 43, an acoustic emission acquisition card 44, an acceleration sensor 45, an acceleration acquisition card 46 and a control computer 5. The dynamometer 41 is installed on the machine tool workbench and connected to the control computer 5 through the dynamometer acquisition card 42. The acoustic emission sensor 43 and the acceleration sensor 45 are installed on the grinding block base 13 or the motor base 23 and connected to the control computer 5 through the acoustic emission acquisition card 44 and the acceleration acquisition card 46.

[0075] like Figure 1 As shown, a method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion is provided, and the method comprises the following steps:

[0076] S1 Rough trimming

[0077] (1) Figure 2As shown in the figure, the machine tool is controlled to move the small-diameter ball-end grinding wheel to the rough dressing module, and the posture of each device in the visual reconstruction module is adjusted to clearly and effectively collect the surface data of the grinding wheel. The laser displacement sensor is used to perform three-dimensional modeling of the grinding wheel and calculate the radial removal amount b of the rough dressing material. The speed a, swing speed ω, and single feed depth a are set in the CNC system. p , feed times n and other process parameters, the grinding wheel and the metal grinding block are ground against each other to achieve rapid removal of the grinding wheel material. After a single rough dressing is completed, the error is measured using a laser displacement sensor, the dressing amount is recalculated, and the dressing is continued. The above measurement-calculation-dressing process is repeated until the grinding wheel profile accuracy meets the rough dressing requirements.

[0078] The three-dimensional modeling and dressing amount calculation of small diameter ball-end grinding wheels include the following steps:

[0079] (11) Figure 3 As shown, start the machine tool, set the small diameter ball head grinding wheel speed a and axial ( Figure 3 The Y direction in the figure sets the moving speed v, and sets the sampling frequency k of the laser displacement sensor accordingly. The axial grinding wheel is controlled to move, and the laser displacement sensor is started at the same time. The laser displacement sensor is turned off after the grinding wheel finishes moving. Invalid data before and after the grinding wheel moves are eliminated. According to the formula T h = k / (a / 60) calculates the theoretical number of sampling points corresponding to one rotation of the grinding wheel.

[0080] (12) The total sampling data is divided into signal cycles, and the data is divided according to the cycle. The phase angle of each sampling point in the corresponding cycle is calculated to obtain the plane coordinates of the sampling point, and the longitudinal movement interval of each sampling point is calculated according to the formula Δz = v / k to obtain the vertical coordinates of the sampling point. The discrete sampling points are plotted as a three-dimensional surface.

[0081] (13) Calculate the distance r from each sampling point to the coordinate origin i , and find the maximum distance r max and the minimum distance r min , then the total deviation is b'=r max -r min In order to leave a margin for fine trimming, the rough trimming coefficient c is set to <1, and the rough trimming amount is b = c·b'.

[0082] (2) During the rough dressing process, an infrared camera is used to monitor the temperature of the grinding wheel. If the maximum temperature exceeds the set threshold temperature T and lasts for more than the threshold time t', the dressing is interrupted for cooling. After the rough dressing is completed, the machine tool is controlled to move the grinding wheel to the fine dressing module.

[0083] S2 Fine Finishing

[0084] (1) Figure 4As shown in the figure, adjusting the posture of each device in the visual reconstruction module can clearly and effectively collect the surface data of the grinding wheel. The initial shooting position of the visual reconstruction module is aligned with the ball equator of the small-diameter ball-end grinding wheel (latitude θ = 0). The equator is the outer contour of the ball-end grinding wheel to be dressed and the plane perpendicular to the grinding wheel rod intersects with the ball-end grinding wheel to be dressed. The circumferential swing feed angle Δθ of the small-diameter ball-end grinding wheel is set each time, and the grinding wheel contour is divided into n = 90° / Δθ slices. Each slice corresponds to the ball head latitude θ i The ball head grinding wheel is controlled to rotate at a speed a, and the laser displacement sensor is used to measure the ball head θ i The normal contour at the slice is extracted, the surface morphology at the same slice position is photographed using an in-situ camera, and the images at different phase angles are spliced ​​to obtain the complete ball head θ i Phase angle expansion diagram of the morphology at the slice.

[0085] (2) Acoustic emission signals, force signals and acceleration signals are collected during the finishing process in the fine finishing experiment. The contour information, surface morphology, acoustic emission signal, force signal and acceleration signal characteristic parameters of each slice layer are extracted and calculated to obtain the contour characteristic parameters, surface characteristic parameters and vibration characteristic parameters of the layered slice.

[0086] The calculation of characteristic parameters of each slice of the grinding wheel ball head includes the following methods: the contour fluctuation parameters measured by the laser displacement sensor are calculated as follows:

[0087] Arithmetic mean roughness:

[0088] RMS roughness:

[0089] Maximum height of the profile: Rz = max(z i )-min(z i )

[0090] Average spacing of profile peaks: Rs = average(T i )

[0091] Skewness of the profile:

[0092] Among them, z i is the height of each sampling point, z mean is the average height of the sampling points in a single cycle, T i is the duration of each cycle detected by the cycle segmentation algorithm, and n is the total number of cycle sampling points.

[0093] The images obtained by stitching the original camera are gray-scaled, Gaussian filtered, and adaptively binarized. The number of white connected domains n in the image is calculated using the seed algorithm. w And each white connected area Si , and calculate the following connected domain related parameters:

[0094] Standard Deviation:

[0095] Maximum connected domain area: S max =max(S i )

[0096] Connected domain skewness:

[0097] Among them, S mean is the average area of ​​the white connected domain, (x c ,y c ) are the centroid coordinates of all white connected domains, (x m ,y m ) are the coordinates of the image center.

[0098] The characteristic parameters of acoustic emission, force signal and acceleration signal are calculated, including the following methods: extracting energy E from the acoustic emission signal, performing spectrum analysis, and extracting the main frequency f of the acoustic emission. AE_main Calculate the average value F of the tangential and normal force signals t_mean 、F n_mean And the fluctuation range F t_p 、F n_p Calculate the average value A of the tangential and normal acceleration signals t_mean , A n_mean and the fluctuation range A t_p , A n_p . Perform spectrum analysis and extract the acceleration main frequency f t_main and f n_main .

[0099] Establish a multi-source information fusion prediction model for the multivariate regression method of grinding wheel slicing dressing time:

[0100] t i =∑α j ·x j +∑β j ·y j +ε

[0101] Among them, t i is the prediction time for each slice trimming, x j is each feature parameter, α j is the multiple regression coefficient of each feature parameter, y j is the processing parameter, β j is the multiple regression coefficient of each processing parameter, and ε is the error term.

[0102] (3) The regression model is used to predict the dressing time of each slice in the fine dressing process, and the grinding wheel is controlled to swing to the corresponding ball head θ i At the slicing point, the grinding wheel is dressed by the fine dressing module. Repeat the above measurement-prediction-dressing steps until the grinding wheel accuracy and morphology meet the requirements.

[0103] The signal period segmentation includes the following steps: the signal point sequence obtained by acquisition is s i , the total length of the signal is n, and the detection window width is w. w should be less than the minimum cycle length. Perform a window sliding scan on the signal sequence, and the detection window center point is from s w / 2 Starting from s n-w / 2 End. Determine the window center position signal s i Is the highest point of all signals in the current detection window? If not, move the window horizontally to detect the next position signal s i +1 is the highest point. If the signal at this position is the highest point, then record the signal point subscript x i , as the periodic boundary. If there is a periodic theoretical calculated value T h , then increase the window coordinates by T h / 2 , and then perform step scan. If the signal period changes irregularly, continue to step scan the detection window directly. After the scan is completed, the periodic boundary point sequence x i , to realize the period division of the sampling signal.

[0104] like Figure 5 As shown, the seed algorithm includes the following steps: Scan the image line by line starting from the position (0, 0) of the image. Let the binary image matrix be img, and let the scanning position coordinates be (x, y). If the img (, y) pixel is white and the position has not been visited in the previous scan, create a seed queue, add (x, y) to the head of the queue as an extended seed, and set the queue pointer to 0 to point to the seed coordinates. Scan the four connected domain pixels above, below, left and right of the coordinates of the position pointed to by the queue pointer. If there are white pixels that have not been scanned, add the pixel coordinates to the end of the queue and mark the pixel as scanned. After the scan of the current seed coordinates is completed, increase the queue pointer by one. Repeat the above steps until the queue is empty and a single complete white connected domain information is obtained. Repeat the above steps until the entire image is scanned and the information of all white connected domains is obtained.

[0105] The present invention will be further described below with reference to specific embodiments.

[0106] The present invention provides an on-site precision dressing method for a small-diameter ball-end grinding wheel based on multi-source information fusion. Specifically, the dressing of a small-diameter diamond ball-end grinding wheel with a diameter of 3 mm is taken as an example to describe the dressing method in detail.

[0107] In this embodiment, the rough trimming includes the following steps:

[0108] Control the machine tool to move the small-diameter ball-end grinding wheel to the rough dressing module, adjust the position of each device in the visual reconstruction module, and collect the surface data of the grinding wheel clearly and effectively. Use the laser displacement sensor to build a three-dimensional model of the grinding wheel and calculate the radial removal amount b of the rough dressing material. Write a program in the CNC system to control the grinding wheel to swing feed motion, set the grinding wheel speed a = 5000rpm, the swing speed ω = 1.5° / s, and the single feed depth a p =2μm, feeding times n=b / a p . The grinding wheel is ground against the metal grinding block to quickly remove the grinding wheel material. After a single rough dressing is completed, a laser displacement sensor is used to measure the error, the dressing amount is recalculated, and the dressing is continued. Repeat the above measurement-calculation-dressing process until the grinding wheel profile accuracy meets the rough dressing requirements. In this embodiment, the rough dressing requirement is that the spherical runout is less than 10μm. During the rough dressing process, an infrared camera is used to monitor the temperature of the grinding wheel. If the maximum temperature exceeds the set threshold temperature T=80° and lasts for more than the threshold time t'=5s, the dressing is interrupted for cooling.

[0109] In this embodiment, the grinding wheel three-dimensional modeling and dressing amount calculation include the following steps:

[0110] Set the small diameter ball head grinding wheel speed a = 1200rpm and the axial moving speed v = 0.5mm / s, and set the laser displacement sensor sampling frequency k = 100kHz, and the sampling time is 4s. Control the grinding wheel to move axially and start the laser displacement sensor at the same time. Turn off the laser displacement sensor after the grinding wheel ends its movement. Eliminate invalid data before and after the grinding wheel moves. According to the formula T h = k / (a / 60) to calculate the theoretical number of sampling points corresponding to one rotation of the grinding wheel, which is 5000. Divide the total sampling data into signal cycles, divide the data according to the cycles, and calculate the phase angle of each sampling point in the corresponding cycle to obtain the plane coordinates of the sampling points. According to the formula Δz=v / k, the longitudinal movement interval of each sampling point is calculated to be 5nm, and the vertical coordinates of each sampling point are obtained. Draw the discrete sampling points as a three-dimensional surface. Calculate the distance r from each sampling point to the origin of the coordinate system. i , find the maximum distance r max = 20 μm and minimum distance r min =5μm, the total radial deviation is b'=r max -r min =15μm. In order to leave a margin for fine trimming, the rough trimming coefficient c is set to 0.9, and the rough trimming radial removal amount is b=c·b'=13.5μm.

[0111] In this embodiment, signal period segmentation includes the following steps:

[0112] The signal point sequence collected is s i , the total length of the signal is 400000. Set the detection window width to w = 2000, and perform window sliding scanning on the signal sequence. The detection window center point is from s w / 2 =1000, to s n-w / 2 =399000. Determine the window center position signal s i Is the highest point of all signals in the current detection window? If not, move the window horizontally to detect the next position signal s i +1 is the highest point. If the signal at this position is the highest point, then record the signal point subscript x i , as the periodic boundary, and increase the window coordinates by T h / 2=2500, and then perform step scan. If the signal period changes irregularly, continue to detect window step scan directly. After the scan is completed, the periodic boundary point sequence x i , to realize the period division of the sampling signal.

[0113] In this embodiment, the fine finishing includes the following steps:

[0114] Move the grinding wheel to the finishing module and adjust the position of each device in the visual reconstruction module to clearly and effectively collect the surface data of the grinding wheel. Align the initial shooting position of the visual reconstruction module to the ball equator of the small-diameter ball-end grinding wheel (latitude θ = 0), set the circumferential swing feed angle of the small-diameter ball-end grinding wheel to Δθ = 10° each time, and divide the grinding wheel contour into n = 90° / Δθ = 9 slices, and each slice corresponds to a latitude of θ i The ball-end grinding wheel is controlled to rotate at a speed of 5000 rpm, and the laser displacement sensor is used to measure the ball end θ i The normal contour at the slice is extracted, the surface morphology at the same slice position is photographed using an in-situ camera, and the images at different phase angles are spliced ​​to obtain the complete ball head θ i The expanded image of the morphology at the slice. In the fine dressing experiment, the acoustic emission signal, force signal and acceleration signal of the dressing process are collected. The contour information, surface morphology, acoustic emission signal, force signal and acceleration signal characteristic parameters of each slice layer are extracted and calculated to obtain the contour characteristic parameters, surface characteristic parameters and vibration characteristic parameters of the layered slice. A multi-source information fusion prediction model of the multivariate regression method for the dressing time of the grinding wheel slice is established:

[0115] t i =∑α j ·x j +∑β j ·y j +ε

[0116] Among them, ti is the prediction time for each slice trimming, x j is each feature parameter, α j is the multiple regression coefficient of each feature parameter. j is the processing parameter, β j is the multivariate regression coefficient of each processing parameter. ε is the error term. The trimming time t of each slice in the fine trimming is predicted by the regression model. i , control the grinding wheel to swing to the corresponding ball head θ i At the slicing point, the grinding wheel is trimmed by the fine trimming module. Repeat the above measurement-prediction-dressing steps until the grinding wheel accuracy and morphology meet the requirements. In this embodiment, the grinding wheel accuracy and morphology requirements are Ra<1μm, Rz<2μm, σ s <100, r c <100.

[0117] In this embodiment, the seed algorithm includes the following steps:

[0118] Scan the image line by line starting from the image position (0, 0). Let the binary image matrix be img, and let the scanning position coordinates be (x, y). If the pixel of img (x, y) is white, and this position has not been visited in the previous scan, create a seed queue, add (x, y) to the head of the queue as an extended seed, and set the queue pointer to 0, pointing to the seed coordinates. Scan the four connected domain pixels above, below, left, and right of the coordinates of the position pointed to by the queue pointer. If there are white pixels that have not been scanned, add the pixel coordinates to the end of the queue and mark the pixel as scanned. After scanning the current seed coordinates, increase the queue pointer by one. Repeat the above steps until the queue is empty and a single complete white connected domain information is obtained. Repeat the above steps until the entire image is scanned and the information of all white connected domains is obtained.

[0119] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion, characterized in that: The method includes: S1 Rough trimming (a) obtaining the coordinates of each point on the ball-end grinding wheel to be dressed, calculating the distance from each point to the center of the ball, obtaining the maximum distance and the minimum distance from the point on the surface of the grinding wheel to be dressed to the center of the ball, setting a rough dressing amount according to the maximum distance and the minimum distance, and dressing the ball-end grinding wheel to be dressed according to the rough dressing amount; (b) repeating step (a) until the difference between the maximum distance and the minimum distance of the ball-end grinding wheel to be dressed meets the preset requirement for rough dressing; S2 Fine Finishing The equator of the ball-end grinding wheel to be dressed is set, and the slice contours of the ball-end grinding wheel to be dressed at different latitudes are obtained; the characteristic parameters of each slice layer contour at different latitudes are calculated, and the dressing time of each slice layer is calculated using the characteristic parameters and the preset processing parameters of the ball-end grinding wheel to be dressed, and the ball-end grinding wheel to be dressed is finely dressed according to the dressing time and the preset processing parameters of the ball-end grinding wheel to be dressed, until the characteristic parameters of each slice layer of the ball-end grinding wheel to be dressed meet the preset requirements for fine dressing.

2. The in-situ precision dressing method for a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 1, characterized in that: In step S1, the coordinates of each point on the ball-end grinding wheel to be dressed are obtained according to the following steps: The ball-end grinding wheel to be dressed rotates around the Y axis and moves up and down along the Y axis at the same time, and the ball-end grinding wheel to be dressed is scanned to obtain sampling points on each slice layer perpendicular to the Y direction, and the coordinates of each sampling point in the XOZ plane are obtained; According to the speed and sampling frequency of the ball-end grinding wheel to be dressed moving up and down along the Y axis, the coordinates of each sampling point direction are calculated to obtain the three-dimensional coordinates of each sampling point, that is, the coordinates of each point on the ball-end grinding wheel to be dressed.

3. A method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 1 or 2, characterized in that: The rough trimming amount is set according to the maximum and minimum distance difference according to the following formula: b=c·(r max -r min ) Where b is the rough trimming amount, c is the rough trimming coefficient, and r max is the maximum distance from the point on the grinding wheel surface to be dressed to the center of the sphere, r min It is the minimum distance from the point on the grinding wheel surface to be dressed to the center of the sphere.

4. The in-situ precision dressing method for a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 1 or 2, characterized in that: In step S2, the equator is the outer contour of the ball-end grinding wheel to be dressed, where a plane passing through the center of the ball-end grinding wheel to be dressed and perpendicular to the grinding wheel rod intersects with the ball-end grinding wheel to be dressed.

5. The in-situ precision dressing method for a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 4, characterized in that: The slicing profile of the ball-end grinding wheel to be dressed at different latitudes is obtained according to the following steps: the ball-end grinding wheel to be dressed rotates while swinging in the YOZ plane, scanning and photographing the surface of the ball-end grinding wheel to be dressed, and obtaining the slicing profiles at different latitudes after image processing.

6. The method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 5, characterized in that: The characteristic parameters include contour characteristic parameters, surface characteristic parameters and vibration characteristic parameters.

7. The method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 6, characterized in that: The characteristic parameters are calculated as follows: The calculation formula of the contour feature parameters is as follows: Arithmetic mean roughness: RMS roughness: Maximum height of the profile: Rz = max(z i )-min(z i ), Average spacing of profile peaks: Rs = average(T i ), Skewness of the profile: The calculation formula of the surface characteristic parameters is as follows: Standard Deviation: Maximum connected domain area: S max =max(S i ), Connected domain skewness: Among them, z i is the height of each sampling point, z mean is the average height of the sampling points in a single cycle, T i is the duration of each cycle detected by the cycle segmentation algorithm, n is the total number of cycle sampling points, S mean is the average area of ​​the white connected domain, x c and c are the horizontal and vertical coordinates of the centroid of all white connected domains, x m and m are the horizontal and vertical coordinates of the image center, n w is the number of white connected domains in a single slice layer, S i is the area of ​​the i-th white connected region in a single slice layer; The vibration characteristic parameters include: the main frequency of acoustic emission, the average value and fluctuation amplitude of the tangential and normal directions of the force signal, and the average value, fluctuation amplitude and main frequency of the tangential and normal directions of the acceleration signal.

8. The method for in-situ precision dressing of a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 7, characterized in that: The total number of white connected regions in the single slice layer and the area of ​​each white connected region are calculated using a seed algorithm, and the steps of the seed algorithm are as follows: (1) Acquire a surface topography image of a single slice layer, perform image processing on the image to obtain a binary image; search for white connected areas in the binary image in the following manner: (2) Determine whether the current pixel i is a black point or a white point; If it is a black dot, i=i+1, determine whether i is less than n, if so, return to step (2); otherwise, end; If it is a white spot, follow the steps below; (a)n w =n w +1, establish a seed queue, and put the current pixel point i as the seed pixel point into the seed queue, (b) searching for four-connected pixels adjacent to the seed pixel and determining whether the adjacent pixels are white or black; (c) If the adjacent pixel is a white point, add the adjacent pixel to the seed queue and mark the pixel i as visited. i =s i +1; Place the white adjacent node in the seed queue as the seed pixel and return to step (b), i=i+1, and determine whether i is less than n. If so, return to step (2), otherwise, end; If the adjacent node is a black dot or a visited node, i=i+1, and determine whether i is less than n. If so, return to step (2), otherwise, end.

9. The in-situ precision dressing method for a small-diameter ball-end grinding wheel based on multi-source information fusion as claimed in claim 1 or 8, characterized in that: The dressing time of each slice layer is calculated by using the characteristic parameters and the preset processing parameters of the ball-end grinding wheel to be dressed, and is calculated according to the following formula: t i =∑α j ·x j +∑β j ·y j +ε Among them, t i is the prediction time for each slice trimming, x j is each feature parameter, α j is the multiple regression coefficient of each feature parameter, y j is the processing parameter, β j is the multiple regression coefficient of each processing parameter, and ε is the error term.

10. A system for dressing using the in-situ precision dressing method for small-diameter ball-end grinding wheels based on multi-source information fusion as described in any one of claims 1 to 9, characterized in that: The system includes a rough trimming module, a fine trimming module, a visual reconstruction module and a signal measurement module, wherein: The visual reconstruction module is used to obtain the coordinates of each point on the ball-end grinding wheel to be dressed during the rough dressing process and calculate the rough dressing amount, and to obtain the slice profile of the ball-end grinding wheel to be dressed at different latitudes during the fine dressing process; The signal measurement module is used to calculate the characteristic parameters of each slice layer profile during the fine trimming process, and calculate the trimming time of each slice layer using the characteristic parameters; The rough dressing module is used to perform rough dressing on the ball-end grinding wheel to be dressed according to the dressing amount calculated by the visual reconstruction module; The fine dressing module is used to perform fine dressing on the ball-end grinding wheel to be dressed according to the dressing time calculated by the signal measuring module.

Images