A system and method for constructing a dataset of abrasive wear data

Abstract

The application provides a kind of construction system of abrasive tool wear dataset, comprising: numerical control grinding machine tool module (100) for performing grinding process;Processing signal acquisition module (200) for acquiring vibration signal, spindle current signal, grinding force signal and acoustic emission signal and other multi-source signals in the grinding process;Abrasive tool wear measurement and acquisition module (300) for acquiring abrasive tool surface topography and measuring the distance value measurement matrix of abrasive tool periphery and spectral confocal displacement sensor;Data processing and control module (400) for presetting machine tool coordinates and angle list, receiving and storing sensor data, running grinding machine tool digital twin model, realizing feedback control of machine tool and sensor and calculating abrasive tool wear amount.The application can realize online acquisition of abrasive tool wear multi-source feature data, in-machine acquisition of wear state and wear amount label data, and solve the problems of low abrasive tool wear dataset construction efficiency, low dataset quality caused by non-processing data pollution and the like.

Description

Technical Field

[0001] This invention belongs to the field of grinding technology, specifically relating to a system and method for constructing a grinding wheel wear dataset. Background Technology

[0002] Grinding is a crucial process in manufacturing, and the wear state of grinding tools directly affects workpiece quality and production efficiency. With increasing demands for machining accuracy and efficiency, real-time monitoring and prediction of grinding tool wear has become increasingly important. Timely understanding of grinding tool wear helps optimize machining parameters and prevent workpiece scrap and machine tool damage due to grinding tool failure. By collecting multi-source signals during the grinding process (such as vibration, spindle current, grinding force, and acoustic emission), a grinding tool wear dataset can be constructed. This provides crucial technical support for wear prediction and remaining life assessment using methods such as deep learning. Therefore, establishing a high-quality grinding tool wear dataset is of great significance for intelligent manufacturing.

[0003] However, existing technologies for constructing grinding wheel wear datasets are still imperfect, mainly due to the following problems: First, traditional methods often involve disassembling the machine after shutdown to detect grinding wheel wear, which is not only inefficient but also introduces clamping errors due to frequent disassembly and assembly, leading to reduced machining accuracy. Second, when collecting data from multiple sensors over a long period, if the actual grinding stage is not distinguished, a large number of invalid signals from idling or non-processing states will contaminate the dataset, generating non-processing data noise and reducing data quality. Third, grinding wheel wear usually needs to be measured offline by dedicated equipment after machining, which not only increases the workload of data collection but also easily leads to a lack of comparability and consistency in wear amount label data due to inconsistent measurement positions or repeated collections. Therefore, there is currently no publicly available grinding wheel wear dataset that includes grinding wheel wear characteristic data, grinding wheel wear state label data, and wear amount label data.

[0004] To address the aforementioned shortcomings, there is an urgent need to develop a technology that can improve the efficiency of constructing grinding wheel wear datasets while ensuring data quality. This invention aims to provide a system and method for constructing grinding wheel wear datasets. By combining technologies such as spectral confocal sensing measurement, digital twin model monitoring and control, and multi-source signal synchronous acquisition, it achieves efficient online acquisition of grinding wheel wear characteristic data and accurate filtering of non-processing data. Furthermore, it acquires grinding wheel wear state label data and wear amount label data on-machine, thereby constructing a high-quality grinding wheel wear dataset. Summary of the Invention

[0005] To address the problems of low efficiency in constructing grinding wheel wear datasets and low dataset quality caused by non-processing data contamination during the dataset construction process, this invention provides a system and method for constructing grinding wheel wear datasets by utilizing technologies such as spectral confocalization and digital twins.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] This invention provides a system for constructing a grinding wheel wear dataset, characterized in that it includes:

[0008] The CNC grinding machine module is used to perform the grinding process;

[0009] The machining signal acquisition module is used to acquire multi-source signals such as vibration signals, spindle current signals, grinding force signals and acoustic emission signals during the grinding process;

[0010] The grinding wheel wear measurement and acquisition module is used to acquire the surface morphology of the grinding wheel and measure the distance values ​​between the grinding wheel periphery and the spectral confocal displacement sensor to obtain the measurement matrix.

[0011] The data processing and control module is used to preset the machine tool coordinates and rotation angle list, receive and store sensor data, run the digital twin model of the grinding machine tool, realize feedback control between the machine tool and sensors, and calculate the wear of the grinding wheel.

[0012] The abrasive wear measurement and acquisition module includes an industrial microscope, a spectral confocal displacement sensor, an industrial microscope fixture, and a spectral confocal displacement sensor fixture;

[0013] The industrial microscope is fixed to the CNC grinding machine tool by an industrial microscope fixture to observe the surface morphology of the grinding wheel and determine whether the grinding wheel is in the initial wear, normal wear or severe wear state, which serves as the grinding wheel wear state label data in the grinding wheel wear dataset.

[0014] The spectral confocal displacement sensor is fixed to a CNC grinding machine tool by a spectral confocal displacement sensor fixture to measure the distance between itself and the periphery of the grinding tool. By using a preset machine tool coordinate and rotation angle list, it can measure the distance value of each measurement point in the list and form a distance value measurement matrix before and after grinding tool wear. The mean of the difference between the distance value measurement matrices before and after grinding tool wear can be calculated as the grinding tool wear amount label data in the grinding tool wear dataset.

[0015] Furthermore, the CNC grinding machine tool module includes a CNC grinding machine tool, a machine tool spindle box, a grinding wheel, a workpiece, and a machine tool worktable, and has a spindle rotation angle recording function, which can realize spindle rotation angle control.

[0016] Furthermore, the machining signal acquisition module includes an accelerometer, a current sensor, a force gauge, and an acoustic emission sensor. The accelerometer is arranged on the machine tool spindle box, the current sensor is arranged on the three-phase AC cable of the machine tool spindle in the machine tool spindle box, the workpiece is fixed on the force gauge, and the force gauge and the acoustic emission sensor are arranged on the machine tool worktable. The vibration signal, spindle current signal, grinding force signal, and acoustic emission signal collected by the above sensors together constitute the grinding tool wear characteristic data in the grinding tool wear dataset.

[0017] Furthermore, the abrasive wear measurement and acquisition module includes an industrial microscope, a spectral confocal displacement sensor, an industrial microscope fixture, and a spectral confocal displacement sensor fixture;

[0018] The industrial microscope is fixed to the CNC grinding machine tool by an industrial microscope fixture to observe the surface morphology of the grinding wheel and determine whether the grinding wheel is in the initial wear, normal wear or severe wear state, which serves as the grinding wheel wear state label data in the grinding wheel wear dataset.

[0019] The spectral confocal displacement sensor is fixed to a CNC grinding machine tool by a spectral confocal displacement sensor fixture to measure the distance between itself and the periphery of the grinding tool. By using a preset machine tool coordinate and rotation angle list, the distance value of each measurement point in the list can be measured to form a distance value measurement matrix before and after grinding tool wear. The mean of the difference between the distance value measurement matrices before and after grinding tool wear can be calculated as the grinding tool wear amount label data in the grinding tool wear dataset.

[0020] Furthermore, the data processing and control module includes a controller, a computer, and a digital twin model of the grinding machine tool. The computer can receive, process, and store data from the CNC grinding machine tool module, the machining signal acquisition module, and the grinding wheel wear measurement and acquisition module.

[0021] The controller can receive signals from the computer and control the movement of the CNC grinding machine tool, as well as control the acquisition and stopping of the accelerometer, current sensor, force gauge, acoustic emission sensor, industrial microscope, and spectral confocal displacement sensor.

[0022] The digital twin model of the grinding machine tool is built in a computer and can realize the virtual-real mapping of the operation of the CNC grinding machine tool. It can reflect the processing situation in real time and simulate the grinding process that will occur based on the machine tool.

[0023] This invention also provides a method for constructing a grinding wheel wear dataset, characterized by comprising:

[0024] S1: Complete initialization operations such as sensor placement, machine tool power supply, sensor power supply and calibration, and mold and workpiece installation;

[0025] S2: Establish communication between the machine tool, sensors, computer and controller; build a digital twin model of the grinding machine tool in the computer; and preset the machine tool coordinates and rotation angle list.

[0026] S3: The computer controls the machine tool to move the grinding wheel to the wear measurement position. The computer collects the distance measurement data matrix corresponding to the unworn grinding wheel, which consists of machine tool coordinates, rotation angle, and distance values ​​between the spectral confocal displacement sensor and the surrounding area of ​​the grinding wheel. The data is then transmitted to the computer.

[0027] S4: Filter the collected distance measurement data matrix in the computer;

[0028] S5: Start grinding. When the grinding wheel and the workpiece are about to come into contact in the machine tool digital twin model, the machining signal acquisition sensor is triggered to collect signals. When the grinding wheel and the workpiece are about to separate, the signal acquisition stops and the signals collected in this process are transmitted to the computer as grinding wheel wear characteristic data.

[0029] S6: When the material volume removed by grinding reaches the threshold after several passes, the machine tool is controlled by computer to move the grinding wheel to the measurement position. The surface morphology of the grinding wheel is observed by an industrial microscope to determine whether the grinding wheel is in the initial wear, normal wear or severe wear state, and this is used as the grinding wheel wear state label data.

[0030] S7: The computer controls the machine tool to move the grinding wheel to the wear measurement position. The distance measurement data matrix after the grinding wheel wears is collected by the spectral confocal displacement sensor, transmitted to the computer and filtered. Then the mean of the difference between the distance measurement data matrix and the distance measurement data matrix corresponding to the grinding wheel without wear is calculated as the grinding wheel wear amount label data.

[0031] S8: Repeat S5 to S7 to construct a grinding wheel wear dataset containing a single grinding wheel;

[0032] S9: Change the grinding wheel and workpiece, repeat S3 to S8, and build a grinding wheel wear dataset containing multiple grinding wheels.

[0033] Furthermore, in step S3, the spectral confocal displacement sensor is fixed, and the machine tool carrying the grinding wheel moves in a manner of first longitudinal translation and then rotation. The cylindrical surface around the grinding wheel can be unfolded into an imaginary plane. The measurement is performed using a preset longitudinal measurement interval Δy and an angular interval. This allows us to determine the coordinates and rotation list of the measurement data matrix, where the lateral measurement interval Δx is... The arc length is defined by the radius of the grinding wheel, which is the distance l between the spectral confocal displacement sensor and the center of the grinding wheel, and the difference ld between the distance d measured by the sensor and the periphery of the grinding wheel.

[0034] In a longitudinal measurement of abrasive wear, the starting point for longitudinal measurement is a preset point where the upper boundary of the abrasive portion of the abrasive part can be measured by a spectral confocal displacement sensor. The abrasive is moved upwards at intervals of Δy in a preset coordinate list, only when the spectral confocal displacement sensor measures a valid value. This continues until the preset lower boundary of the abrasive portion of the abrasive part is measured, i.e., the measurement termination point. Once a valid measurement value is obtained, a longitudinal measurement of abrasive wear is completed. The abrasive is then controlled to rotate. Then move downwards until the spectral confocal displacement sensor measures the upper boundary of the preset abrasive part, and the second longitudinal measurement can begin, that is, the data measurement of the second column of the preset list can begin on the plane around the imaginary abrasive. Repeat this operation to obtain the data of all measurement points in the preset list.

[0035] During this process, the measurement interval is set to be less than the machine tool's movement interval. Therefore, the computer will repeatedly check whether the machine tool's current coordinates and rotation angles are elements in the preset coordinates and rotation angles list. If the machine tool's movement interval is too long, it will cause the computer to check and compare too many times, and an error message will be thrown. The computer controls the spectral confocal displacement sensor to perform a measurement process only when the machine tool reaches the measurement point of the preset coordinates and rotation angles. Then, the measurement will be repeated when the machine tool moves to the next preset measurement point, so as to avoid the problem of repeated measurement of the same measurement point.

[0036] Furthermore, step S3 can be further refined into the following steps:

[0037] S31: Initialize parameters, preset machine tool coordinates and rotation angle list, and assign measurement sequence numbers;

[0038] S32: Determine if the error count has reached the upper limit. If not, obtain the current machine tool coordinates and rotation angle, and proceed to step S33. If yes, report an error and exit the measurement program.

[0039] S33: Determine whether the obtained machine tool coordinates and rotation angle are the same as the coordinates and rotation angle to be measured corresponding to the current measurement sequence number in the preset list. If not, the error count value is accumulated and the process returns to S32. If yes, the computer will obtain the measurement distance value of the spectral confocal displacement sensor, add the machine tool coordinates, rotation angle and sensor measurement value to the measurement data matrix, update the measurement sequence number to the next measurement point, accumulate the valid coordinate count value, reset the error count value to zero, and proceed to S34.

[0040] S34: Determine if the valid coordinate count has reached the upper limit. If not, wait for the machine tool to proceed with the next movement before taking the measurement, and return to S32. If yes, output the measurement matrix to the computer.

[0041] In step S5, when the digital twin model of the grinding machine tool detects that the grinding wheel and the workpiece are about to come into contact, it triggers the acceleration sensor, current sensor, force gauge and acoustic emission sensor to collect vibration, spindle current, grinding force and acoustic emission signals during the processing. When the model detects that the grinding wheel and the workpiece are about to separate, it stops collecting signals and transmits the signals collected in the process to the computer as grinding wheel wear characteristic data.

[0042] In step S7, the method for obtaining the distance value measurement data matrix is ​​the same as in step S3;

[0043] In this process, although the transverse measurement interval Δx will decrease as the wear amount Δd of the grinding wheel increases, due to Δy and The number of longitudinal and transverse measurement points N is preset and remains unchanged. y With N x All remain unchanged. Even as the wear of the grinding wheel increases, this method can still accurately obtain the corresponding measurement points before and after the wear of the grinding wheel by dynamically adjusting the transverse measurement interval, and ensure that the total number of measurement points remains unchanged, thereby establishing an accurate mapping relationship of the same point before and after wear.

[0044] Compared with the prior art, the present invention has the following advantages and significant effects:

[0045] (1) By using acceleration sensors, current sensors, force gauges and acoustic emission sensors arranged on the machine tool, the wear characteristics data of grinding tools such as vibration, spindle current, grinding force and acoustic emission signals during grinding are collected online. Furthermore, by using industrial microscopes and spectral confocal displacement sensors fixed on the machine tool through fixtures, the wear status label data and wear amount label data of grinding tools are acquired on-machine, thereby improving the efficiency of dataset construction.

[0046] (2) By leveraging the ability of the digital twin model of the grinding machine tool to map the virtual and real, the timing of the grinding wheel and the workpiece about to come into contact and about to leave contact can be accurately captured. This can be used as the start and end of the grinding process. The acquisition and stopping time of the acceleration sensor, current sensor, force gauge and acoustic emission sensor can be precisely controlled by the computer and controller. In this way, the signal data during the processing can be accurately collected to establish a high-quality grinding wheel wear dataset, avoiding the contamination of the dataset by non-processing data.

[0047] (3) A method for constructing a distance measurement data matrix based on a preset list of machine tool coordinates and rotation angles is proposed. On the one hand, the computer controls the spectral confocal displacement sensor to perform a measurement process only when the machine tool reaches the next preset coordinate and rotation angle measurement point. Then, the measurement will be repeated when the machine tool moves to the next preset measurement point again, thus avoiding the problem of repeated measurement of the same measurement point. On the other hand, even if the wear of the grinding wheel increases, the method can still accurately obtain the corresponding measurement points before and after the wear of the grinding wheel through the dynamic adjustment of the lateral measurement interval, and ensure that the total number of measurement points remains unchanged, thereby establishing an accurate mapping relationship of the same point before and after wear, and further ensuring the high quality of the grinding wheel wear label data.

[0048] Therefore, the grinding wheel wear dataset construction system and method provided by the present invention can realize the online acquisition of multi-source feature data of grinding wheel wear, and the on-machine acquisition of grinding wheel wear state label data and grinding wheel wear amount label data, while having the advantages of high dataset construction efficiency and high dataset quality. Attached Figure Description

[0049] Appendix Figure 1 This is a schematic diagram of the modules in the grinding wheel wear dataset construction system.

[0050] Appendix Figure 2 This is a schematic diagram showing the components and relationships of each module in the grinding wheel wear dataset construction system.

[0051] Appendix Figure 3 This is a flowchart of the method for constructing a grinding wheel wear dataset.

[0052] Appendix Figure 4 This is a schematic diagram of the longitudinal measurement process of the wear of the grinding wheel and the surrounding plane of the grinding wheel and the measurement point.

[0053] Appendix Figure 5 This is a schematic diagram showing the transition to the next longitudinal measurement after the completion of a longitudinal measurement of the wear of the grinding wheel, as well as the unfolding of the grinding wheel periphery into a plane and the measurement point.

[0054] Appendix Figure 6 This is a schematic diagram showing the changes in the corresponding measurement points when the spindle rotates at a certain angle before and after the grinding wheel wears out, as well as the changes in the complete measurement points on the surrounding plane of the grinding wheel.

[0055] Appendix Figure 7 It is a flowchart for constructing a distance measurement data matrix consisting of preset machine tool coordinates, rotation angles, spectral confocal displacement sensors, and distance values ​​around the mold.

[0056] Among them: 100—CNC grinding machine tool module, 101—CNC grinding machine tool, 102—machine tool spindle box, 103—grinding wheel, 104—workpiece, 105—machine tool worktable; 200—machining signal acquisition module, 201—accelerometer, 202—current sensor, 203—force gauge, 204—acoustic emission sensor; 300—grinding wheel wear measurement and acquisition module, 301—industrial microscope, 302—spectral confocal displacement sensor, 303—industrial microscope fixture, 304—spectral confocal displacement sensor fixture; 400—data processing and control module, 401—controller, 402—computer, 403—digital twin model of grinding machine tool. Detailed Implementation

[0057] To facilitate understanding by those skilled in the art, the following description is provided in conjunction with the appendix. Figure 1 -8 provides a detailed description of the present invention.

[0058] This embodiment provides a system for constructing a grinding wheel wear dataset, including:

[0059] The CNC grinding machine tool module 100 includes a CNC grinding machine tool 101, a machine tool spindle box 102, a grinding wheel 103, a workpiece 104, and a machine tool worktable 105. It has a spindle rotation angle recording function and can realize spindle rotation angle control.

[0060] The machining signal acquisition module 200 includes an accelerometer 201, a current sensor 202, a force gauge 203, and an acoustic emission sensor 204. The accelerometer 201 is mounted on the machine tool spindle box 102 to measure vibration signals during machining. The current sensor 202 is mounted on the three-phase AC cable of the machine tool spindle in the machine tool spindle box 102 to measure the spindle current signal during machining. The workpiece 104 is fixed on the force gauge 203. The force gauge 203 and the acoustic emission sensor 204 are mounted on the machine tool worktable 105 to measure the grinding force signal and acoustic emission signal during machining, respectively. The vibration signal, spindle current signal, grinding force signal, and acoustic emission signal together constitute the grinding tool wear characteristic data in the grinding tool wear dataset.

[0061] The grinding wheel wear measurement and acquisition module 300 includes an industrial microscope 301, a spectral confocal displacement sensor 302, an industrial microscope fixture 303, and a spectral confocal displacement sensor fixture 304. The industrial microscope 301 is fixed to the CNC grinding machine tool 101 via the industrial microscope fixture 303 to observe the surface morphology of the grinding wheel 103 and determine whether the grinding wheel 103 is in an initial wear, normal wear, or severe wear state, which serves as the grinding wheel wear state label data in the grinding wheel wear dataset. The spectral confocal displacement sensor 302 is fixed to the CNC grinding machine tool 101 via the spectral confocal displacement sensor fixture 304 to measure the distance values ​​between itself and the surrounding area of ​​the grinding wheel 103. By using a preset machine tool coordinate and rotation angle list, it can measure the distance values ​​of each measurement point in the list and form a distance value measurement matrix of the grinding wheel 103 before and after wear. The mean of the difference between the distance value measurement matrices of the grinding wheel 103 before and after wear can be calculated, which serves as the grinding wheel wear amount label data in the grinding wheel wear dataset.

[0062] The data processing and control module 400 includes a controller 401, a computer 402, and a digital twin model 403 of the grinding machine tool. The computer 402 can receive, process, and store data collected by the accelerometer 201, current sensor 202, force gauge 203, acoustic emission sensor 204, industrial microscope 301, and spectral confocal displacement sensor 302. The controller 401 can receive signals from the computer 402 and control the movement of the CNC grinding machine tool 101, as well as control the acquisition and stopping of data from the accelerometer 201, current sensor 202, force gauge 203, acoustic emission sensor 204, industrial microscope 301, and spectral confocal displacement sensor 302. The digital twin model 403 of the grinding machine tool is built within the computer 402 and can realize the virtual-real mapping of the operation of the CNC grinding machine tool 101, which can reflect the processing status in real time and simulate the grinding process that is about to occur. The grinding machine digital twin model 403 detects that the grinding wheel 103 and workpiece 104 are about to contact to begin grinding. It then transmits a signal to the computer 402, which in turn transmits a control signal to the controller 401. The controller 401 controls the accelerometer 201, current sensor 202, force gauge 203, and acoustic emission sensor 204 to begin collecting processing signals. When the grinding wheel 103 and workpiece 104 are about to disengage and the grinding process is about to end, the digital twin model 403 transmits a signal to the computer 402. The computer 402 transmits a control signal to the controller 401, which then controls the accelerometer 201, current sensor 202, force gauge 203, and acoustic emission sensor 204 to stop collecting signals. This process accurately collects signal data during processing to establish a grinding wheel wear dataset, avoiding contamination of the dataset by non-processing data.

[0063] This invention also provides a method for constructing a grinding wheel wear dataset, comprising:

[0064] Step S1: Complete initialization operations such as sensor placement, machine tool power supply, sensor power supply and calibration, and mold and workpiece installation;

[0065] Step S2: Establish communication between the machine tool, sensors, computer and controller; build a digital twin model of the grinding machine tool in the computer; and preset the machine tool coordinates and rotation angle list.

[0066] Step S3: The computer controls the machine tool to move the grinding wheel to the wear measurement position. The distance measurement data matrix corresponding to the unworn grinding wheel is collected by the spectral confocal displacement sensor. The matrix consists of machine tool coordinates, rotation angle, and distance values ​​between the spectral confocal displacement sensor and the surrounding area of ​​the grinding wheel. The data is then transmitted to the computer.

[0067] like Figure 4 , Figure 5 , Figure 6 As shown, during measurement, the spectral confocal displacement sensor is fixed, and the machine tool carrying the grinding wheel moves in a manner of first longitudinal translation and then rotation. The cylindrical surface around the grinding wheel can be unfolded into an imaginary plane. The measurement is performed by preset longitudinal measurement interval Δy and angular interval. This allows us to determine the coordinates and rotation list of the measurement data matrix, where the lateral measurement interval Δx is... The arc length is centered on the radius of the grinding wheel, which is the difference (ld) between the distance l between the spectral confocal displacement sensor and the center of the grinding wheel and the distance d between the sensor and the periphery of the grinding wheel.

[0068] Taking the longitudinal measurement of abrasive wear as an example, the starting point O(1,1) of the longitudinal measurement is the point where the upper boundary of the abrasive part of the abrasive is measurable by the spectral confocal displacement sensor. The abrasive is moved upwards gradually at intervals of Δy in the preset coordinate list until the lower boundary of the abrasive part of the abrasive is measured, i.e., the measurement termination point O(1,N). y And obtain valid measurement values ​​d(1,N) y At this time, a longitudinal measurement of the wear of the grinding wheel is completed, and then the grinding wheel is rotated. And move down (N) y -1)×Δy, that is, when the spectral confocal displacement sensor measures the upper boundary of the preset grinding wheel with abrasive part, the second longitudinal measurement can begin, that is, the data measurement of the second column of the preset list can begin on the plane around the imaginary grinding wheel. Repeating this operation can obtain the data of all measurement points in the preset list.

[0069] During this process, the measurement interval is set to be less than the machine tool's movement interval. Therefore, the computer will repeatedly check whether the machine tool's current coordinates and rotation angles are elements in the preset coordinates and rotation angles list. If the machine tool's movement interval is too long, it will cause the computer to check and compare too many times, and an error message will be thrown. The computer controls the spectral confocal displacement sensor to perform a measurement process only when the machine tool reaches the measurement point of the preset coordinates and rotation angles. Then, the measurement will be repeated when the machine tool moves to the next preset measurement point, so as to avoid the problem of repeated measurement of the same measurement point.

[0070] The flowchart for constructing the distance measurement data matrix is ​​as follows: Figure 7 As shown, step S3 can be further broken down into the following steps:

[0071] Step S31: Initialize parameters, preset machine tool coordinates and rotation angle list, and assign measurement sequence numbers, with the aim of traversing all measurement sequence numbers and obtaining the distance values ​​of all measurement points later;

[0072] Step S32: Determine whether the error count has reached the upper limit. If not, obtain the current machine tool coordinates and rotation angle, and proceed to step S33. If yes, report an error and exit the measurement program.

[0073] Step S33: Determine whether the obtained machine tool coordinates and rotation angles are the same as the measured coordinates and rotation angles corresponding to the current measurement sequence number in the preset list. If not, or because the machine tool movement interval time has not been reached and the measurement value has been obtained, then the error count value is accumulated and the process returns to S32. If yes, the computer will obtain the measurement distance value of the spectral confocal displacement sensor, add the machine tool coordinates, rotation angles and sensor measurement values ​​to the measurement data matrix, update the measurement sequence number to the next measurement point, accumulate the valid coordinate count value, reset the error count value to zero, and proceed to S34.

[0074] Step S34: Determine whether the valid coordinate count has reached the upper limit. If not, wait for the machine tool to proceed to the next movement before taking the measurement, and return to S32. If yes, it means that the measurement values ​​of the entire coordinate list have been obtained, and output the measurement matrix to the computer.

[0075] Step S4: Filter the collected distance measurement data matrix in the computer to reduce the impact of random noise caused by factors such as machine tool vibration;

[0076] Step S5: Start grinding. When the digital twin model of the grinding machine tool detects that the grinding wheel and the workpiece are about to contact, it triggers the acceleration sensor, current sensor, force gauge, and acoustic emission sensor to collect vibration, spindle current, grinding force, and acoustic emission signals during the processing. When the model detects that the grinding wheel and the workpiece are about to separate, it stops collecting signals and transmits the signals collected during the process to the computer as grinding wheel wear characteristic data. This accurately obtains the signal data during processing to establish a grinding wheel wear dataset, avoiding contamination of the dataset by non-processing data.

[0077] Step S6: After several passes or grinding, when the material volume removed reaches the threshold, the machine tool is controlled by computer to move the grinding wheel to the measurement position. The surface morphology of the grinding wheel is observed by an industrial microscope to determine whether the grinding wheel is in the initial wear, normal wear, or severe wear state, which is used as the grinding wheel wear state label data.

[0078] Step S7: The machine tool is controlled by computer to move the grinding wheel to the wear measurement position. Using the method described in step S3, the distance measurement data matrix after the grinding wheel wear is collected by the spectral confocal displacement sensor, transmitted to the computer and filtered. Then the mean of the difference between the distance measurement data matrix and the distance measurement data matrix corresponding to the grinding wheel without wear is calculated as the grinding wheel wear amount label data.

[0079] like Figure 6 As shown, Δy and The advantage of using a pre-set fixed quantity design for machine tool coordinates and rotation lists is that the transverse measurement interval Δx will decrease as the wear amount Δd of the grinding wheel increases, but due to Δy and The number of longitudinal and transverse measurement points N is preset and remains unchanged. y With N x Since all parameters remain unchanged, it can be seen that even if the wear of the grinding wheel increases continuously, this method can still accurately obtain the corresponding measurement points before and after the wear of the grinding wheel by dynamically adjusting the transverse measurement interval (Δx and Δx' before and after wear, respectively), such as O(N). x N y ) and O'(N x N y And ensure that the total number of measurement points remains unchanged, so as to establish an accurate mapping relationship of the same point before and after wear.

[0080] Step S8: Repeat S5 to S7 to construct a grinding wheel wear dataset containing a single grinding wheel;

[0081] Step S9: Replace the grinding wheel and workpiece, repeat S3 to S8, and construct a grinding wheel wear dataset containing multiple grinding wheels.

Claims

1. A system for constructing a grinding wheel wear dataset, characterized in that, include: CNC grinding machine module (100) for performing grinding processes; The machining signal acquisition module (200) is used to acquire multi-source signals such as vibration signals, spindle current signals, grinding force signals and acoustic emission signals during the grinding process; A grinding wheel wear measurement and acquisition module (300) is used to acquire the surface morphology of the grinding wheel and measure the distance value measurement matrix between the grinding wheel periphery and the spectral confocal displacement sensor; The data processing and control module (400) is used to preset the machine tool coordinates and rotation angle list, receive and store sensor data, run the digital twin model of the grinding machine tool, realize feedback control between the machine tool and the sensor, and calculate the wear of the grinding wheel. The abrasive wear measurement and acquisition module (300) includes an industrial microscope (301), a spectral confocal displacement sensor (302), an industrial microscope fixture (303), and a spectral confocal displacement sensor fixture (304); The industrial microscope (301) is fixed on the CNC grinding machine (101) by the industrial microscope fixture (303) to observe the surface morphology of the grinding wheel (103) and determine whether the grinding wheel (103) is in the initial wear, normal wear or severe wear state, and serve as the grinding wheel wear state label data in the grinding wheel wear dataset; The spectral confocal displacement sensor (302) is fixed on the CNC grinding machine (101) by the spectral confocal displacement sensor fixture (304) to measure the distance between itself and the periphery of the grinding wheel (103). By using a preset machine tool coordinate and rotation angle list, it can measure the distance value of each measurement point in the list and form a distance value measurement matrix of the grinding wheel (103) before and after wear. The mean of the difference between the distance value measurement matrices of the grinding wheel (103) before and after wear can be calculated as the grinding wheel wear amount label data in the grinding wheel wear dataset.

2. The system for constructing a grinding wheel wear dataset as described in claim 1, characterized in that, The CNC grinding machine tool module (100) includes a CNC grinding machine tool (101), a machine tool spindle box (102), a grinding wheel (103), a workpiece (104), and a machine tool worktable (105). It has a spindle rotation angle recording function and can realize spindle rotation angle control.

3. The system for constructing a grinding wheel wear dataset as described in claim 1, characterized in that, The machining signal acquisition module (200) includes an acceleration sensor (201), a current sensor (202), a force gauge (203), and an acoustic emission sensor (204). The acceleration sensor (201) is arranged on the machine tool spindle box (102), the current sensor (202) is arranged on the three-phase AC cable of the machine tool spindle in the machine tool spindle box (102), the workpiece (104) is fixed on the force gauge (203), and the force gauge (203) and the acoustic emission sensor (204) are arranged on the machine tool worktable (105). The vibration signal, spindle current signal, grinding force signal, and acoustic emission signal collected by the above sensors together constitute the grinding tool wear characteristic data in the grinding tool wear dataset.

4. The system for constructing a grinding wheel wear dataset as described in claim 1, characterized in that, The data processing and control module (400) includes a controller (401), a computer (402), and a digital twin model of the grinding machine tool (403). The computer (402) can receive, process, and store data from the CNC grinding machine tool module (100), the machining signal acquisition module (200), and the grinding wheel wear measurement and acquisition module (300). The controller (401) can receive signals from the computer (402) and control the movement of the CNC grinding machine tool (101) as well as control the acquisition and stopping of the accelerometer (201), current sensor (202), force gauge (203), acoustic emission sensor (204), industrial microscope (301), and spectral confocal displacement sensor (302). The digital twin model (403) of the grinding machine tool is built in the computer (402) and can realize the virtual-real mapping of the operation of the CNC grinding machine tool (101). It can reflect the processing situation in real time and simulate the grinding processing situation that will occur based on the machine tool.

5. The dataset construction method of the grinding wheel wear dataset construction system according to claim 1, characterized in that, include: S1: Complete initialization operations such as sensor placement, machine tool power supply, sensor power supply and calibration, and mold and workpiece installation; S2: Establish communication between the machine tool, sensors, computer and controller; build a digital twin model of the grinding machine tool in the computer; and preset the machine tool coordinates and rotation angle list. S3: The computer controls the machine tool to move the grinding wheel to the wear measurement position. The computer collects the distance measurement data matrix corresponding to the unworn grinding wheel, which consists of machine tool coordinates, rotation angle, and distance values ​​between the spectral confocal displacement sensor and the surrounding area of ​​the grinding wheel. The data is then transmitted to the computer. S4: Filter the collected distance measurement data matrix in the computer; S5: Start grinding. When the grinding wheel and the workpiece are about to come into contact in the machine tool digital twin model, the machining signal acquisition sensor is triggered to collect signals. When the grinding wheel and the workpiece are about to separate, the signal acquisition stops and the signals collected in this process are transmitted to the computer as grinding wheel wear characteristic data. S6: When the material volume removed by grinding reaches the threshold after several passes, the machine tool is controlled by computer to move the grinding wheel to the measurement position. The surface morphology of the grinding wheel is observed by an industrial microscope to determine whether the grinding wheel is in the initial wear, normal wear or severe wear state, and this is used as the grinding wheel wear state label data. S7: The computer controls the machine tool to move the grinding wheel to the wear measurement position. The distance measurement data matrix after the grinding wheel wears is collected by the spectral confocal displacement sensor, transmitted to the computer and filtered. Then the mean of the difference between the distance measurement data matrix and the distance measurement data matrix corresponding to the grinding wheel without wear is calculated as the grinding wheel wear amount label data. S8: Repeat S5 to S7 to construct a grinding wheel wear dataset containing a single grinding wheel; S9: Change the grinding wheel and workpiece, repeat S3 to S8, and build a grinding wheel wear dataset containing multiple grinding wheels.

6. The dataset construction method of the grinding wheel wear dataset construction system as described in claim 5, characterized in that, In step S3, the spectral confocal displacement sensor is fixed, and the machine tool carries the grinding wheel and moves it in a manner of first longitudinal translation and then rotation. The cylindrical surface around the grinding wheel can be unfolded into an imaginary plane. By preset longitudinal measurement interval Δy and rotation interval Δφ, the coordinates and rotation list of the measurement data matrix can be determined. The transverse measurement interval Δx is the arc length with Δφ as the center and the grinding wheel radius, i.e., the distance l between the spectral confocal displacement sensor and the center of the grinding wheel and the difference ld between the distance value d measured by the sensor and the distance value d between the sensor and the periphery of the grinding wheel, as the radius. In a longitudinal measurement of abrasive wear, the starting point of the longitudinal measurement is the point where the upper boundary of the abrasive part of the abrasive is measured by the spectral confocal displacement sensor. If and only if the spectral confocal displacement sensor measures a valid value, the abrasive moves upward step by step at intervals of Δy in the preset coordinate list until the lower boundary of the abrasive part of the abrasive is measured, i.e., the measurement termination point. When a valid value is obtained, the longitudinal measurement of abrasive wear is completed. Then, the abrasive is rotated by Δφ and moved downward until the spectral confocal displacement sensor measures the upper boundary of the abrasive part of the abrasive. The second longitudinal measurement can then begin, i.e., the data measurement of the second column of the preset list is started on the imaginary abrasive peripheral plane. Repeating this operation can obtain the data of all measurement points in the preset list. The measurement interval is set to be less than the machine tool's movement interval. The computer repeatedly checks whether the machine tool's current coordinates and rotation angles are elements in the preset coordinates and rotation angles list. If the machine tool's movement interval is too long, the computer will perform too many checks and comparisons, and will throw an error message. The computer controls the spectral confocal displacement sensor to perform a measurement process only when the machine tool reaches the preset coordinates and rotation angle measurement point. Then, when the machine tool moves to the next preset measurement point, the measurement is repeated to avoid repeated measurement of the same measurement point.

7. The dataset construction method of the grinding wheel wear dataset construction system as described in claim 5, characterized in that, Step S3 can be further broken down into the following steps: S31: Initialize parameters, preset machine tool coordinates and rotation angle list, and assign measurement sequence numbers; S32: Determine if the error count has reached the upper limit. If not, obtain the current machine tool coordinates and rotation angle, and proceed to step S33. If yes, report an error and exit the measurement program. S33: Determine whether the obtained machine tool coordinates and rotation angle are the same as the coordinates and rotation angle to be measured corresponding to the current measurement sequence number in the preset list. If not, the error count value is accumulated and the process returns to S32. If yes, the computer will obtain the measurement distance value of the spectral confocal displacement sensor, add the machine tool coordinates, rotation angle and sensor measurement value to the measurement data matrix, update the measurement sequence number to the next measurement point, accumulate the valid coordinate count value, reset the error count value to zero, and proceed to S34. S34: Determine if the valid coordinate count has reached the upper limit. If not, wait for the machine tool to proceed with the next movement before taking the measurement, and return to S32. If yes, output the measurement matrix to the computer.

8. The dataset construction method of the grinding wheel wear dataset construction system as described in claim 5, characterized in that, In step S5, when the digital twin model of the grinding machine tool detects that the grinding wheel and the workpiece are about to come into contact, it triggers the acceleration sensor, current sensor, force gauge, and acoustic emission sensor to collect vibration, spindle current, grinding force, and acoustic emission signals during the processing. When the model detects that the grinding wheel and the workpiece are about to separate, it stops collecting signals and transmits the signals collected in the process to the computer as grinding wheel wear characteristic data.

9. The dataset construction method of the grinding wheel wear dataset construction system as described in claim 5, characterized in that, In step S7, the method for obtaining the distance value measurement data matrix is ​​the same as in step S3; During this process, the transverse measurement interval Δx will decrease as the wear amount Δd of the grinding wheel increases. However, since Δy and Δφ are preset to remain constant, the number of longitudinal and transverse measurement points N will remain constant. y With N x All remain unchanged. Even as the wear of the grinding wheel increases, the corresponding measurement points before and after the wear of the grinding wheel can still be accurately obtained through the dynamic adjustment of the transverse measurement interval, and the total number of measurement points remains unchanged, thereby establishing an accurate mapping relationship between the same point before and after wear.

Images