Grinding force online detection control device and control method

Abstract

The invention discloses a grinding force online detection control device and method, and relates to the technical field of precision grinding machining force measurement, a spring chuck is arranged in a containing cavity, and the outer wall of a detection shaft section is a tapered conical surface; the threaded gland is connected with the connecting shaft section to axially limit the spring chuck, the mounting handle is provided with a first conical cavity, the threaded gland is provided with an avoiding hole, the piezoelectric sensors are mounted on the detection shaft section, the grinding wheel spindle acts on the piezoelectric sensors through the detection shaft section, and the piezoelectric sensors are used for detecting tangential force, normal force and axial force. The signal amplification processing module can convert charge signals output by the piezoelectric sensor into digital signals and output the digital signals to the terminal, and the terminal can control logic to control the grinding head to maintain constant-force grinding according to a preset grinding force target value. According to the method, all-directional real-time detection and control over the grinding force are achieved, the accuracy and the real-time performance are improved, and a new control method is provided for low-damage machining of hard and brittle materials.

Description

Technical Field

[0001] The present application relates to the field of precision grinding force measurement technology, and in particular to an on-line detection and control device for grinding force. Meanwhile, the present application also relates to a grinding force control method. Background Art

[0002] In the field of precision manufacturing, hard and brittle materials (such as silicon carbide, silicon nitride, optical glass, etc.) are widely used in high-end fields such as aerospace, semiconductor devices, and optical components due to their high hardness, high temperature resistance, and excellent chemical stability. During the grinding process, hard and brittle materials usually have surface damage in the form of cracks, pits, plastic scratches, etc., as well as subsurface damage in the form of subsurface cracks, which in turn affects the grinding surface quality and product performance.

[0003] When grinding hard and brittle materials, the size of the grinding force has an important influence on the processing quality. Excessive grinding force can cause serious damage such as cracks and pits on the material surface. At the same time, subsurface cracks and microscopic damage will deepen, weakening the mechanical properties of the material. In addition, although a small grinding force means a smaller removal amount and it is easier to obtain a smooth surface, too small a grinding force may lead to low processing efficiency, and the grinding wheel is prone to passivation, increasing the dressing cost. It can be seen that controlling the grinding force is the key to ensuring processing quality. The grinding force of the grinding wheel can be divided into the normal force F n With tangential force F t , a larger tangential force can increase the material removal rate, but it may cause surface crack damage and reduce surface finish and precision. A larger normal force aggravates the stress concentration on the subsurface, causing subsurface cracks and plastic deformation, weakening the mechanical properties of the material, and accelerating the wear of the grinding wheel. Therefore, controlling the balance between the tangential force and the normal force during grinding is crucial to optimizing the grinding quality and efficiency. According to the formula for the maximum undeformed chip thickness, it can be concluded that by increasing the grinding wheel speed and reducing the grinding feed rate, the maximum undeformed chip thickness of grinding can be reduced, thereby ensuring that the number of cracks and pits on the grinding surface is reduced.

[0004] According to the theory of median / lateral crack system, the grinding normal force has the main influence on the subsurface cracks of the material. The subsurface cracks penetrate deep into the material and reduce the material performance. The tangential force will increase the material breakage and improve the grinding efficiency to a certain extent. However, when the tangential force is too large, it will accelerate the dislocation and slip movement, and it is easy to produce dislocation accumulation at the grain boundary and the second phase particles, which will form conditions for crack nucleation and accelerate damage. When the tangential force is too small, the dislocation activity is insufficient, and the energy may be released directly through the cracks, forming a large number of microcracks.

[0005] At present, there is little structural and theoretical support for the detection and control of grinding forces during the grinding process of hard and brittle materials in the existing technology, which has at least the following significant limitations: (1) Traditional grinding force detection mostly uses single-axis force sensors (such as strain gauge sensors) or indirect estimation methods (such as spindle current monitoring). Such solutions can only obtain force components in a single direction and cannot fully reflect the dynamic distribution of three-dimensional grinding forces. For example, some existing radial force detection devices can measure normal forces, but are sensitive to the coupling effect of tangential forces, resulting in high decoupling errors. In addition, the response frequency of strain gauge sensors is usually lower than 1kHz, which makes it difficult to capture microsecond force fluctuations during the grinding process, causing control signal lag and failing to meet ultra-precision machining requirements.

[0006] (2) Existing grinding force control is mostly based on fixed-parameter PID algorithms, whose proportional, integral, and differential coefficients need to be repeatedly debugged based on manual experience. However, the processing window for hard and brittle materials is extremely narrow. Once the PID parameters are solidified, it is difficult to adapt to dynamic changes such as grinding wheel wear and material batch differences. Relevant literature points out that after one hour of continuous processing, the grinding force drift caused by grinding wheel passivation in traditional PID control can reach 30%, forcing frequent shutdowns for grinding wheel dressing, which greatly reduces processing efficiency.

[0007] (3) In surface grinding, although the axial force is relatively small, its influence is often ignored by existing technologies. In actual working conditions, the runout of the grinding wheel spindle or the tilt of the workpiece clamping will cause abnormal axial force. For example, in actual applications, for the sapphire substrate grinding process, a sudden increase in axial force may cause the workpiece to "warp" or the grinding wheel axis system to resonate, exacerbating the surface waviness. However, due to the limitations of sensor layout, traditional devices cannot effectively monitor the axial force and can only detect defects through post-workpiece inspection, resulting in waste of raw materials. Summary of the invention

[0008] The present application provides an online grinding force detection and control device to improve or solve the technical problems of inaccurate grinding force measurement, poor real-time performance, and difficulty in precise control in the detection and control of grinding force during the grinding process of hard and brittle materials.

[0009] The technical solution adopted in this application is: A grinding force online detection and control device is used to detect and control the grinding force of a grinding head during the grinding process of a hard and brittle material surface, wherein the grinding force includes tangential force, normal force and axial force, and comprises a threaded gland, a mounting handle, a plurality of piezoelectric sensors, a spring chuck with a hollow interior, and a signal amplification and processing module, wherein the threaded gland and the mounting handle are connected and cooperated to form an accommodating cavity for accommodating the spring chuck, wherein the spring chuck includes a connecting shaft section and a detection shaft section in sequence along the axial direction, and the outer wall of the detection shaft section is a conical surface; along the direction from the connecting shaft section to the detection shaft section, the detection shaft section gradually shrinks inwardly; the threaded gland limits the axial position of the spring chuck, and the mounting handle A first conical cavity is provided which is adapted to the outer wall of the detection shaft section, and the threaded cover is provided with an avoidance hole corresponding to the center of the spring chuck, and the grinding wheel spindle can be inserted into the spring chuck through the avoidance hole and clamped, and the multiple piezoelectric sensors are installed on the detection shaft section, so that the grinding wheel spindle acts on the multiple piezoelectric sensors through the detection shaft section, and the multiple piezoelectric sensors detect the tangential force, the normal force and the axial force, and the signal amplification and processing module can convert the charge signal output by the piezoelectric sensor into a digital signal and output it to the terminal, and the terminal can control the grinding head to maintain constant force grinding with a preset grinding force target value according to a preset control logic.

[0010] In the technical scheme, the spring chuck is confined in the accommodating cavity surrounded by the threaded cover and the mounting handle. The grinding wheel spindle can be inserted into the spring chuck through the avoidance hole of the threaded cover and clamped by the spring chuck. Then, the conical detection shaft section of the spring chuck cooperates with the first conical cavity of the mounting handle. The spring chuck reduces vibration interference in the grinding process, improves the signal acquisition accuracy of the piezoelectric sensor, and realizes accurate force transmission of the grinding wheel spindle. Moreover, the rigid force transmission path between the grinding wheel spindle and the piezoelectric sensor effectively overcomes the defect of large decoupling error of traditional single-axis sensors. The piezoelectric sensors are distributed on the detection shaft section, and directly sense Knowing the tangential force, normal force and axial force of the grinding force, the high sensitivity and fast response characteristics of the piezoelectric sensor can capture the slight changes in the grinding force in real time, provide accurate input signals for constant force grinding control, and solve the control delay problem caused by the lag of traditional strain gauge sensors. Combined with the signal amplification processing module, the charge signal is converted into a digital signal, realizing the real-time detection of the tangential, normal, axial and other dimensions of the grinding force, building a full-dimensional data foundation, providing high-precision input for subsequent control, and through the detection of grinding axial force, avoiding workpiece warping or grinding wheel resonance caused by the lack of axial force monitoring. In this solution, through the integrated structure of the spring chuck, threaded cover and mounting handle, and the arrangement of the piezoelectric sensor, high rigidity, vibration resistance and convenient installation are achieved, which not only improves the reliability and adaptability of the device, but also reduces the manufacturing and maintenance costs. By setting the control logic at the terminal, it is enabled to adjust at least one of the feed speed, grinding feed amount, and grinding wheel speed to achieve constant force grinding. Then, by accurately controlling the grinding force in real time, the grinding normal force, tangential force, and axial force are stabilized within the material damage threshold, directly inhibiting the propagation of sub-surface cracks and reducing surface roughness, while maximizing the material removal rate while reducing abnormal wear of the grinding wheel.

[0011] The grinding force online detection and control device includes four piezoelectric sensors, which are evenly distributed along the circumference of the detection shaft segment, wherein the force measuring ends of three piezoelectric sensors are arranged along the radial direction of the detection shaft segment for detecting the tangential force and the normal force, and the force measuring end of another piezoelectric sensor is arranged along the axial direction of the detection shaft segment for detecting the axial force.

[0012] In this technical solution, four piezoelectric sensors are arranged in three radial directions and one axial direction, and the three radial piezoelectric sensors accurately decouple the tangential force and the normal force through vector synthesis, taking into account accuracy, redundancy and cost. Although in theory some functions can be achieved with fewer sensors, in a dynamic grinding environment, the three-radial piezoelectric sensor solution can significantly improve reliability and avoid data distortion caused by single-point failure or directional coupling. An axial piezoelectric sensor is dedicated to detecting tiny but dangerous axial forces, which can provide early warning of workpiece clamping tilt during surface grinding and reduce scrap rate. The above design not only avoids excessive redundancy in the number of sensors, but also ensures the comprehensiveness of three-dimensional force detection, thereby improving measurement efficiency and accuracy.

[0013] The detection shaft section is provided with sensor mounting countersunk holes which are matched one by one with the four piezoelectric sensors. The openings of the sensor mounting countersunk holes are covered by a cover plate, which presses and limits the piezoelectric sensors. The outer surface of the cover plate smoothly transitions with the conical surface of the detection shaft section to form a continuous curved surface. The outer surface of the cover plate has a consistent generatrix curvature with the conical surface to maintain the integrity of the conical surface.

[0014] In this technical solution, the piezoelectric sensor is rigidly fixed in the detection shaft segment through the coordinated design of the sensor mounting countersunk hole and the cover plate. The outer surface of the cover plate and the conical surface have a smooth transition, thereby maintaining the structural integrity and mechanical properties of the conical surface of the detection shaft segment, reducing stress concentration and pneumatic noise during high-speed rotation, extending the fatigue life of the spring chuck, and preventing cutting fluid from invading the area where the piezoelectric sensor is located.

[0015] A wire hole is provided in the sensor mounting countersunk hole, and the wire hole passes through the end of the detection shaft segment away from the connecting shaft segment. The interior of the mounting handle is hollow, and the piezoelectric sensor is connected to a data transmission line, and the data transmission line passes through the wire hole and the interior of the mounting handle.

[0016] In this technical solution, the wire hole and the internal cavity of the mounting handle form a closed wiring channel, and the data transmission line is led out through the end of the detection shaft segment and the internal cavity of the mounting handle, which optimizes the wiring path, replaces traditional exposed cables (which are susceptible to mechanical damage), avoids external cable entanglement or wear, improves device reliability, and simplifies the assembly and maintenance process.

[0017] The detection shaft section and the connecting shaft section are separated by a circumferentially extending limiting groove, the inner wall of the threaded cover is provided with a limiting protrusion adapted to the limiting groove, and the limiting groove and the limiting protrusion cooperate to limit the axial position of the spring chuck; the outer wall of the connecting shaft section is a conical surface, the conical surface of the connecting shaft section and the conical surface of the detection shaft section have opposite taper directions, and the threaded cover is provided with a second conical cavity adapted to the outer wall of the connecting shaft section.

[0018] In this technical solution, the connecting shaft section cooperates with the threaded gland through the limiting groove and the limiting protrusion to realize reliable axial limiting of the spring collet. Moreover, the contraction and expansion characteristics of the spring collet are utilized to facilitate the cooperation of the limiting groove and the limiting protrusion with the threaded gland. The taper directions of the connecting shaft section and the detection shaft section are opposite, forming a bidirectional self-centering locking structure, which helps to reduce the coaxiality error of the spring collet assembly, avoid axial force abnormality caused by the runout of the grinding wheel spindle, enhance the connection stability between the spring collet and the grinding wheel spindle, reduce vibration offset during grinding, improve force transmission symmetry, and help to suppress the plastic deformation of the spring collet in heavy-load grinding and extend its service life.

[0019] The spring chuck is provided with a plurality of first elastic grooves extending from the connecting shaft section toward the detecting shaft section and a plurality of second elastic grooves extending from the detecting shaft section toward the connecting shaft section, and the first elastic grooves and the second elastic grooves are alternately arranged along the circumference of the spring chuck.

[0020] In the present technical solution, the spring chuck is provided with alternately distributed first elastic grooves and second elastic grooves, which significantly improves the radial elastic deformation capacity of the spring chuck, ensures that the grinding wheel spindle is subjected to uniform force, and avoids plastic deformation or fatigue fracture caused by excessive local stress.

[0021] Preferably, the threaded gland is provided with an internal thread, and the mounting handle is provided with an external thread, and the threaded gland and the mounting handle are detachably connected by screwing the internal thread and the external thread together.

[0022] In this technical solution, the threaded cover and the mounting handle are detachably assembled through a threaded connection, which is convenient for quick maintenance or replacement of spring chucks or sensor modules of different specifications, adapting to diverse workpiece processing needs while reducing maintenance costs.

[0023] Another object of the present application is to propose a grinding force control method, which is applied to the grinding force online detection and control device as described above, including: setting the initial grinding wheel speed, initial feed speed and initial grinding feed amount, and setting the target value of the grinding force based on the critical crack pressure of the hard and brittle material being ground; the piezoelectric sensor collects the grinding force signal in real time, and the grinding force signal is processed by the signal amplification processing module and output to the PID feedback control system, and the PID feedback control system adopts a quasi-ductile critical adjustment model according to the deviation between the actual grinding force detection value and the target value, and dynamically coordinates and controls the grinding wheel speed and / or feed speed in real time to maintain constant force grinding at the target value; during the control process of the PID feedback control system, sensor signals, processing parameters and control output data are collected to construct a training sample set for the feedforward deep learning network; supervised learning is used to train the deep learning model to learn the mapping relationship from sensor input to control signal output; the proportion of feedforward control and feedback control is dynamically adjusted based on the control strategy, and the control weight is gradually transitioned from feedback control to feedforward control.

[0024] The quasi-ductility critical adjustment model specifically includes: a) Construct the composite control quantity equation: , where p is the dynamic adjustment factor, p is 0.1-0.5, τ is the system response time constant, is the composite control quantity, is the actual detection value of grinding force, is the target value of grinding force, and the real-time grinding force error is mapped to the adjustment amount of grinding wheel speed and feed speed through the dynamic adjustment factor p and system response time constant 𝜏; b) Implement a secondary error response mechanism: i) When the absolute value of the grinding force error is greater than 15% of the target value, adjust the feed speed; ii) When the absolute value of the grinding force error is less than or equal to 15% of the target value, adjust the grinding wheel speed.

[0025] In a preferred embodiment, in the quasi-ductility critical adjustment model, a dynamic limit mechanism can also be used to constrain the parameter adjustment range, such as limiting the feed speed to 1.2 times the initial value, the speed fluctuation does not exceed ±20%, and limiting the single adjustment range (≤5% of the current value). hour, ;when hour, ; Intelligent adjustment factor design is used for adjustment factors, , in order to achieve rapid adjustment when the error is large and fine control when the error is small, among which, is the feed speed, is the initial feed speed, is the maximum adjustment of feed speed, is the grinding wheel speed, is the initial grinding wheel speed, is the maximum adjustment of the grinding wheel speed, is the initial adjustment factor.

[0026] The control method further includes: the control strategy adopts a dynamic weighted control strategy, and the dynamic weighted control strategy specifically defines a feedforward control weight With feedback control weight , initially =0, =1; when the mean square error E between the feedforward control signal output by the deep learning model and the PID feedback signal is less than the set threshold, gradually increase and reduce , until =1, =0.

[0027] The control method further includes: optimizing the control action by using the deep deterministic policy gradient (DDPG) algorithm, generating a continuous control signal through the Actor-Critic network; designing a reward function, combining force error, action smoothness and processing efficiency to optimize the control strategy; and realizing a smooth transition from PID control to feedforward control through experience replay and target network soft update mechanism. The control system required for grinding force control consists of multiple modules, including an environment modeling module, an Actor-Critic network module, an experience replay module, a reward design module and a dynamic weighting module. The environment modeling module transforms the grinding force control problem into a Markov decision process in reinforcement learning, defining the state (sensor signal, processing parameters, force error, etc.), action (servo motor control signal) and reward function (combining force error, oscillation amplitude and action smoothness).

[0028] The technical effects that can be achieved by adopting the above-mentioned grinding force control method are as follows: a grinding force threshold setting method based on the critical crack pressure is combined with PID feedback control to adjust the grinding wheel parameters in real time, and a feedforward control model is constructed through deep learning to achieve the transition from experience dependence to intelligent prediction, solving the problems of response lag and parameter solidification in traditional control. Based on the composite control quantity equation: through the dynamic coupling of the dynamic adjustment factor p and the time constant τ, the grinding force error is mapped to the joint adjustment amount of the speed and feed, solving the response lag problem caused by the traditional PID parameter solidification and shortening the system response time. When the error is large, the feed speed is adjusted first, and when the error is small, the speed and feed amount are adjusted jointly, taking into account rapid correction and stability. The control weight is gradually transitioned through a dynamic weighting strategy. In the initial stage, PID feedback is relied on to ensure stability. As the data accumulates, the proportion of feedforward control is enhanced, and finally low-latency, high-precision constant force grinding is achieved, balancing the system robustness and response speed. The DDPG algorithm is introduced to optimize the control action, and the reward function is designed in combination with force error, action smoothness and processing efficiency. Through the Actor-Critic network and experience replay mechanism, the system can autonomously adapt to complex working conditions such as grinding wheel wear and material changes, thereby improving the generalization ability of control. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 Assembly of the grinding force online detection and control device provided in the embodiment of the present application Figure 1 ; Figure 2 Assembly of the grinding force online detection and control device provided in the embodiment of the present application Figure 2 ; Figure 3 A cross-sectional view of a grinding force online detection and control device provided in an embodiment of the present application; Figure 4 A schematic diagram of the cooperation between the spring clamp and the cover plate provided in the embodiment of the present application; Figure 5 A cross-sectional view of a threaded gland provided in an embodiment of the present application; Figure 6 A cross-sectional view of a mounting handle provided in an embodiment of the present application; Figure 7 The structure of the spring chuck provided in the embodiment of the present application is schematically shown. Figure 1 ; Figure 8 The structure of the spring chuck provided in the embodiment of the present application is schematically shown. Figure 2 ; Fig. 9A cross-sectional view of the cooperation between the spring chuck and the piezoelectric sensor provided in an embodiment of the present application; Fig.10 A control flow chart of the grinding force control method provided in an embodiment of the present application; Fig.11 This is a DDPG reinforcement learning control flow chart provided in an embodiment of the present application.

[0030] List of parts and reference numerals: 1 threaded gland, 11 avoidance hole, 12 limiting protrusion, 13 second tapered cavity, 14 internal thread; 2 mounting handle, 21 first tapered cavity, 22 external thread; 3. Piezoelectric sensor; 4 spring chuck, 41 connecting shaft section, 42 detecting shaft section, 43 sensor mounting countersunk hole, 44 wire passing hole, 45 limiting groove, 46 first elastic groove, 47 second elastic groove; 5 Cover. DETAILED DESCRIPTION

[0031] In order to more clearly illustrate the overall concept of the present application, a detailed description is given below in an illustrative manner in conjunction with the accompanying drawings.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present application is not limited to the specific embodiments disclosed below.

[0033] In addition, in the description of the present application, it should be understood that the terms "upper", "lower", "top", "bottom", "inside", "outside", "axial", "radial", "circumferential", "lateral", "longitudinal", etc., indicating orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0034] In this application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, or a communication; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0035] In the present application, unless otherwise clearly specified and limited, a first feature "above" or "below" a second feature may be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples.

[0036] The present application provides an online detection and control device for grinding force. For the convenience of explanation and understanding, the following contents provided in the present application are all described on the basis of the illustrated product structure. Of course, it can be understood by those skilled in the art that the above structure is only used as a specific example and schematic description, and cannot constitute a specific limitation on the technical solution provided in the present application.

[0037] The grinding force online detection and control device of the present application is used to detect and control the grinding force of the grinding head during the grinding process of the surface of hard and brittle materials (such as silicon carbide, silicon nitride, optical glass, etc.). The grinding force includes tangential force, normal force and axial force. In terms of overall composition, Figures 1 to 9 As shown, it includes a threaded gland 1, a mounting handle 2, a plurality of piezoelectric sensors 3, a spring chuck 4 with a hollow interior, and a signal amplification processing module. The mounting handle 2 can be connected to the main shaft of a machine tool. The threaded gland 1 and the mounting handle 2 are connected and cooperated to form a receiving cavity for receiving the spring chuck 4. The spring chuck 4 includes a connecting shaft section 41 and a detecting shaft section 42 in sequence along the axial direction. The outer wall of the detecting shaft section 42 is a conical surface. The detecting shaft section 42 gradually shrinks inwardly along the direction from the connecting shaft section 41 to the detecting shaft section 42. The threaded gland 1 is connected to the connecting shaft section 41 to limit the axial position of the spring chuck 4. The mounting handle 2 is provided with a connection with the detecting shaft section 42. The outer wall of the first conical cavity 21 is adapted to the outer wall, the threaded cover 1 is provided with an avoidance hole 11 corresponding to the center of the spring chuck 4, the grinding wheel spindle can be inserted into the spring chuck 4 through the avoidance hole 11 and clamped, a plurality of piezoelectric sensors 3 are installed on the detection shaft section 42, the grinding wheel spindle acts on the plurality of piezoelectric sensors 3 through the detection shaft section 42, the plurality of piezoelectric sensors 3 detect the tangential force, normal force and axial force, the signal amplification processing module can convert the charge signal output by the piezoelectric sensor 3 into a digital signal and output it to the terminal, and the terminal can control the grinding head to maintain constant force grinding with a preset grinding force target value according to the preset control logic.

[0038] It should be noted that the present application does not specifically limit the connection and matching method between the threaded gland 1 and the mounting handle 2. The two can be connected by non-detachable methods such as ultrasonic welding. As a preferred solution, the two can also be detachably connected by means of snap connection, threaded connection, etc. In a preferred embodiment, Figure 2 , Figure 3 , Figure 5 and Figure 6 As shown, the threaded cover 1 is provided with an internal thread 14, and the mounting handle 2 is provided with an external thread 22. The threaded cover 1 and the mounting handle 2 are detachably connected by screwing the internal thread 14 and the external thread 22 together, thereby facilitating the rapid replacement of spring chucks 4 or sensor modules of different specifications, adapting to diverse workpiece processing requirements, and reducing maintenance costs.

[0039] In addition, the present application does not specifically limit the manner in which the threaded gland 1 axially limits the spring chuck 4. In a preferred embodiment, Figure 3 , Figure 5 and Figure 7 As shown, the detection shaft section 42 and the connecting shaft section 41 are separated by a circumferentially extending limiting groove 45, and the inner wall of the threaded gland 1 is provided with a limiting protrusion 12 adapted to the limiting groove 45. The limiting groove 45 cooperates with the limiting protrusion 12 to reliably limit the axial direction of the spring collet 4. Moreover, by utilizing the contraction and expansion characteristics of the spring collet 4, it is also convenient for the limiting groove 45 and the limiting protrusion 12 to cooperate and disengage with the threaded gland 1. In other embodiments, the threaded gland 1 can also use other methods to limit the axial direction of the spring collet 4, for example, a screw-in groove is provided on the spring collet 4, and a screw-in protrusion is provided on the threaded gland 1, so that the spring collet 4 and the threaded gland 1 are rotationally matched through the screw-in groove and the screw-in protrusion.

[0040] In the present technical solution, the spring chuck 4 is confined in the accommodating cavity surrounded by the threaded cover 1 and the mounting handle 2. The grinding wheel spindle can be inserted into the spring chuck 4 through the avoidance hole 11 of the threaded cover 1 and clamped by the spring chuck 4. Then, the conical detection shaft section 42 of the spring chuck 4 cooperates with the first conical cavity 21 of the mounting handle 2. The spring chuck 4 reduces the vibration interference in the grinding process, improves the signal acquisition accuracy of the piezoelectric sensor 3, and realizes the precise force transmission of the grinding wheel spindle. Moreover, the rigid force transmission path between the grinding wheel spindle and the piezoelectric sensor 3 effectively overcomes the defect of large decoupling error of the traditional single-axis sensor; the piezoelectric sensor 3 is distributed in the detection The measuring shaft section 42 directly senses the tangential force, normal force and axial force of the grinding force. Through the high sensitivity and fast response characteristics of the piezoelectric sensor 3, it can capture the slight changes of the grinding force in real time, provide accurate input signals for constant force grinding control, and solve the control delay problem caused by the hysteresis of the traditional strain gauge sensor. Combined with the signal amplification processing module, the charge signal is converted into a digital signal, and the tangential, normal, axial and other multi-dimensional real-time detection of the grinding force is realized, and a full-dimensional data foundation is constructed to provide high-precision input for subsequent control. By detecting the grinding axial force, the workpiece warping or grinding wheel resonance caused by the lack of axial force monitoring is avoided. In this scheme, through the integrated structure of the spring chuck 4, the threaded gland 1 and the mounting handle 2, and the arrangement of the piezoelectric sensor 3, high rigidity, vibration resistance and convenient installation are achieved, which not only improves the reliability and adaptability of the device, but also reduces the manufacturing and maintenance costs. Specifically, the signal amplification processing module includes a signal conditioning module and a data transmission / acquisition / processing and analysis module. The signal amplification processing module converts the charge signal output by the piezoelectric sensor 3 into a digital signal and transmits it to the terminal. By setting the control logic at the terminal, the constant force grinding is achieved by adjusting the feed speed, grinding feed amount, grinding wheel speed, etc., and then the grinding normal force, tangential force, and axial force are stabilized within the material damage threshold by real-time and precise control of the grinding force, directly inhibiting the subsurface crack extension and reducing the surface roughness, while maximizing the material removal rate on the basis of reducing abnormal wear of the grinding wheel. There is no limitation on the control logic set at the terminal. For example, PID control logic, PID-feedforward composite control logic, PID feedback + feedforward compensation + dynamic limit three-level control logic, etc. can be used. The control logic of the specific example is specifically expanded in the control method.

[0041] Regarding the structure of the connecting shaft section 41, in a preferred embodiment, as Figure 3 , Figure 5 and Figure 7As shown, the outer wall of the connecting shaft segment 41 is a conical surface, and the conical surface of the connecting shaft segment 41 and the conical surface of the detection shaft segment 42 have opposite taper directions, and the threaded gland 1 is provided with a second conical cavity 13 adapted to the outer wall of the connecting shaft segment 41. It can be understood by those skilled in the art that the taper directions of the connecting shaft segment 41 and the detection shaft segment 42 are opposite, forming a bidirectional self-centering locking structure, which helps to reduce the coaxiality error of the spring chuck 4 during assembly, avoids abnormal axial force caused by the runout of the grinding wheel spindle, enhances the connection stability of the spring chuck 4 and the grinding wheel spindle, reduces vibration offset during grinding, improves the symmetry of force transmission, and helps to suppress the plastic deformation of the spring chuck 4 during heavy-load grinding and prolongs its service life.

[0042] In order to realize that the spring clamp 4 has a reliable and stable elastic deformation capability, in a preferred embodiment, Figure 7 As shown, the spring chuck 4 is provided with a plurality of first elastic grooves 46 extending from the connecting shaft section 41 toward the detection shaft section 42 and a plurality of second elastic grooves 47 extending from the detection shaft section 42 toward the connecting shaft section 41, and the first elastic grooves 46 and the second elastic grooves 47 are alternately arranged along the circumference of the spring chuck 4. In the present technical scheme, the spring chuck 4 is provided with the first elastic grooves 46 and the second elastic grooves 47 which are alternately distributed, which significantly improves the radial elastic deformation ability of the spring chuck 4, ensures that the grinding wheel spindle is evenly stressed, and avoids plastic deformation or fatigue fracture caused by excessive local stress. Moreover, through the above design, the proportion of effective expansion or contraction of the spring chuck 4 is large, which can effectively reduce the volume, making it easy to operate, and the expansion or contraction amplitude of the detection shaft section 42 is larger, which is helpful for clamping the grinding wheel spindle in the hole.

[0043] In a preferred embodiment, the piezoelectric sensor 3 can be a piezoelectric ceramic sensor, which can produce a more significant charge change when subjected to external force. It can sensitively sense and convert tiny changes in physical quantities such as pressure and vibration into electrical signals, thereby accurately detecting weak external stimuli and responding to external forces in a very short time, and quickly converting mechanical signals into electrical signals. This fast response characteristic makes it suitable for dynamic measurement of grinding force, capture of rapidly changing signals, and the like.

[0044] Regarding the number and installation method of the piezoelectric sensors 3, in a preferred embodiment, as shown in FIG. Figure 3 , Figure 4 and Fig. 9As shown, the grinding force online detection control device includes four piezoelectric sensors 3, and the four piezoelectric sensors 3 are evenly distributed along the circumference of the detection shaft segment 42, wherein the force measuring ends of three piezoelectric sensors 3 are arranged along the radial direction of the detection shaft segment 42 for detecting tangential force and normal force, and the force measuring end of another piezoelectric sensor 3 is arranged along the axial direction of the detection shaft segment 42 for detecting axial force. The four piezoelectric sensors 3 are arranged in three radial directions and one axial direction, wherein the three radial piezoelectric sensors can accurately decouple the tangential force and the normal force by vector synthesis, taking into account accuracy, redundancy and cost. Although in theory some functions can be achieved with fewer sensors, in a dynamic grinding environment, the three radial piezoelectric sensor solution can significantly improve reliability and avoid data distortion caused by single point failure or directional coupling. An axial piezoelectric sensor is dedicated to detecting small but dangerous axial forces, which can warn of the tilt of the workpiece clamping in advance during surface grinding and reduce the scrap rate. Through the above design, it is possible to avoid excessive redundancy of the number of sensors, and ensure the comprehensiveness of three-dimensional force detection, thereby improving measurement efficiency and accuracy.

[0045] Furthermore, if Figure 4 , Figure 7 , Figure 8 and Fig. 9 As shown, the detection shaft section 42 is provided with sensor mounting countersunk holes 43 that are matched one by one with four piezoelectric sensors 3. The opening of the sensor mounting countersunk hole 43 is covered by a cover plate 5. The cover plate 5 presses and limits the piezoelectric sensor 3. The outer surface of the cover plate 5 smoothly transitions with the conical surface of the detection shaft section 42 to form a continuous curved surface. The outer surface of the cover plate 5 is consistent with the generatrix curvature of the conical surface to maintain the integrity of the conical surface. Specifically, the force measuring ends of the three radially arranged piezoelectric sensors 3 can be close to the bottom of the groove of the sensor mounting countersunk hole 43, and the force measuring end of the axially arranged piezoelectric sensor 3 can be close to the groove side of the sensor mounting countersunk hole 43 and face the connecting shaft section 41. Preferably, a fixing glue can be used to fix the piezoelectric sensor 3 to the spring chuck 4 in the sensor mounting countersunk hole 43. Through the matching design of the sensor installation countersunk hole 43 and the cover plate 5, the piezoelectric sensor 3 is rigidly fixed in the detection shaft section 42, and the outer surface of the cover plate 5 and the conical surface are smoothly transitioned, maintaining the structural integrity and mechanical properties of the conical surface of the detection shaft section 42, reducing stress concentration and pneumatic noise during high-speed rotation, extending the fatigue life of the spring chuck 4, and preventing cutting fluid from invading the area where the piezoelectric sensor 3 is located. Preferably, the cover plate 5 can also be fixed to the spring chuck 4 by fixing glue and limit the piezoelectric sensor 3.

[0046] Furthermore, if Figure 7 and Figure 8As shown, a wire hole 44 is provided in the sensor mounting counterbore 43, and the wire hole 44 passes through the end of the detection shaft section 42 away from the connecting shaft section 41. The interior of the mounting handle 2 is hollow, and the piezoelectric sensor 3 is connected to the data transmission line (not shown in the figure), and the data transmission line passes through the wire hole 44 and the internal cavity of the mounting handle 2. Since the detection shaft section 42 is provided with sensor mounting counterbore 43 that matches the four piezoelectric sensors 3 one by one, and there are four sensor mounting counterbore 43 in total, a wire hole 44 can be provided in each sensor mounting counterbore 43, and the data transmission line connected to the piezoelectric sensor 3 passes through the spring clamp 4 through the wire hole 44. The wire hole 44 and the internal cavity of the mounting handle 2 form a closed wiring channel, and the data transmission line is led out through the end of the detection shaft section 42 and the internal cavity of the mounting handle 2, which optimizes the wiring path, replaces the traditional exposed cable (susceptible to mechanical damage) to avoid external cable entanglement or wear, improves the reliability of the device, and simplifies the assembly and maintenance process.

[0047] The assembly process of the grinding force online detection and control device: First, the four piezoelectric sensors 3 are respectively installed in the sensor mounting countersunk holes 43 of the detection shaft segment 42, and the data transmission line passes through the wire hole 44. The four piezoelectric sensors 3 are fixed by fixing glue, and the force measuring ends of three piezoelectric sensors 3 are arranged along the radial direction of the detection shaft segment 42 and close to the bottom of the groove, and the other is arranged along the axial direction and close to the groove side in the direction of the connecting shaft segment 41. Then, the cover plate 5 is used to cover the opening of the sensor mounting countersunk hole 43. The cover plate 5 is fixed by fixing glue to press and limit the piezoelectric sensor 3 to ensure a smooth transition between the outer surface of the cover plate 5 and the conical surface of the detection shaft segment 42; then, the connecting shaft segment 41 of the spring chuck 4 is matched with the second conical cavity 13 of the threaded cover 1, and the limiting protrusion 12 is matched with the limiting groove 45 to realize the assembly of the spring chuck 4 and the threaded cover 1; thereafter, the threaded cover 1 is threadedly connected to the mounting handle 2, so that the detection shaft segment 42 is matched with the first conical cavity 21 of the mounting handle 2, and the data transmission line passes through the internal cavity of the mounting handle 2 to the outside of the detection device, and then is connected to the signal processing module.

[0048] The grinding force control method provided in this application is applied to the above grinding force online detection and control device, refer to Fig.10 and Fig.11 As shown, including: Set the initial grinding wheel speed, initial feed speed and initial grinding feed, and set the target value of the grinding force based on the critical crack pressure of the hard and brittle material being ground; The piezoelectric sensor 3 collects the grinding force signal in real time, and the grinding force signal is processed by the signal amplification processing module and output to the PID feedback control system. The PID feedback control system adopts a quasi-ductility critical adjustment model according to the deviation between the actual grinding force detection value and the target value, and adjusts the grinding wheel speed and / or feed speed in real time through dynamic coordination to maintain constant force grinding at the target value; In the control process of the PID feedback control system, sensor signals, processing parameters and control output data are collected to build a training sample set for the feedforward deep learning network; supervised learning is used to train the deep learning model to learn the mapping relationship from sensor input to control signal output; Based on the control strategy, the proportion of feedforward control and feedback control is dynamically adjusted, and the control weight is gradually transitioned from feedback control to feedforward control.

[0049] Preferably, the quasi-ductility critical adjustment model specifically includes: a) Construct the composite control quantity equation: , where p is the dynamic adjustment factor (0.1-0.5), τ is the system response time constant, is the composite control quantity, is the actual detection value of grinding force, is the target value of grinding force, and the real-time grinding force error is mapped to the adjustment amount of grinding wheel speed and feed speed through the dynamic adjustment factor p and system response time constant 𝜏; b) Implement a secondary error response mechanism: i) When the absolute value of the grinding force error is greater than 15% of the target value, adjust the feed speed; ii) When the absolute value of the grinding force error is less than or equal to 15% of the target value, adjust the grinding wheel speed.

[0050] In a preferred embodiment, in the quasi-ductility critical adjustment model, a dynamic limit mechanism can also be used to constrain the parameter adjustment range, such as limiting the feed speed to 1.2 times the initial value, the speed fluctuation does not exceed ±20%, and limiting the single adjustment range (≤5% of the current value). hour, ;when hour, ; Intelligent adjustment factor design is used for adjustment factors, , in order to achieve rapid adjustment when the error is large and fine control when the error is small, among which, is the feed speed, is the initial feed speed, is the maximum adjustment of feed speed, is the grinding wheel speed, is the initial grinding wheel speed, is the maximum adjustment of the grinding wheel speed, is the initial adjustment factor.

[0051] Preferably, the control strategy adopts a dynamic weighted control strategy: .

[0052] in, is the output of the feedforward network, is the output of the PID feedback network, is the feedforward control weight, is the feedback control weight.

[0053] Initial setup =0, =1, gradually increases with the increase of sample data and reduce , when the error between the output of the feedforward control and the output of the feedback control is less than the set threshold, =1, = 0. The mean square error can be used to measure the difference between the feedforward control output and the feedback control output:

[0054] When the mean square error E between the feedforward control signal output by the deep learning model and the PID feedback signal is less than the set threshold, a complete transition from PID control to feedforward control is achieved.

[0055] Preferably, the control method further includes: optimizing the control action by using a deep deterministic policy gradient (DDPG) algorithm, generating a continuous control signal through an Actor-Critic network; designing a reward function, combining force error, action smoothness and processing efficiency to optimize the control strategy; and achieving a smooth transition from PID control to feedforward control through experience replay and target network soft update mechanism.

[0056] The technical effects that can be achieved by adopting the above-mentioned grinding force control method are as follows: a grinding force threshold setting method based on the critical crack pressure, combined with PID feedback control to adjust the grinding wheel parameters in real time, and a feedforward control model is constructed through deep learning to achieve the transition from experience dependence to intelligent prediction, solving the problems of response lag and parameter solidification in traditional control. The control weight is gradually transitioned through a dynamic weighted strategy. In the initial stage, PID feedback is relied on to ensure stability. As data accumulates, the proportion of feedforward control is enhanced, and finally low-latency, high-precision constant force grinding is achieved, balancing the system robustness and response speed. The DDPG algorithm is introduced to optimize the control action, and the reward function is designed in combination with force error, action smoothness and processing efficiency. Through the Actor-Critic network and experience playback mechanism, the system can autonomously adapt to complex working conditions such as grinding wheel wear and material changes, thereby improving the generalization ability of control.

[0057] The specific implementation route of this method is to calculate the critical pressure of crack generation of the hard and brittle material to be ground based on theoretical low damage calculation (for example, the critical normal load for the initiation of cracks on the surface of the material during grinding can be calculated based on the fracture toughness and elastic modulus of the hard and brittle material through the median / lateral crack theoretical model, and the safety threshold range of the grinding force is set accordingly to achieve low damage grinding), set the initial grinding force target value, including the grinding tangential force safety target value, the grinding normal force safety target value and the grinding axial force safety target value, set the initial grinding wheel speed of the servo motor, the initial feed speed and the initial grinding feed amount, during the grinding process, the force signal is taken by the piezoelectric ceramic sensor, and the signal amplification processing module implements data processing and transmits it to the initial PID feedback control system, which is fully responsible for the grinding process, and proposes the concept of "quasi-ductility" control through the deviation between the actual grinding force detection value and the target value, realizes the nonlinear decoupling of the grinding force and the motion parameters through the composite control quantity equation, and designs an intelligent response mechanism based on error classification to maintain the grinding wheel grinding at the target value constant force. At the same time, data is collected through the feedback control stage to provide samples for the feedforward deep learning network. The deep learning model learns the behavior of the PID feedback control system through supervised learning to achieve the mapping from sensor input to control signal output. In order to achieve dynamic adjustment of the proportion of feedforward control and feedback control, a weighted control strategy is introduced to achieve a gradual transition from feedback control to feedforward control; a feedforward control strategy based on deep deterministic policy gradient (DDPG) is adopted to achieve direct mapping from sensor signals to control signals through reinforcement learning. First, in the early stage of grinding, the PID feedback control system is responsible for constant force grinding, and dynamically collects force signals, processing parameters and control outputs as training samples for the reinforcement learning network. Subsequently, the DDPG algorithm learns the control strategy through the Actor-Critic architecture, in which the Actor network generates continuous control signals, the Critic network evaluates the value of the action, and stable training is achieved through experience replay and target network. The entire control system consists of multiple modules, including an environment modeling module, an Actor-Critic network module, an experience replay module, a reward design module and a dynamic weighting module. The environmental modeling module transforms the grinding force control problem into a Markov decision process in reinforcement learning, defining states (sensor signals, processing parameters, force errors, etc.), actions (servo motor control signals), and reward functions (combining force errors, oscillation amplitude, and action smoothness). The Actor network is responsible for generating control signals, and the Critic network evaluates the value of actions. Both are stably trained through experience playback and the soft update mechanism of the target network. The working path is as follows: First, in the early stage of grinding, the PID controller is responsible for constant force grinding. The state St is collected in real time through sensors, and combined with the feedforward control signal generated by the Actor network. and , using the dynamic weighting module according to the weight and Calculate the final control signal , then, by interactively recording the state, action, reward, and next state, the data is stored in the experience playback module, and the Actor-Critic network is optimized in each training cycle. In the reward design module, the reward function is optimized by combining the target force error, oscillation amplitude, and action smoothness to guide the system to learn a more efficient control strategy. With the accumulation and training of data, the feedforward control weight is gradually increased, and finally the transition to full reliance on feedforward control is achieved to achieve precise and stable constant force grinding control.

[0058] Anything not described in this application can be achieved by adopting or drawing on existing technologies.

[0059] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0060] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A grinding force online detection and control device, used to detect and control the grinding force of a grinding head during grinding of a hard and brittle material surface, wherein the grinding force includes tangential force, normal force and axial force, and is characterized in that: It comprises a threaded gland, a mounting handle, a plurality of piezoelectric sensors, a spring chuck with a hollow interior, and a signal amplification processing module. The threaded gland and the mounting handle are connected and matched to form a receiving cavity for receiving the spring chuck. The spring chuck is provided with a connecting shaft section and a detecting shaft section along the axial direction. The outer wall of the detecting shaft section is a conical surface. Along the direction from the connecting shaft section to the detecting shaft section, the detecting shaft section gradually shrinks inwards. The threaded cover limits the axial position of the spring chuck, the mounting handle is provided with a first conical cavity adapted to the outer wall of the detection shaft section, the threaded cover is provided with an avoidance hole corresponding to the center of the spring chuck, so that the grinding wheel spindle can be inserted into the spring chuck through the avoidance hole and clamped, the multiple piezoelectric sensors are installed on the detection shaft section, the grinding wheel spindle acts on the multiple piezoelectric sensors through the detection shaft section, the multiple piezoelectric sensors detect the tangential force, the normal force and the axial force, the signal amplification and processing module can convert the charge signal output by the piezoelectric sensor into a digital signal and output it to the terminal, and the terminal can control the grinding head to maintain constant force grinding with a preset grinding force target value according to a preset control logic.

2. The grinding force online detection and control device according to claim 1, characterized in that: The grinding force online detection and control device includes four piezoelectric sensors, which are evenly distributed along the circumference of the detection shaft segment, wherein the force measuring ends of three piezoelectric sensors are arranged along the radial direction of the detection shaft segment for detecting the tangential force and the normal force, and the force measuring end of another piezoelectric sensor is arranged along the axial direction of the detection shaft segment for detecting the axial force.

3. The grinding force online detection and control device according to claim 2, characterized in that: The detection shaft section is provided with sensor mounting countersunk holes which are matched one by one with the four piezoelectric sensors. The openings of the sensor mounting countersunk holes are covered by a cover plate, which presses and limits the piezoelectric sensors. The outer surface of the cover plate smoothly transitions with the conical surface of the detection shaft section to form a continuous curved surface. The outer surface of the cover plate has a consistent generatrix curvature with the conical surface to maintain the integrity of the conical surface.

4. The grinding force online detection and control device according to claim 3, characterized in that: A wire hole is provided in the sensor mounting countersunk hole, and the wire hole passes through the end of the detection shaft segment away from the connecting shaft segment. The interior of the mounting handle is hollow, and the piezoelectric sensor is connected to a data transmission line, and the data transmission line passes through the wire hole and the interior of the mounting handle.

5. The grinding force online detection and control device according to claim 1, characterized in that: The detection shaft section and the connecting shaft section are separated by a circumferentially extending limiting groove, the inner wall of the threaded cover is provided with a limiting protrusion adapted to the limiting groove, and the limiting groove and the limiting protrusion cooperate to limit the axial position of the spring chuck; the outer wall of the connecting shaft section is a conical surface, the conical surface of the connecting shaft section and the conical surface of the detection shaft section have opposite taper directions, and the threaded cover is provided with a second conical cavity adapted to the outer wall of the connecting shaft section.

6. The grinding force online detection and control device according to claim 1, characterized in that: The spring chuck is provided with a plurality of first elastic grooves extending from the connecting shaft section toward the detecting shaft section and a plurality of second elastic grooves extending from the detecting shaft section toward the connecting shaft section, and the first elastic grooves and the second elastic grooves are alternately arranged along the circumference of the spring chuck.

7. A grinding force control method, applied to the grinding force online detection and control device according to any one of claims 1 to 6, characterized in that: include: Set the initial grinding wheel speed, initial feed speed and initial grinding feed, and set the target value of the grinding force based on the critical crack pressure of the hard and brittle material being ground; The piezoelectric sensor collects the grinding force signal in real time, and the grinding force signal is processed by the signal amplification processing module and output to the PID feedback control system. The PID feedback control system adopts a quasi-ductility critical adjustment model according to the deviation between the actual grinding force detection value and the target value, and adjusts the grinding wheel speed and / or feed speed in real time through dynamic coordination to maintain constant force grinding at the target value; In the control process of the PID feedback control system, sensor signals, processing parameters and control output data are collected to build a training sample set for the feedforward deep learning network; supervised learning is used to train the deep learning model to learn the mapping relationship from sensor input to control signal output; Based on the control strategy, the proportion of feedforward control and feedback control is dynamically adjusted, and the control weight is gradually transitioned from feedback control to feedforward control.

8. The grinding force control method according to claim 7, characterized in that: The quasi-ductility critical adjustment model specifically includes: a) Construct the composite control quantity equation: , where p is the dynamic adjustment factor, τ is the system response time constant, is the composite control quantity, is the actual detection value of grinding force, is the target value of grinding force, and the real-time grinding force error is mapped to the adjustment amount of grinding wheel speed and feed speed through the dynamic adjustment factor p and system response time constant 𝜏; b) Implement a secondary error response mechanism: i) When the absolute value of the grinding force error is greater than 15% of the target value, adjust the feed speed; ii) When the absolute value of the grinding force error is less than or equal to 15% of the target value, adjust the grinding wheel speed.

9. The grinding force control method according to claim 7, characterized in that: The control strategy adopts a dynamic weighted control strategy, and the dynamic weighted control strategy is specifically: Defining feedforward control weights With feedback control weight , initially =0, =1; When the mean square error E between the feedforward control signal output by the deep learning model and the PID feedback signal is less than the set threshold, gradually increase and reduce , until =1, =0.

10. The grinding force control method according to claim 7, characterized in that: The control method further comprises: The Deep Deterministic Policy Gradient (DDPG) algorithm is used to optimize control actions and generate continuous control signals through the Actor-Critic network; Design reward functions to optimize control strategies by combining force error, motion smoothness, and processing efficiency; Through experience replay and target network soft update mechanism, a smooth transition from PID control to feedforward control is achieved.

Images