A polishing device for automobile accessory machining

Abstract

The application relates to a polishing device for automobile accessory machining, which comprises a polishing execution unit, a detection unit and a controller, the controller comprising a contact state characteristic spectrum construction module, a wear morphology inversion module and a heterogeneous parameter compensation control module. The contact state characteristic spectrum construction module fuses force-torque frequency domain characteristics, acoustic emission resonance peak frequency shift characteristics and temperature gradient direction characteristics into a real-time contact state characteristic spectrum; the wear morphology inversion module online inverts the eccentric wear azimuth angle, wear curvature radius and central concave depth of the polishing head based on a physical correlation vector; and the heterogeneous parameter compensation control module generates posture deflection angle instructions, axial micro-vibration instructions and rotational speed feed ratio instructions according to different wear morphologies, so that targeted active compensation of non-uniform wear is realized. The application has the effect of maintaining the high consistency of complex curved surface polishing quality within the service life of a single polishing head.

Description

Technical Field

[0001] This application relates to the field of automotive parts processing equipment, and in particular to a grinding device for automotive parts processing. Background Technology

[0002] In the automotive parts processing field, grinding equipment is widely used for surface finishing of complex curved parts such as engine cylinder head intake ports and turbocharger turbine housings. Existing grinding equipment typically incorporates one or more of force sensors, acoustic emission sensors, or vision sensors to address grinding head wear during the grinding process through constant force control, fixed-cycle feed compensation, or threshold-triggered shutdown. For example, some devices monitor and maintain the normal contact force in real time to compensate for the contact pressure attenuation caused by overall grinding head wear; others use acoustic emission signals to monitor the grinding head's passivation level, triggering a shutdown and replacement when the signal characteristics exceed a preset threshold; still others employ periodic axial feed, advancing the grinding head a preset distance towards the workpiece at fixed time intervals to compensate for the reduction in the overall height of the grinding head.

[0003] However, the aforementioned existing technologies all rely on the implicit assumption that "grinding head wear is uniform," meaning they assume that the wear degree of each part of the grinding head is consistent, requiring only compensation for changes in overall height or overall sharpness reduction. However, in the actual grinding of complex curved surfaces, due to the drastic changes in the curvature of the component's contour, there are significant differences in the contact arc length, relative linear velocity, and pressure distribution between different areas of the grinding head and the workpiece. This inevitably leads to non-uniform wear patterns in the grinding head, such as edge collapse, central concavity, or unilateral wear. Such non-uniform wear will cause distortion of the contact pressure field. Even if the total pressure or total feed is maintained at the target value, over-grinding streaks or missed grinding defects will still appear in local areas, severely disrupting the consistency of curved surface grinding. Existing technologies cannot detect the current wear pattern of the grinding head online, nor can they specifically compensate for the deterioration of local contact conditions caused by non-uniform wear morphology. Ultimately, they can only rely on frequent grinding head replacements to ensure processing quality, resulting in high production costs and low equipment uptime.

[0004] To address the aforementioned technical problems, this invention proposes a grinding device for processing automotive parts. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a grinding device for automotive parts processing. It constructs a real-time contact state feature spectrum by integrating force-torque combined characteristics, acoustic emission resonance peak frequency shift characteristics, and temperature gradient characteristics. Based on this spectrum, it online inverts the non-uniform wear morphology feature parameters of the grinding head, thereby generating heterogeneous compensation commands corresponding to specific wear patterns. This achieves targeted active compensation for non-uniform wear, maintaining a high degree of consistency in the grinding quality of complex curved surfaces throughout the single grinding head lifespan. The specific technical solution adopted is as follows: A grinding device for processing automotive parts, comprising: The grinding execution unit has a grinding head and a robotic arm that drives the grinding head to perform multi-axis motion; The detection unit includes at least a force sensor for collecting the grinding contact force, an acoustic emission sensor for collecting acoustic emission signals, an infrared array sensor for collecting the temperature distribution in the grinding area, and a vision sensor for acquiring the spatial position of the accessory and the grinding head. A controller, connected to the detection unit and the grinding execution unit, is characterized in that the controller comprises: The contact state feature spectrum construction module is used to acquire signals from the force sensor, acoustic emission sensor and infrared array sensor in real time, extract force-torque joint features, acoustic emission resonance peak frequency shift features and temperature gradient features from them, and fuse the three into a real-time contact state feature spectrum; the real-time contact state feature spectrum includes a torque fluctuation frequency domain energy distribution vector, resonance peak frequency shift scalar and temperature gradient direction vector. The wear morphology inversion module is used to invert the non-uniform wear morphology feature parameters of the grinding head online based on the real-time contact state feature spectrum. The non-uniform wear morphology feature parameters include at least the wear azimuth angle, wear curvature radius and center depression depth. The heterogeneous parameter compensation control module is used to generate heterogeneous compensation instructions based on the non-uniform wear morphology characteristic parameters. The heterogeneous compensation instructions include at least one of the following: grinding head attitude deviation angle instruction, axial micro-vibration parameter instruction, and spindle speed to feed rate ratio instruction. The module also controls the grinding execution unit to perform the corresponding compensation actions.

[0006] Preferably, the contact state feature spectrum construction module includes: The force-torque joint feature extraction unit is used to extract torque fluctuation features related to the rotation frequency of the grinding head based on the normal force and torque signals collected by the force sensor. The acoustic emission feature extraction unit is used to perform spectral analysis on the acoustic emission signal, lock the resonance peak of a specific frequency band, and monitor its frequency offset in real time. The temperature gradient feature extraction unit is used to calculate the temperature gradient vector along the front and rear edges of the grinding contact area based on the temperature distribution data collected by the infrared linear array sensor.

[0007] Preferably, the force-torque joint feature extraction unit is specifically used to: perform spectral analysis on the real-time torque signal, extract the energy distribution corresponding to the fundamental frequency and harmonics of the grinding head rotation, and use the energy distribution as a force-torque joint feature characterizing the direction and degree of uneven wear.

[0008] Preferably, the acoustic emission feature extraction unit is specifically used to: continuously analyze acoustic emission signals during the grinding process, identify the frequency shift of a specific resonance peak caused by changes in the contact arc length and contact stiffness between the grinding head and the workpiece, and use the frequency shift as a frequency shift feature of the acoustic emission resonance peak that characterizes the absolute magnitude of wear.

[0009] Preferably, the temperature gradient feature extraction unit is specifically used to: acquire the temperature field at the front and rear edges of the grinding contact point using an infrared linear array sensor, calculate the direction and magnitude of the maximum temperature gradient, and use the direction of the gradient vector as the temperature gradient feature pointing to the area of ​​most severe wear.

[0010] Preferably, the wear morphology inversion module includes: The calibration storage unit stores the physical correlation vector between the pre-calibrated basic wear morphology library and the corresponding contact state feature spectrum; The online solution unit is used to decompose the contact state feature spectrum acquired in real time into the physical correlation vector space, and to calculate the wear azimuth angle, wear curvature radius and center depression depth of the current grinding head in real time through a weighted fusion algorithm.

[0011] Preferably, the heterogeneous parameter compensation control module includes: The attitude deviation angle compensation unit is used to generate a dynamic yaw angle command based on the wear azimuth angle and output it to the robotic arm to guide the center of contact pressure toward the area of ​​slight wear. An active micro-vibration modulation unit is used to generate axial micro-vibration parameter commands based on the depth of the central depression and apply them to the main shaft to homogenize the contact pressure distribution through dynamic impact. The spindle speed feed linkage unit is used to dynamically adjust the ratio of spindle speed to feed speed according to the wear radius of curvature.

[0012] Preferably, the grinding execution unit integrates a piezoelectric actuator within its spindle; the active micro-vibration modulation unit is specifically used to: calculate the vibration amplitude and frequency based on the depth of the central recess, and drive the piezoelectric actuator to generate axial micro-amplitude vibrations that are not integer ratios to the spindle rotation frequency.

[0013] Preferably, the controller further includes: The forward-looking allowance redistribution module is used to determine the matching relationship between the severely worn area of ​​the grinding head and the high removal amount area of ​​the workpiece on the subsequent grinding path based on the workpiece allowance distribution information on the grinding path obtained by the vision sensor and the non-uniform wear morphology feature parameters, and adjust the grinding allowance allocation strategy of each area accordingly.

[0014] Preferred options also include: Online trimmer and online self-healing control module; The online self-repair control module is used to trigger the online trimmer to trim the edge of the grinding head in situ according to a preset trimming path when the parameters output by the wear morphology inversion module indicate that the grinding head has experienced edge collapse, and to control the grinding execution unit to continue grinding work after the trimming is completed.

[0015] In summary, this application includes at least one of the following beneficial technical effects: This application, through the collaboration of the contact state feature spectrum construction module and the wear morphology inversion module, integrates force-torque frequency domain features, acoustic emission resonance peak frequency shift features, and temperature gradient direction features into a real-time contact state feature spectrum. Based on the physical correlation vector, it online inverts the non-uniform wear morphology feature parameters of the grinding head, including the wear azimuth angle, wear curvature radius, and central depression depth. For the first time, it realizes online perception of the non-uniform wear morphology of the grinding head without adding dedicated wear morphology measurement hardware. This upgrades the controller from only being able to determine "whether wear exists" to being able to know "the specific morphology and distribution of wear", providing a complete information foundation for subsequent targeted compensation.

[0016] This application uses a heterogeneous parameter compensation control module to generate corresponding heterogeneous compensation commands for different wear morphology characteristics. It generates dynamic yaw angle commands based on the wear azimuth angle to guide the center of contact pressure to shift to the area with slight wear, generates axial micro-vibration parameter commands based on the depth of the center indentation to homogenize the contact pressure distribution through dynamic impact, and dynamically adjusts the ratio of spindle speed to feed speed based on the wear curvature radius to avoid local over-wear. This breaks the mindset of existing technologies that only compensate for wear with a single isomorphic parameter such as "increasing total pressure" or "overall feed", and achieves precise adaptation of "different wear morphologies and different compensation strategies".

[0017] This application sets the axial micro-vibration frequency to a non-integer ratio with the spindle rotation frequency, so that the vibration phase relationship between two adjacent rotations changes continuously. This homogenizes the contact pressure distribution in terms of time averaging, eliminates the conditions for the generation of periodic vibration stripes in terms of mechanism, and avoids the formation of fixed grinding marks on the workpiece surface.

[0018] This application uses a forward-looking allowance redistribution module to feed forward the non-uniform wear morphology feature parameters output by the wear morphology inversion module to the grinding path planning stage. Before grinding is executed, it predicts the matching conflict between the severely worn area of ​​the grinding head and the high removal area of ​​the workpiece on the subsequent grinding path, and actively adjusts the grinding allowance allocation strategy of each area. This upgrades "passively bearing the consequences of wear" to "actively using wear information for global optimization", further ensuring the consistency of grinding quality in batch processing.

[0019] This application utilizes an online self-healing control module. When the parameters output by the wear morphology inversion module indicate that the grinding head has experienced edge collapse, the online dresser is automatically triggered to perform in-situ trimming of the grinding head edge. After trimming, the reference parameters are updated and processing continues, transforming the grinding head from a "disposable consumable" into a "quasi-tool that can be maintained online." This significantly extends the effective service life of the grinding head, reduces the frequency of downtime for replacement, and improves equipment uptime.

[0020] Because it employs a wear morphology inversion method based on physical correlation vectors, the inversion process in this application has clear physical interpretability and does not rely on a large amount of training data. When the grinding head or workpiece material is changed, only a small amount of offline calibration is required for rapid adaptation, demonstrating excellent engineering deployability. Attached Figure Description

[0021] Figure 1 This is an overall structural block diagram of the grinding device for processing automotive parts in the embodiments of this application; Figure 2 This is a block diagram of the contact state feature spectrum construction module in an embodiment of this application; Figure 3 This is a block diagram of the wear morphology inversion module in the embodiments of this application; Figure 4 This is a block diagram of the heterogeneous parameter compensation control module in an embodiment of this application; Figure 5 This is a schematic diagram of the workflow of an embodiment of this application.

[0022] Explanation of reference numerals in the attached drawings: 10. Grinding execution unit; 11. Grinding head; 12. Robotic arm; 13. Piezoelectric actuator; 14. In-line dressing device; 20. Detection unit; 21. Force sensor; 22. Acoustic emission sensor; 23. Infrared array sensor; 24. Vision sensor; 30. Controller; 31. Contact State Feature Spectrum Construction Module; 311. Force-Torque Joint Feature Extraction Unit; 312. Acoustic Emission Feature Extraction Unit; 313. Temperature Gradient Feature Extraction Unit; 32. Wear Morphology Inversion Module; 321. Calibration Storage Unit; 322. Online Solving Unit; 33. Heterogeneous Parameter Compensation Control Module; 331. Attitude Angle Compensation Unit; 332. Active Micro-Vibration Modulation Unit; 333. Rotational Speed ​​Feed Linkage Unit; 34. Look-Ahead Margin Redistribution Module; 35. Online Self-Repair Control Module. Detailed Implementation

[0023] The following is in conjunction with the appendix Figures 1-5 This application will be described in further detail.

[0024] This application discloses a grinding apparatus for processing automotive parts. To facilitate understanding of the technical solution of this application, the key terms involved in this application will first be explained in a unified manner.

[0025] Real-time contact state feature spectrum: refers to a multi-dimensional feature vector formed by fusing the torque fluctuation frequency domain energy distribution vector, the resonance peak frequency offset scalar, and the temperature gradient direction vector, which is used to comprehensively characterize the physical feature set of the contact state between the grinding head and the workpiece at the current moment.

[0026] Torque fluctuation frequency domain energy distribution vector: refers to the N+1-dimensional vector formed by extracting the energy values ​​at the fundamental frequency of the grinding head rotation and its Nth harmonic (N is a positive integer, N=4 in this embodiment) after performing a fast Fourier transform on the time-domain torque signal collected by the torque sensor.

[0027] Resonance frequency offset scalar: refers to the offset of a specific resonance frequency value of the acoustic emission signal currently monitored in real time relative to the initial calibrated resonance frequency value of the grinding head in its brand-new state, in Hz.

[0028] Temperature gradient direction vector: This refers to the direction of the maximum temperature gradient calculated based on the temperature field distribution of the polishing area obtained from the infrared array sensor. Its direction points to the location where the temperature rises the fastest in the contact area, and physically corresponds to the area with the most severe wear.

[0029] Azimuth angle of uneven wear: Defined as an angle value in the coordinate system of the grinding head end face, it represents the azimuth angle of the direction of the most severe non-uniform wear of the grinding head relative to the preset reference axis, and the value ranges from 0° to 360°.

[0030] Wear radius of curvature: The radius of curvature value obtained by fitting the local morphology of the grinding head surface after wear. The smaller the radius of curvature, the sharper the wear and the more concentrated the local contact stress.

[0031] Center concavity depth: The axial height difference between the center area of ​​the grinding head and the unworn edge area, measured in μm, used to characterize the degree of concavity in the center area of ​​the grinding head.

[0032] Reference Figure 1 A grinding device for processing automotive parts includes a grinding execution unit 10, a detection unit 20, and a controller 30.

[0033] The grinding execution unit 10 includes a grinding head 11 and a robotic arm 12 that drives the grinding head 11 to perform multi-axis movements. In this embodiment, the grinding head 11 is a rubber-bonded abrasive grinding head, which is mounted to the end of the robotic arm 12 via a spindle flange. The robotic arm 12 is a six-axis industrial robot capable of driving the grinding head 11 to move along a planned path in three-dimensional space under the command of the controller 30. A piezoelectric actuator 13 is integrated within the spindle to generate high-frequency micro-amplitude vibrations in the axial direction.

[0034] The detection unit 20 includes a force sensor 21, an acoustic emission sensor 22, an infrared array sensor 23, and a vision sensor 24.

[0035] Force sensor 21, in this embodiment, is a six-dimensional force sensor, installed between the spindle flange of the grinding execution unit 10 and the end of the robotic arm 12. It is used to collect normal contact force and triaxial torque signals during the grinding process in real time. The sampling frequency is set to 2kHz to meet the frequency resolution requirements of subsequent spectrum analysis. Acoustic emission sensor 22 is installed on the side of the spindle housing near the grinding head 11. Its frequency response range is 50kHz to 500kHz, and the sampling frequency is set to 2MHz. It is used to collect acoustic emission signals generated during the grinding process in real time. Infrared array sensor 23 consists of multiple infrared sensing units arranged back and forth along the grinding movement direction. It is installed on a bracket on the side of the grinding head 11. Its field of view covers the grinding contact point and its front and rear edges, and is used to acquire temperature distribution data of the grinding area in a non-contact manner. Vision sensor 24 is a 3D structured light camera, installed at the end of the robotic arm 12 or above the grinding station. It is used to acquire three-dimensional point cloud data of the parts before grinding, and to acquire the spatial positional relationship between the parts and the grinding head 11 and the workpiece allowance distribution information on the grinding path in real time during the grinding process.

[0036] The controller 30 is electrically connected to the detection unit 20 and the grinding execution unit 10. In this embodiment, the controller 30 is an industrial embedded controller with a built-in FPGA acceleration card for hardware acceleration of signal processing, ensuring the real-time performance of each module. The controller 30 includes a contact state feature spectrum construction module 31, a wear morphology inversion module 32, a heterogeneous parameter compensation control module 33, a look-ahead margin reallocation module 34, and an online self-repair control module 35.

[0037] This application integrates multi-physical sensor signals into a multi-dimensional feature spectrum characterizing the wear morphology of the grinding head through a contact state feature spectrum construction module 31. The wear morphology inversion module 32 then uses this spectrum to online invert the non-uniform wear morphology feature parameters of the grinding head. Subsequently, the heterogeneous parameter compensation control module 33 generates corresponding heterogeneous compensation commands for different wear morphologies, achieving closed-loop control of non-uniform wear through "sensing → inversion → compensation." Furthermore, the look-ahead margin redistribution module 34 feeds wear morphology information forward to the grinding path planning stage, avoiding conflicts between severely worn areas and high removal areas before grinding execution. The online self-repair control module 35 triggers in-situ repair when wear exceeds limits, further extending the effective service life of the grinding head. The coordinated operation of these modules enables the device to maintain a high degree of consistency in the grinding quality of complex curved surfaces throughout a single grinding head life cycle, solving the industry problem of large grinding quality fluctuations and low grinding head utilization caused by the inability to sense and specifically compensate for non-uniform wear in existing technologies.

[0038] Reference Figure 2 The contact state feature spectrum construction module 31 is used to acquire signals from the force sensor 21, acoustic emission sensor 22, and infrared array sensor 23 in real time, extract force-torque joint features, acoustic emission resonance peak frequency shift features, and temperature gradient features from them, and fuse the three into a real-time contact state feature spectrum. The contact state feature spectrum construction module 31 includes a force-torque joint feature extraction unit 311, an acoustic emission feature extraction unit 312, and a temperature gradient feature extraction unit 313.

[0039] The force-torque joint feature extraction unit 311 is used to extract torque fluctuation features related to the rotation frequency of the grinding head 11 based on the normal force and torque signals collected by the force sensor 21. Its specific working process is as follows: First, the real-time torque signal collected by the force sensor 21 is windowed; in this embodiment, a Hanning window is used to reduce spectral leakage. Then, a fast Fourier transform is performed on the windowed data to obtain the spectrum of the torque signal. Next, the amplitudes at the fundamental frequency f0 of the grinding head 11 and its first four harmonics (2f0, 3f0, 4f0, 5f0) are extracted from the spectrum to form a 5-dimensional energy distribution vector. Finally, this vector is normalized, and the normalized energy distribution vector is used as the force-torque joint feature characterizing the direction and degree of uneven wear. It should be noted that when the grinding head 11 is in an ideal uniform wear state, the torque fluctuation is mainly based on the fundamental frequency, and the harmonic energy is low; however, when the grinding head 11 experiences unilateral wear, the contact force in the wear area changes abruptly with each rotation, and the harmonic energy (especially the 2nd and 3rd harmonics) increases significantly. Therefore, the change in this energy distribution vector can sensitively reflect the occurrence and direction of the wear.

[0040] The acoustic emission feature extraction unit 312 is used to perform spectral analysis on the acoustic emission signal collected by the acoustic emission sensor 22, lock the resonance peak in a specific frequency band, and monitor its frequency offset in real time. Its specific working process is as follows: With the grinding head 11 in a brand-new state, a baseline run is first performed. One or more peak frequencies with stable amplitude and a signal-to-noise ratio higher than a preset threshold in the acoustic emission spectrum under this state are recorded, and the most stable one is selected as the baseline resonance peak frequency f_r0. In this embodiment, for a rubber grinding head of a specific specification and an aluminum alloy workpiece for the cylinder head air passage of an automobile, the measured baseline resonance peak frequency is approximately 180kHz. During subsequent continuous grinding, the acoustic emission feature extraction unit 312 continuously performs short-time Fourier transform on the acoustic emission signal, uses a peak search algorithm to track the current center frequency f_r of the resonance peak in real time, and calculates the real-time frequency offset Δf = f_r - f_r0. As the wear of the grinding head intensifies, the actual contact arc length between the grinding head and the workpiece increases and the contact stiffness decreases. As a result, the resonance peak will monotonically drift towards the lower frequency direction. Therefore, this frequency drift can be used as a characteristic of the acoustic emission resonance peak frequency shift to characterize the absolute magnitude of wear.

[0041] The temperature gradient feature extraction unit 313 is used to calculate the temperature gradient vector along the front and rear edges of the grinding contact area based on the temperature distribution data collected by the infrared array sensor 23. Specifically, each sensing unit of the infrared array sensor 23 acquires the temperature values ​​of the grinding contact point and its front and rear edges (extending approximately 2mm along the grinding movement direction) in real time, forming a one-dimensional temperature field distribution. The temperature gradient feature extraction unit 313 uses the central difference method to calculate the temperature change rate between adjacent sensing units, searching for the maximum temperature gradient value and its corresponding direction. If the direction of the maximum temperature gradient basically points to a certain position in front of the contact point, then that direction is the temperature gradient direction vector. Due to non-uniform wear causing local pressure concentration in the contact area, a higher frictional temperature rise will occur at this location, forming a characteristic "hot spot tail" phenomenon. Therefore, the direction of this gradient vector can point to the area with the most severe wear.

[0042] After obtaining the above three features, the contact state feature spectrum construction module 31 fuses them using vector concatenation, that is, concatenating the torque fluctuation frequency domain energy distribution vector (5-dimensional), the resonance peak frequency offset scalar (1-dimensional), and the temperature gradient direction vector (2-dimensional, including direction and amplitude) end to end into an 8-dimensional real-time contact state feature spectrum. This fusion method has low computational cost, high real-time performance, and fully preserves the physical meaning of each feature. The above feature construction process is executed every 10ms to ensure that the update of the contact state feature spectrum can reflect the dynamic changes of the grinding process in a timely manner.

[0043] Reference Figure 3The wear morphology inversion module 32 is used to invert the non-uniform wear morphology characteristic parameters of the grinding head 11 online based on the real-time contact state characteristic spectrum, including the wear azimuth angle, wear curvature radius, and center depression depth. The wear morphology inversion module 32 includes a calibration storage unit 321 and an online solution unit 322.

[0044] The calibration storage unit 321 stores the physical correlation vector between the pre-calibrated basic wear morphology library and the corresponding contact state feature spectrum. The process of establishing the basic wear morphology library is as follows: In the offline stage, a standard grinding head 11 and a standard workpiece sample of the same model as the online grinding condition are selected, and five basic wear morphologies are artificially prepared or naturally worn out, namely: (a) uniform wear type (no obvious non-uniform features); (b) central concave type (wear depth in the central area is about 50 μm); (c) edge collapse type (wear depth in the edge area is about 40 μm); (d) 0° azimuth unilateral wear type (wear depth in the 0° direction of the end face reference coordinate system is about 30 μm); (e) 90° azimuth unilateral wear type (wear depth in the 90° direction is about 30 μm). For each basic wear morphology, the grinding process is run under standard working conditions, and the contact state feature spectrum corresponding to the morphology is obtained through the contact state feature spectrum construction module 31. The feature spectrum is then normalized to obtain the physical correlation vector corresponding to the wear morphology. The above five sets of physical correlation vectors, along with the corresponding wear azimuth angle, wear curvature radius, and center indentation depth parameters, are stored together in calibration storage unit 321. It should be noted that the "physical correlation vectors" here are not weight parameters obtained through neural network training, but feature template vectors that are directly calibrated through physical experiments and characterize the causal relationship between specific wear morphology and multiple physical features.

[0045] The online solution unit 322 is used to decompose the real-time acquired contact state feature spectrum into a physical correlation vector space, and to calculate the wear azimuth angle, wear curvature radius, and center depression depth of the current grinding head 11 in real time through a weighted fusion algorithm. The specific solution process is as follows: First, the currently acquired contact state feature spectrum is normalized in the same way as the calibration process; then, the cosine similarity between the normalized real-time feature spectrum and the five sets of physical correlation vectors stored in the calibration storage unit 321 is calculated to obtain five similarity values ​​S_i (i=1 to 5); next, using each similarity value S_i as a weight, a weighted average is calculated for each set of corresponding wear morphology parameters (wear azimuth angle θ_i, wear curvature radius R_i, and center depression depth D_i) in the calibration library, i.e.: θ = Σ(S_i × θ_i) / ΣS_i R = Σ(S_i × R_i) / ΣS_i D = Σ(S_i × D_i) / ΣS_i Therefore, the online solution unit 322 outputs three non-uniform wear morphology characteristic parameters of the current grinding head 11 in real time. This weighted fusion algorithm has low computational complexity and can complete a solution within 1ms under the hardware conditions of the controller 30, fully meeting the real-time control requirements.

[0046] To facilitate understanding, a specific inversion example is given below: Assume that the cosine similarity between the current real-time contact state feature spectrum and the edge collapse type in the basic wear morphology library is 0.82, the similarity with (c) uniform wear type is 0.15, and the similarity with other types is less than 0.05. After normalization, the weight coefficients for each type are: edge collapse type 0.84, uniform wear type 0.16. The parameters corresponding to the edge collapse type in the library are θ=0°, R=15mm, D=5μm; the parameters corresponding to the uniform wear type are θ=0°, R=50mm, D=2μm. Then, the weighted average output is: wear azimuth angle 0°, wear curvature radius R=0.84×15+0.16×50≈20.6mm, center depression depth D=0.84×5+0.16×2≈4.5μm. This result comprehensively reflects the current composite wear state of the grinding head, characterized primarily by edge collapse, while also exhibiting a certain degree of uniform wear. The wear morphology inversion module 32 performs a complete inversion calculation every 100ms, retaining the previous output parameter values ​​between two inversions.

[0047] This embodiment achieves online inversion of the non-uniform wear morphology (wear azimuth angle, wear curvature radius, and central depression depth) of the grinding head without adding dedicated wear morphology measurement hardware by physically fusing force-torque frequency domain characteristics, acoustic emission resonance peak frequency shift, and temperature gradient direction. This upgrades the controller from "only knowing that wear exists" to "knowing the specific form of wear," providing an information basis for subsequent targeted compensation.

[0048] Reference Figure 4 The heterogeneous parameter compensation control module 33 is used to generate heterogeneous compensation commands based on the non-uniform wear morphology characteristic parameters output by the wear morphology inversion module 32, and control the grinding execution unit 10 to perform corresponding compensation actions. The heterogeneous parameter compensation control module 33 includes an attitude deviation angle compensation unit 331, an active micro-vibration modulation unit 332, and a speed feed linkage unit 333.

[0049] The attitude deviation compensation unit 331 generates a dynamic yaw angle command based on the wear azimuth angle and outputs it to the robotic arm 12 to guide the center of contact pressure towards the area of ​​slight wear. Its specific compensation logic is as follows: the yaw direction is set to the opposite direction of the wear azimuth angle (i.e., wear azimuth angle + 180°), and the yaw amplitude is proportional to the severity of wear. The severity of wear is quantified by the wear curvature radius: when the wear curvature radius is greater than 40mm, wear is considered slight, and the yaw amplitude is set to 0.5°; when the wear curvature radius is between 20mm and 40mm, the yaw amplitude is set to 1.0°; and when the wear curvature radius is less than 20mm, the yaw amplitude is set to 1.5°. For example, when the inversion output wear azimuth angle is 30° and the wear curvature radius is 18mm, the attitude angle compensation unit 331 generates a dynamic yaw angle command that causes the grinding head 11 to yaw 1.5° towards the 210° azimuth (30°+180°), thereby shifting the contact pressure center of the grinding head 11 towards the 210° azimuth (i.e. the side with slight wear) to compensate for the contact pressure attenuation caused by wear at the 30° azimuth.

[0050] The active micro-vibration modulation unit 332 is used to generate axial micro-vibration parameter commands based on the center depression depth and apply them to the spindle to homogenize the contact pressure distribution through dynamic impact. Its specific compensation logic is as follows: when the inverted center depression depth is greater than a preset threshold (3μm in this embodiment), active micro-vibration modulation is activated. The vibration frequency f_v is set to be a non-integer ratio to the spindle rotation frequency f_s; in this embodiment, f_v = 2.37 × f_s. When the spindle speed is 6000 rpm, the rotation frequency f_s = 100 Hz, and the vibration frequency f_v = 237 Hz. This non-integer ratio setting has significant technical implications: if the vibration frequency is an integer multiple of the spindle rotation frequency, the phase relationship of the vibration is fixed during each rotation, resulting in periodic vibration stripes on the workpiece surface; however, by using a non-integer ratio, the vibration phase of adjacent rotations continuously changes, achieving a homogenized distribution of contact pressure through time averaging and avoiding the generation of periodic grinding marks. The vibration amplitude A_v is dynamically calculated based on the central indentation depth D. In this embodiment, a linear relationship A_v = k × D is adopted, where k is a proportionality coefficient determined by a combination of the elastic modulus of the grinding head 11 and the hardness of the workpiece material. The typical value range is 0.02 to 0.05. When D = 5 μm and k = 0.03, the vibration amplitude is 0.15 μm. The active micro-vibration modulation unit 332 converts the above frequency and amplitude parameters into a drive signal and outputs it to the piezoelectric actuator 13 integrated in the spindle to generate axial micro-amplitude vibration.

[0051] The spindle speed feed linkage unit 333 is used to dynamically adjust the ratio of spindle speed to feed rate based on the wear radius of curvature. Its specific linkage logic is as follows: the ratio of spindle speed to feed rate, R_vf, is inversely proportional to the wear radius of curvature, R. When the grinding head 11 is in a normal state (with a large wear radius of curvature, such as R ≥ 50mm), R_vf takes the baseline value; when the wear radius of curvature decreases (indicating sharper wear and increased local contact stress), the R_vf value is increased, i.e., the spindle speed is appropriately increased or the feed rate is decreased, to reduce the amount of material cut per pass and prevent scratches and over-grinding of the workpiece surface. In this embodiment, R_vf = R_vf0 × (R0 / R), where R_vf0 is the baseline spindle speed feed ratio, R0 is the initial wear radius of curvature, and R is the current wear radius of curvature. For example, when R decreases from the initial 50mm to 25mm, R_vf automatically increases to twice its original value, thereby achieving adaptive matching of cutting parameters under changing wear conditions.

[0052] By setting up the heterogeneous parameter compensation control module 33, the existing technology breaks away from the conventional thinking of compensating for wear solely through a single isomorphic parameter such as "increasing total pressure" or "overall feed." It proposes a heterogeneous compensation strategy of "attitude guidance + vibration homogenization + parameter linkage," which can implement the most suitable compensation action for different wear patterns such as uneven wear, indentation, and sharp edges, achieving precise control of "compensation targeting the specific wear location." In particular, the non-integer ratio design of the axial micro-vibration frequency to the spindle rotation frequency eliminates the conditions for the generation of periodic vibration stripes from a mechanistic perspective.

[0053] The look-ahead allowance redistribution module 34 is used to determine the matching relationship between the severely worn areas of the grinding head 11 and the high removal amount areas of the workpiece on the subsequent grinding path based on the workpiece allowance distribution information on the grinding path obtained by the vision sensor 24 and the non-uniform wear morphology feature parameters, and adjust the grinding allowance distribution strategy of each area accordingly. Its specific working logic is as follows: Before grinding, the vision sensor 24 scans the workpiece to obtain the three-dimensional morphology of the workpiece surface and compares it with a standard CAD model to calculate the actual allowance distribution of each area on the grinding path. Based on this, the look-ahead allowance redistribution module 34 divides the end face of the grinding head 11 into several sectors according to the azimuth angle. Based on the wear azimuth angle and wear curvature radius output by the wear morphology inversion module 32, it marks the severely worn sectors on the end face of the grinding head 11. Then, it determines whether there are any high removal amount areas of the workpiece (i.e., areas with allowance exceeding a preset range) on the subsequent grinding path that will come into contact with the aforementioned severely worn sectors. If such a matching relationship is determined to exist, the forward-looking allowance reallocation module 34 will adjust the grinding strategy for that area: pre-allocate some of the grinding allowance from the high removal area to adjacent path segments that can be covered by the sharper sector of the grinding head 11, or appropriately increase the number of grinding passes in that area to reduce the removal amount per pass, thereby avoiding processing quality defects caused by processing high removal amounts in severely worn areas. This module feeds back the adjusted allowance reallocation strategy to the path planning unit of the robotic arm 12, completing the forward-looking adjustment before grinding execution.

[0054] By setting up the forward-looking allowance reallocation module 34, the wear morphology information is expanded from "real-time compensation" to "forward-looking path planning". Before grinding is performed, the matching conflict between the severely worn area and the high removal area of ​​the workpiece is predicted, and the allowance allocation strategy is actively adjusted, upgrading "passive response to wear" to "active use of wear information for global optimization".

[0055] The online self-repair control module 35 is used to trigger the pre-set online dresser 14 to perform in-situ trimming of the edge of the grinding head 11 according to a preset trimming path when the parameters output by the wear morphology inversion module 32 indicate that the grinding head 11 has experienced edge collapse. After the trimming is completed, the online dresser 14 controls the grinding execution unit 10 to continue grinding. In this embodiment, the online dresser 14 is a diamond dressing pen, which is fixedly installed on one side of the grinding station, and its tip position has been pre-calibrated in the coordinate system of the controller 30. The threshold for judging edge collapse is set as follows: the wear curvature radius is less than 10mm and the central depression depth is greater than 8μm. When the parameters output by the wear morphology inversion module 32 simultaneously meet the above two conditions, it indicates that the edge of the grinding head 11 has experienced significant edge collapse, and the grinding quality cannot be maintained by parameter compensation alone. At this time, the online self-repair control module 35 controls the robotic arm 12 to move the grinding head 11 to the online dresser 14 and trim the edge of the grinding head 11 according to a preset arc trimming path. The trimming path is a full circular arc movement along the edge of the grinding head 11. The trimming depth is set according to the depth of the central depression to ensure that the collapsed edge area is removed and the end face of the grinding head 11 is restored to a usable geometric state. After trimming, the reference resonance peak frequency f_r0 and the reference value of the physical correlation vector in the wear morphology inversion module 32 are automatically updated to adapt to the new state of the grinding head 11 after trimming. Then, the controller 30 controls the grinding execution unit 10 to return to the grinding position interrupted before trimming and continue to complete the remaining grinding work.

[0056] By setting the online self-repair control module 35, when the wear morphology parameters determine that the grinding head has experienced edge collapse, it can trigger in-situ repair online without disassembling the grinding head. After the repair is completed, the reference parameters are automatically updated and processing continues, turning the grinding head from a "disposable consumable" into a "quasi-tool that can be maintained online", which significantly extends the effective service life of the grinding head.

[0057] Reference Figure 5 The complete workflow of this application embodiment is as follows: S1. Preparation stage before grinding: Vision sensor 24 performs a 3D scan of the workpiece, acquires the actual shape of the workpiece surface and compares it with the standard CAD model, generating the grinding path and the initial allowance distribution information of each area. At the same time, the working status of each sensor is checked to ensure it is normal.

[0058] S2, Start Grinding: The controller 30 controls the robotic arm 12 to drive the grinding head 11 to grind the workpiece according to the planned path. The sensors of the detection unit 20 collect signals in real time at a set sampling frequency.

[0059] S3. Construction of contact state feature spectrum. The contact state feature spectrum construction module 31 acquires signals from force sensor 21, acoustic emission sensor 22 and infrared array sensor 23 in real time with a period of 10ms, extracts force-torque joint features, acoustic emission resonance peak frequency shift features and temperature gradient features according to the aforementioned method, and fuses them into a real-time contact state feature spectrum.

[0060] S4. Real-time wear morphology inversion: The wear morphology inversion module 32 uses a period of 100ms to solve the current wear azimuth angle, wear curvature radius and central depression depth of the grinding head 11 online according to the current real-time contact state feature spectrum and the weighted cosine similarity matching and weighted interpolation method.

[0061] S5. Heterogeneous Parameter Compensation Control: The heterogeneous parameter compensation control module 33 generates corresponding heterogeneous compensation commands in real time based on the non-uniform wear morphology characteristic parameters output in step S4: the attitude deviation angle compensation unit 331 generates a dynamic sway angle command based on the wear azimuth angle; the active micro-vibration modulation unit 332 generates an axial micro-vibration parameter command based on the center concavity depth; and the speed feed linkage unit 333 generates a speed feed ratio command based on the wear curvature radius. The above commands are output to the robotic arm 12, the piezoelectric actuator 13, and the spindle driver for execution, respectively.

[0062] S6. Forward-looking margin redistribution: The forward-looking margin redistribution module 34 determines whether there is a situation where a severely worn area matches a high removal amount area based on the margin distribution of the path to be polished and the current wear morphology characteristic parameters obtained by the vision sensor 24. If so, the grinding margin distribution of each area is adjusted according to the aforementioned strategy.

[0063] S7. Online Self-Repair Judgment: The online self-repair control module 35 monitors the wear morphology characteristic parameters and determines whether the edge collapse conditions of a wear curvature radius of less than 10mm and a center depression depth of greater than 8μm are met simultaneously. If met, the robotic arm 12 is controlled to move the grinding head 11 to the online dresser 14 for in-situ dressing. After dressing is completed, the reference parameters are updated and the robot returns to the grinding position to continue processing.

[0064] S8. Determine if the polishing task is completed: If not, return to step S3 and repeat; if completed, end the current polishing cycle.

[0065] To verify the technical effectiveness of this application, a comparative experiment was conducted under the same working conditions. The experimental conditions were: consistent grinding head model, workpiece material of aluminum alloy automotive cylinder head air passage, and target roughness Ra ≤ 0.8 μm. In the comparative scheme using traditional constant force control + fixed-cycle feed compensation, within a single grinding head life cycle (approximately 80 workpieces), the roughness Ra of the last 20 workpieces fluctuated between 0.52 μm and 1.23 μm, with a fluctuation amplitude of 0.71 μm, and 7 workpieces exhibited localized over-grinding stripes, resulting in non-conforming workpieces. However, using the scheme of this application, within the same life cycle, the roughness Ra of all 80 workpieces fluctuated between 0.68 μm and 0.92 μm, with a fluctuation amplitude of only 0.24 μm, and no non-conforming workpieces appeared. Furthermore, due to the effects of heterogeneous compensation and look-ahead margin redistribution, the scheme of this application can maintain grinding quality even in the severe wear stage, extending the effective service life of the grinding head by approximately 45% compared to the comparative scheme.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions are also within the scope of protection of the present invention. The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete or otherwise adjust the features in the various embodiments of the present invention as appropriate without conflict or creative effort, thereby obtaining different technical solutions that do not depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. A grinding device for processing automotive parts, characterized in that, include: The grinding execution unit has a grinding head and a robotic arm that drives the grinding head to perform multi-axis motion; The detection unit includes at least a force sensor for collecting the grinding contact force, an acoustic emission sensor for collecting acoustic emission signals, an infrared array sensor for collecting the temperature distribution in the grinding area, and a vision sensor for acquiring the spatial position of the accessory and the grinding head. A controller, connected to the detection unit and the grinding execution unit, the controller comprising: The contact state feature spectrum construction module is used to acquire signals from the force sensor, acoustic emission sensor and infrared array sensor in real time, extract force-torque joint features, acoustic emission resonance peak frequency shift features and temperature gradient features from them, and fuse the three into a real-time contact state feature spectrum; the real-time contact state feature spectrum includes a torque fluctuation frequency domain energy distribution vector, resonance peak frequency shift scalar and temperature gradient direction vector. The wear morphology inversion module is used to invert the non-uniform wear morphology feature parameters of the grinding head online based on the real-time contact state feature spectrum. The non-uniform wear morphology feature parameters include at least the wear azimuth angle, wear curvature radius and center depression depth. The heterogeneous parameter compensation control module is used to generate heterogeneous compensation instructions based on the non-uniform wear morphology characteristic parameters. The heterogeneous compensation instructions include at least one of the following: grinding head attitude deviation angle instruction, axial micro-vibration parameter instruction, and spindle speed to feed rate ratio instruction. The module also controls the grinding execution unit to perform the corresponding compensation actions.

2. The grinding device for processing automotive parts according to claim 1, characterized in that, The contact state feature spectrum construction module includes: The force-torque joint feature extraction unit is used to extract torque fluctuation features related to the rotation frequency of the grinding head based on the normal force and torque signals collected by the force sensor. The acoustic emission feature extraction unit is used to perform spectral analysis on the acoustic emission signal, lock the resonance peak of a specific frequency band, and monitor its frequency offset in real time. The temperature gradient feature extraction unit is used to calculate the temperature gradient vector along the front and rear edges of the grinding contact area based on the temperature distribution data collected by the infrared linear array sensor.

3. The grinding device for processing automotive parts according to claim 2, characterized in that, The force-torque joint feature extraction unit is specifically used to: perform spectral analysis on the real-time torque signal, extract the energy distribution corresponding to the fundamental frequency and harmonics of the grinding head rotation, and use the energy distribution as a force-torque joint feature characterizing the direction and degree of uneven wear.

4. The grinding device for processing automotive parts according to claim 2, characterized in that, The acoustic emission feature extraction unit is specifically used to: continuously analyze acoustic emission signals during the grinding process, identify the frequency shift of a specific resonance peak caused by changes in the contact arc length and contact stiffness between the grinding head and the workpiece, and use this frequency shift as the acoustic emission resonance peak frequency shift feature characterizing the absolute magnitude of wear.

5. A grinding device for processing automotive parts according to claim 2, characterized in that, The temperature gradient feature extraction unit is specifically used to: acquire the temperature field at the front and rear edges of the grinding contact point using an infrared linear array sensor, calculate the direction and magnitude of the maximum temperature gradient, and use the direction of the gradient vector as the temperature gradient feature pointing to the area of ​​most severe wear.

6. A grinding device for processing automotive parts according to claim 1, characterized in that, The wear morphology inversion module includes: The calibration storage unit stores the physical correlation vector between the pre-calibrated basic wear morphology library and the corresponding contact state feature spectrum; The online solution unit is used to decompose the contact state feature spectrum acquired in real time into the physical correlation vector space, and to calculate the wear azimuth angle, wear curvature radius and center depression depth of the current grinding head in real time through a weighted fusion algorithm.

7. A grinding device for processing automotive parts according to claim 1, characterized in that, The heterogeneous parameter compensation control module includes: The attitude deviation angle compensation unit is used to generate a dynamic yaw angle command based on the wear azimuth angle and output it to the robotic arm to guide the center of contact pressure toward the area of ​​slight wear. An active micro-vibration modulation unit is used to generate axial micro-vibration parameter commands based on the depth of the central depression and apply them to the main shaft to homogenize the contact pressure distribution through dynamic impact. The spindle speed feed linkage unit is used to dynamically adjust the ratio of spindle speed to feed speed according to the wear radius of curvature.

8. A grinding device for processing automotive parts according to claim 7, characterized in that, The grinding execution unit integrates a piezoelectric actuator within its spindle; the active micro-vibration modulation unit is specifically used to: calculate the vibration amplitude and frequency based on the depth of the central depression, and drive the piezoelectric actuator to generate axial micro-amplitude vibrations that are not integer ratios to the spindle rotation frequency.

9. A grinding device for processing automotive parts according to claim 1, characterized in that, The controller also includes: The forward-looking allowance redistribution module is used to determine the matching relationship between the severely worn area of ​​the grinding head and the high removal amount area of ​​the workpiece on the subsequent grinding path based on the workpiece allowance distribution information on the grinding path obtained by the vision sensor and the non-uniform wear morphology feature parameters, and adjust the grinding allowance allocation strategy of each area accordingly.

10. A grinding device for processing automotive parts according to claim 1, characterized in that, Also includes: Online trimmer and online self-healing control module; The online self-repair control module is used to trigger the online trimmer to trim the edge of the grinding head in situ according to a preset trimming path when the parameters output by the wear morphology inversion module indicate that the grinding head has experienced edge collapse, and to control the grinding execution unit to continue grinding work after the trimming is completed.

Images