A flexible grinding method and system based on real-time detection of grinding wheel wear

Abstract

The present application belongs to the technical field of industrial robot polishing system, and discloses a flexible polishing method and system based on real-time detection of grinding wheel wear, which comprises the following steps: constructing a polishing parameter regression model; acquiring real-time grinding wheel wear state characteristic parameters; constructing a set of multi-dimensional control matrix, i.e., a grinding wheel wear state real-time detection module, based on the grinding wheel wear state characteristic parameters and in combination with the regression model parameters in the polishing process; and dynamically adjusting the polishing path, polishing force and process parameters according to the multi-dimensional control matrix to realize adaptive polishing control in different wear stages. The present application realizes intelligent polishing with dynamic adjustment of polishing strategy by fusing the grinding wheel wear parameters and the regression model parameters in the polishing process to construct a polishing control matrix.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of industrial robot polishing system, and particularly relates to a flexible polishing method and system based on real-time detection of grinding wheel wear. BACKGROUND

[0002] In the industrial fields of steel and mechanical processing, the polishing process is an important process for ensuring the surface quality of workpieces. With the continuous improvement of industrial automation level, industrial robots are gradually introduced into the polishing operation to replace the traditional manual operation mode, significantly improving the polishing efficiency and consistency. However, in the polishing process, the grinding wheel as a key tool will continue to wear itself. At different wear stages, the abrasive state on the surface of the grinding wheel will change significantly, such as bonding, passivation and grinding force reduction, etc., which will change the surface topography, cutting performance and contact characteristics of the grinding wheel, and ultimately affect the material removal rate and polishing quality.

[0003] At present, most robot polishing systems still use fixed parameter models, which fail to fully consider the nonlinear dynamic changes brought about by the wear of the grinding wheel, and therefore it is difficult to achieve adaptive regulation and control of the polishing force, polishing depth and path strategy. Although some systems introduce wear compensation strategies, they are mostly based on periodic manual detection or empirical adjustment, lacking real-time sensing ability of the grinding wheel wear state and intelligent modeling ability according to the state changes, which to a great extent limits the flexible regulation performance and polishing accuracy of the polishing system.

[0004] Therefore, it is urgent to build a flexible polishing system that integrates real-time detection of grinding wheel wear and modeling analysis of the polishing process. By constructing a set of multi-dimensional control matrixes with the grinding wheel wear characteristic parameters and the polishing process regression model parameters, the polishing state can be dynamically reflected and the polishing strategy can be adjusted in real time, realizing closed-loop optimization of the polishing process, and thus significantly improving the intelligent level of the polishing system and the stability of the polishing quality. SUMMARY

[0005] To solve the technical problems of unstable polishing quality, fixed model control and lack of adaptive ability caused by grinding wheel wear in the existing industrial robot polishing process, the present application provides a flexible polishing method and system based on real-time detection of grinding wheel wear, which aims to obtain the grinding wheel wear state characteristic parameters in real time, combine them with the regression model parameters in the polishing process, and construct a set of multi-dimensional control matrixes for dynamically adjusting the polishing path, polishing force and process parameters, realizing adaptive polishing control at different wear stages. The system not only can improve the uniformity of material removal and polishing accuracy, but also has the ability of continuous learning and optimization, significantly enhancing the flexibility, intelligence and stability of the robot polishing system.

[0006] To achieve the above-mentioned purpose, the present application provides the following solutions:

[0007] A flexible grinding method based on real-time detection of grinding wheel wear, the method comprising:

[0008] Construct a regression model for the grinding parameters;

[0009] Real-time acquisition of grinding wheel wear characteristic parameters;

[0010] Based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process, a set of multi-dimensional control matrices is constructed, namely the grinding wheel wear state real-time detection module.

[0011] Based on the multidimensional control matrix, the grinding path, grinding force and process parameters are dynamically adjusted to achieve adaptive grinding control for different wear stages.

[0012] The preferred regression model for the polishing parameters is as follows:

[0013] ;

[0014] Where D is the grinding depth, F is the grinding force, and V is the grinding wheel feed rate. s denoted as the linear velocity of the grinding wheel, μ as the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains, and σ as the standard deviation of the abrasive grain height distribution on the grinding wheel, reflecting surface roughness and uniformity. ~ Regression coefficients obtained by fitting experimental data.

[0015] Preferred methods for real-time acquisition of grinding wheel wear characteristic parameters include:

[0016] Set a fixed time interval or trigger sampling based on the grinding cycle, use a structured light camera to measure the morphology of the grinding wheel surface, and generate high-density point cloud data or three-dimensional morphology map;

[0017] The generated high-density point cloud data or 3D topography map is transmitted to the robot's host controller in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics and extracts representative probability distribution parameters.

[0018] Preferably, the method for adaptive grinding control at different wear stages by dynamically adjusting the grinding path, grinding force, and process parameters based on a multi-dimensional control matrix includes:

[0019] Construct a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area:

[0020] ;

[0021] Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model, and D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers;

[0022] A linear decreasing model is used to set the grinding depth in layers:

[0023] ;

[0024] in, For the first Polishing depth in the extended layer area The step size for decreasing grinding depth can be set according to material properties and grinding precision requirements. , indicating the number of extended ladder levels.

[0025] The present invention also provides a flexible grinding system based on real-time detection of grinding wheel wear. The system is used to implement the aforementioned method and includes: a model building module, a parameter acquisition module, a matrix building module, and a grinding control module.

[0026] The model building module is used to build a regression model for grinding parameters;

[0027] The parameter acquisition module is used to acquire the characteristic parameters of the grinding wheel wear state in real time;

[0028] The matrix construction module is used to construct a set of multi-dimensional control matrices, namely the real-time detection module for grinding wheel wear state, based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process.

[0029] The grinding control module is used to dynamically adjust the grinding path, grinding force and process parameters according to the multi-dimensional control matrix, so as to realize adaptive grinding control for different wear stages.

[0030] The preferred regression model for the polishing parameters is as follows:

[0031] ;

[0032] Where D is the grinding depth, F is the grinding force, and V is the grinding wheel feed rate. s denoted as the linear velocity of the grinding wheel, μ as the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains, and σ as the standard deviation of the abrasive grain height distribution on the grinding wheel, reflecting surface roughness and uniformity. ~ Regression coefficients obtained by fitting experimental data.

[0033] Preferably, the parameter acquisition module includes: a generation unit and an analysis unit;

[0034] The generation unit is used to set a fixed time interval or trigger sampling based on the grinding cycle, use a structured light camera to measure the morphology of the grinding wheel surface, and generate high-density point cloud data or three-dimensional morphology map.

[0035] The analysis unit is used to transmit the generated high-density point cloud data or three-dimensional topography map to the robot's host controller in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics and extracts representative probability distribution parameters.

[0036] Preferably, the process of dynamically adjusting the grinding path, grinding force, and process parameters based on a multi-dimensional control matrix to achieve adaptive grinding control for different wear stages includes:

[0037] Construct a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area:

[0038] ;

[0039] Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model, and D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers;

[0040] A linear decreasing model is used to set the grinding depth in layers:

[0041] ;

[0042] in, For the first Polishing depth in the extended layer area The step size for decreasing grinding depth can be set according to material properties and grinding precision requirements. , indicating the number of extended ladder levels.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] This invention introduces a real-time detection mechanism for grinding wheel wear and combines it with a multivariate grinding parameter regression model to construct a flexible grinding system with adaptive control capabilities, which has the following beneficial effects:

[0045] 1. Improve polishing precision and consistency:

[0046] This invention achieves online adjustment of grinding path and parameters by real-time monitoring of grinding wheel wear (including the probability distribution characteristics of abrasive grain protrusion) and dynamically updating key variables in the grinding model. This effectively overcomes the error accumulation problem that occurs in traditional fixed parameter models during long-term grinding, and improves the uniformity of material removal and the consistency of surface quality.

[0047] 2. Achieve dynamic modeling and updating of the polishing model:

[0048] This invention constructs a multi-dimensional parameter matrix for grinding control by using the grinding wheel linear velocity (composed of angular velocity and radius) and surface morphology parameters as input variables for a regression model. This supports the system in updating the model and adaptively adjusting the grinding strategy based on the real-time status, thus solving the problem that traditional models cannot reflect the nonlinear changes in the wear process.

[0049] 3. Enhance the system's flexibility and intelligence:

[0050] This invention utilizes a structured light vision system to acquire grinding wheel surface data in real time, automatically extract the feature parameters (such as μ, σ, etc.) required for the grinding model, and embed them into the control logic, enabling the system to have state perception, model update and parameter optimization functions, and achieve highly compliant and flexible grinding control.

[0051] 4. Enhance the ability to adaptively handle complex defect morphologies:

[0052] For areas where the depth drops sharply at the edge of defects, this invention proposes a stepped grinding strategy. Depth gradient control logic is set on the periphery of the grinding path to give the grinding process better transition characteristics, avoid local over-cutting or residue, and improve the overall smoothness and precision of grinding.

[0053] 5. Possesses good adaptability for promotion and industrial application value:

[0054] This invention can be widely applied in industrial scenarios such as metal processing, machinery manufacturing, ship repair, and grinding of rail transit components. It is compatible with various types of robot grinding units and grinding tools, has good system integration and versatility, and has broad prospects for promotion. Attached Figure Description

[0055] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of the stepped edge grinding process according to an embodiment of the present invention;

[0057] Figure 2 This is a flowchart of the grinding wheel morphology detection process according to an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of a flexible grinding method based on real-time detection of grinding wheel wear according to an embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Example 1

[0062] like Figure 3 As shown, this invention provides a flexible grinding method based on real-time detection of grinding wheel wear, the method comprising:

[0063] Construct a regression model for the grinding parameters;

[0064] Real-time acquisition of grinding wheel wear characteristic parameters;

[0065] Based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process, a set of multi-dimensional control matrices is constructed, namely the grinding wheel wear state real-time detection module.

[0066] Based on the multidimensional control matrix, the grinding path, grinding force and process parameters are dynamically adjusted to achieve adaptive grinding control for different wear stages.

[0067] In this embodiment, the grinding parameter regression model is constructed as follows:

[0068] Through multiple grinding experiments, key grinding parameters were collected, including grinding force (F), robot moving speed (v), grinding wheel rotational angular velocity (ω), grinding wheel radius (r), and actual grinding depth (D). Simultaneously, by combining feature detection of the grinding wheel surface condition, the protrusion height distribution function of the grinding wheel surface was obtained. and the distribution function of wear variation at the angle of attack of abrasive particles .

[0069] ;

[0070] ;

[0071] in, The protrusion height of the abrasive grains, R is the angle of attack of the abrasive grains, and R is the radius of the grinding wheel.

[0072] In Matlab, an adaptive algorithm is used to process the data (the distribution of protrusion height on the grinding wheel surface and the distribution of the number of abrasive grains) to obtain probability distribution indicators of the protrusion height of abrasive grains on the grinding wheel surface, such as the mean (μ) and standard deviation (σ).

[0073] Considering the linear velocity V of the grinding wheel s It is a key factor affecting the polishing effect, defined as V s = ω · r, which is one of the core variables in the regression model.

[0074] Using the polishing depth D as the target variable, a multiple linear regression model for the polishing process is constructed as follows:

[0075] ;

[0076] Where D is the grinding depth (mm); F is the grinding force (N); V is the grinding wheel feed rate (mm / s); V s σ is the linear velocity of the grinding wheel, which is equal to the product of the grinding wheel's angular velocity and radius (mm / s); μ is the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains; σ is the standard deviation of the abrasive grain height distribution on the grinding wheel, reflecting the surface roughness and uniformity. ~ The regression coefficients obtained by fitting experimental data; e represents the base of the natural logarithm.

[0077] This model integrates the inherent relationship between the wear state of the grinding wheel and the parameters of the grinding process, and repeatedly takes into account the real-time dynamic changes in the abrasive particles on the grinding wheel surface. It can provide a theoretical basis and quantitative evidence for subsequent grinding strategy regulation and adaptive control.

[0078] In this embodiment, a real-time grinding wheel wear state detection module is constructed: a surface state detection device is integrated into the grinding system. This device consists of a laser rangefinder and a 3D surface profile camera. The surface state detection device acquires images of the grinding wheel surface state in real time and transmits the data to a built-in data processing program in Matlab. Through image recognition, laser measurement, or structured light technology, the current abrasive grain distribution characteristics on the grinding wheel surface are obtained in real time. Parameters such as the distribution mean (μ) and standard deviation (σ) are extracted, and feature quantization is performed. The corresponding grinding process parameters are then output: force F, grinding wheel feed rate V, and grinding wheel rotational angular velocity V0. s wait.

[0079] In this embodiment, the real-time grinding wheel wear state detection module can update the grinding wheel wear state in real time, providing the grinding system with input parameters of the current grinding wheel state. Together with key process parameters such as grinding force and robot speed, it forms a state input that can be updated in real time.

[0080] Dynamic updates and parameter tuning of regression models:

[0081] An innovative real-time prediction and dynamic update mechanism for the grinding process was constructed. The core of this mechanism lies in dynamically adjusting the coefficient weights of the grinding force regression model using grinding wheel wear state parameters (such as the mean μ and standard deviation σ of the protrusion height distribution), enabling the model to possess real-time adaptive capabilities. This method differs not only from traditional static model regression methods but also from existing online training methods based on black-box neural networks. Instead, it employs a hybrid model structure combining distribution state perception and physical modeling with embedded weight control—namely, the protrusion height distribution function of the grinding wheel surface. and the distribution function of wear variation at the angle of attack of abrasive particles .

[0082] Traditional methods treat the mean μ and standard deviation σ as constants, considering them only during model construction. The innovation of this invention lies in:

[0083] The surface abrasive protrusion height distribution of the current grinding wheel is obtained through real-time sensors or image processing (via point cloud or contour map).

[0084] Extract statistics (such as mean μ, standard deviation σ);

[0085] Dynamically calculate the predicted value.

[0086] This dynamic updating method can fully consider the impact of grinding wheel wear on the grinding effect, thereby ensuring precise control and optimization of the grinding process.

[0087] Considering the nonlinear evolution of the wear state of the grinding wheel during use, a dynamic weight control mechanism is proposed to enable the model to adapt to the grinding wheel state at different stages.

[0088] This invention introduces a wear stage factor λ(t), defined as:

[0089] ;

[0090] Where R0 is the initial grinding wheel radius; R(t) is the current grinding wheel radius (which can be indirectly inferred through laser measurement or linear velocity); R min The minimum usable radius.

[0091] Based on the current grinding wheel state statistics, a dynamic weight adjustment function is defined:

[0092] ;

[0093] Where, β i 0 γ is the initial coefficient; i A coefficient adjustment factor related to the degree of wear (reflecting "macroscopic wear"); δ iA fine-tuning factor related to distribution characteristics (reflecting "microscopic wear"); f(μ) t ,σ t) It can be a linear or nonlinear function.

[0094] In this embodiment, the closed-loop feedback control system integrates:

[0095] This invention further proposes integrating a closed-loop feedback control mechanism into the grinding system, deriving the model-predicted force F, grinding wheel feed rate V, and grinding wheel rotational angular velocity V from the equation process of the grinding depth D. s The grinding process allows for closed-loop control of grinding quality, ensuring that the grinding quality meets the expected requirements and greatly increasing grinding efficiency.

[0096] The robot can adjust the grinding strategy in real time according to the different wear stages of the grinding wheel and the actual depth of the defect, ensuring that each grinding can be optimally adjusted according to the current state of the grinding wheel, adapting to different types of defects, and ultimately improving the stability and accuracy of the grinding effect.

[0097] Specifically, after the robot receives the grinding area location, it will perform secondary processing on the grinding defect area. The width of each pass of the grinding wheel is w, and the distance from the next multiple of w to the next width of the square is L. Therefore, we process the initial grinding area by expanding the long side of the grinding square to a multiple of w (see appendix). Figure 1 ):

[0098] ;

[0099] in, The x-axis coordinates of each point before the secondary processing; These are the x-axis coordinates of each point after secondary processing; For each point before secondary processing Axis coordinates; For each point after secondary processing Axis coordinates.

[0100] After the second processing, the robot will move from point A to point B, and then continue moving from point A' to point B' in the next pass, repeating this process until all passes have been completed and the defective area has been polished. (See attached image) Figure 1 ).

[0101] Specifically, after the defective area is polished, some areas exhibit abrupt changes in defect depth, particularly at the edges where significant height jumps or "cliff-like" drops may occur. This can easily lead to poor transitions at the polished edges and uneven surface quality. To improve the smoothness of the polished edge transition and surface consistency, this invention proposes a "stepped edge polishing strategy." This method further expands the polishing process at the defective edge area, setting multi-level polishing depth control to create a gradually decreasing depth gradient from the center outwards. This achieves a smooth transition at the edge, effectively eliminating abrupt step effects at the polished edges and improving overall surface quality and polishing consistency.

[0102] The specific implementation method is as follows:

[0103] The system constructs a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area:

[0104] ;

[0105] Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model; D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers;

[0106] To control the smoothness of edge polishing, the system uses a linear decreasing model to set the polishing depth in layers:

[0107] ;

[0108] in, For the first Polishing depth of the extended layer area; The grinding depth decreasing step size can be set according to the material properties and grinding precision requirements; , indicating the number of extended ladder levels;

[0109] This linear decreasing model reduces the grinding depth gradually in the edge region, forming a gradient grinding layer. This effectively alleviates the abrupt grinding edge phenomenon caused by sudden changes in defect depth, achieving a smooth transition in the grinding effect.

[0110] Furthermore, this model can be used in conjunction with grinding wheel wear state parameters (such as S = {μ, σ}) to assess Δd or individual layers. Dynamic adjustments are made to intelligently optimize the grinding edge compensation strategy under different grinding wheel wear stages, further improving the system's compliant grinding capability and the consistency of workpiece surface quality.

[0111] Specifically, to achieve accurate perception of the wear state of the grinding wheel and dynamic updating of the grinding model, this invention introduces a three-dimensional blue structured light camera as a real-time detection device for the surface condition of the grinding wheel. This camera possesses high-precision, high-resolution three-dimensional topography acquisition capabilities, and can periodically scan and sample the grinding wheel surface during the grinding process to obtain complete distribution information of the current abrasive grain protrusion height.

[0112] The acquisition process is as follows: The system sets a fixed time interval or triggers sampling based on the grinding cycle. The structured light camera measures the morphology of the grinding wheel surface and generates high-density point cloud data or three-dimensional morphology map. The required grinding wheel surface morphology data, distribution mean (μ) and standard deviation (σ) and other parameters are extracted.

[0113] Subsequently, the system transmits the collected data (high-density point cloud data or 3D topographic maps) to the robot's host controller (host computer) in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics, extracting representative probability distribution parameters such as mean (μ) and standard deviation (σ). These parameters serve as key input variables in the grinding regression model, directly participating in grinding force prediction and adaptive strategy adjustment. This enables the robot to self-adapt and adjust during the grinding process to achieve the desired grinding quality.

[0114] This process constructs a closed-loop processing flow from "sensor acquisition → parameter calculation → model invocation," ensuring that the system can dynamically adjust the grinding model according to the actual wear state of the grinding wheel, achieving true wear adaptive control (see appendix). Figure 2 ).

[0115] Example 2

[0116] The present invention also provides a flexible grinding system based on real-time detection of grinding wheel wear. The system is used to implement the method described in Embodiment 1. The system includes: a model building module, a parameter acquisition module, a matrix building module, and a grinding control module.

[0117] The model building module is used to build a regression model for polishing parameters.

[0118] The parameter acquisition module is used to acquire the characteristic parameters of the grinding wheel wear state in real time;

[0119] The matrix construction module is used to construct a set of multi-dimensional control matrices, namely the real-time detection module for grinding wheel wear state, based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process.

[0120] The grinding control module is used to dynamically adjust the grinding path, grinding force and process parameters according to the multi-dimensional control matrix, so as to realize adaptive grinding control for different wear stages.

[0121] In this embodiment, the constructed grinding parameter regression model is as follows:

[0122] ;

[0123] Where D is the grinding depth, F is the grinding force, and V is the grinding wheel feed rate. s denoted as the linear velocity of the grinding wheel, μ as the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains, and σ as the standard deviation of the abrasive grain height distribution on the grinding wheel, reflecting surface roughness and uniformity. ~ Regression coefficients obtained by fitting experimental data.

[0124] In this embodiment, the parameter acquisition module includes: a generation unit and an analysis unit;

[0125] The generation unit is used to set a fixed time interval or trigger sampling based on the grinding cycle, use a structured light camera to measure the morphology of the grinding wheel surface, and generate high-density point cloud data or three-dimensional morphology map.

[0126] The analysis unit is used to transmit the generated high-density point cloud data or 3D topography map to the robot's host controller in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics and extracts representative probability distribution parameters.

[0127] In this embodiment, the process of dynamically adjusting the grinding path, grinding force, and process parameters according to the multi-dimensional control matrix to achieve adaptive grinding control for different wear stages includes:

[0128] Construct a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area:

[0129] ;

[0130] Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model, and D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers;

[0131] A linear decreasing model is used to set the grinding depth in layers:

[0132] ;

[0133] in, For the first Polishing depth in the extended layer area The step size for decreasing grinding depth can be set according to material properties and grinding precision requirements. , indicating the number of extended ladder levels.

[0134] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A flexible grinding method based on real-time detection of grinding wheel wear, characterized in that, The method includes: Construct a regression model for the grinding parameters; Real-time acquisition of grinding wheel wear characteristic parameters; Based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process, a set of multi-dimensional control matrices is constructed, namely the grinding wheel wear state real-time detection module. Based on the multidimensional control matrix, the grinding path, grinding force and process parameters are dynamically adjusted to achieve adaptive grinding control for different wear stages. The constructed grinding parameter regression model is as follows: ； in, D To refine the depth, F For polishing power, V This refers to the grinding wheel feed rate. V s The linear velocity of the grinding wheel. μ This represents the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains. σ The standard deviation of the abrasive grain height distribution on the grinding wheel reflects surface roughness and uniformity. ~ Regression coefficients obtained by fitting experimental data; Methods for real-time acquisition of grinding wheel wear characteristic parameters include: Set a fixed time interval or trigger sampling based on the grinding cycle, use a structured light camera to measure the morphology of the grinding wheel surface, and generate high-density point cloud data or three-dimensional morphology map; The generated high-density point cloud data or 3D topography map is transmitted to the robot's host controller in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics and extracts representative probability distribution parameters. Based on a multidimensional control matrix, methods for adaptive grinding control at different wear stages include dynamically adjusting the grinding path, grinding force, and process parameters. Construct a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area: ； Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model, and D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers; A linear decreasing model is used to set the grinding depth in layers: ； in, For the first Polishing depth in the extended layer area The step size for decreasing grinding depth can be set according to material properties and grinding precision requirements. , indicating the number of extended ladder levels.

2. A flexible grinding system based on real-time detection of grinding wheel wear, the system being used to implement the method of claim 1, characterized in that, The system includes: a model building module, a parameter acquisition module, a matrix building module, and a polishing control module; The model building module is used to build a regression model for grinding parameters; The parameter acquisition module is used to acquire the characteristic parameters of the grinding wheel wear state in real time; The matrix construction module is used to construct a set of multi-dimensional control matrices, namely the real-time detection module for grinding wheel wear state, based on the characteristic parameters of grinding wheel wear state and combined with the regression model parameters in the grinding process. The grinding control module is used to dynamically adjust the grinding path, grinding force and process parameters according to the multi-dimensional control matrix, so as to realize adaptive grinding control for different wear stages.

3. The system according to claim 2, characterized in that, The constructed grinding parameter regression model is as follows: ； in, D To refine the depth, F For polishing power, V This refers to the grinding wheel feed rate. V s The linear velocity of the grinding wheel. μ This represents the average height of the abrasive grains protruding from the grinding wheel surface, reflecting the average cutting ability of the overall abrasive grains. σ The standard deviation of the abrasive grain height distribution on the grinding wheel reflects surface roughness and uniformity. ~ Regression coefficients obtained by fitting experimental data.

4. The system according to claim 2, characterized in that, The parameter acquisition module includes: a generation unit and an analysis unit; The generation unit is used to set a fixed time interval or trigger sampling based on the grinding cycle, use a structured light camera to measure the morphology of the grinding wheel surface, and generate high-density point cloud data or three-dimensional morphology map. The analysis unit is used to transmit the generated high-density point cloud data or three-dimensional topography map to the robot's host controller in real time. The embedded parameter extraction and modeling algorithm module automatically performs statistical analysis on the abrasive grain height distribution characteristics and extracts representative probability distribution parameters.

5. The system according to claim 2, characterized in that, The process of dynamically adjusting the grinding path, grinding force, and process parameters based on a multi-dimensional control matrix to achieve adaptive grinding control for different wear stages includes: Construct a grinding depth control matrix M to describe the grinding depth distribution in the defect area and its edge extension area: ； Where D0 is the grinding depth of the core defect area, determined by the grinding parameter regression model, and D1, D2, ..., D n The polishing depth of the edge areas extending outward from the core area, where n is the number of layers; A linear decreasing model is used to set the grinding depth in layers: ； in, For the first Polishing depth in the extended layer area The step size for decreasing grinding depth can be set according to material properties and grinding precision requirements. , indicating the number of extended ladder levels.

Images