Self-adaptive force control grinding system and method for complex curved surface machining robot

Abstract

The invention relates to the technical field of grinding, in particular to a self-adaptive force control grinding system and method for a complex curved surface machining robot. Three-dimensional point cloud data of a workpiece are obtained through machine vision, gridding processing is carried out, and normal vector and curvature characteristic parameters are calculated; performing operation area division and operation path planning based on curvature characteristics; calculating operation parameters according to the normal vector and curvature characteristics; a self-adaptive impedance control algorithm is adopted, the polishing force is monitored in real time through a force sensor, and the rigidity and damping parameters of the robot are dynamically adjusted; and performing defect identification on the surface image of the processed workpiece by using the deep learning model, and when defects are identified, re-planning the operation path. According to the method, self-adaptive grinding of the complex curved surface workpiece is achieved, and the grinding efficiency and quality of the complex curved surface can be improved.

Description

Technical Field

[0001] The present application relates to the field of polishing technology, and specifically to an adaptive force-controlled polishing system and method for a complex surface processing robot. Background Art

[0002] As modern manufacturing evolves towards high precision, high quality, and high efficiency, the demand for polishing complex curved workpieces is growing. Traditional manual polishing methods struggle to meet the demands of modern industrial production, suffering from issues such as low efficiency, inconsistent quality, and high labor intensity. Therefore, research into automated and intelligent robotic polishing technology is of great significance.

[0003] Currently, existing robotic grinding technology primarily relies on offline programming based on CAD models to generate grinding trajectories and automatically grind complex surfaces. This approach relies on the workpiece CAD model, making it difficult to adapt to errors and deformations encountered during actual machining. Furthermore, setting grinding parameters relies heavily on experience and lacks an adaptive adjustment mechanism for workpiece surface features, resulting in unstable grinding quality.

[0004] On the other hand, existing technologies for polishing quality assessment primarily rely on manual inspection and lack intelligent surface defect recognition and classification capabilities. This results in low efficiency and accuracy in polishing quality assessment, making it difficult to achieve real-time feedback and optimization. Furthermore, existing technologies often require manual repair of defects that occur during the polishing process, lacking automated compensatory polishing mechanisms, hindering further improvements in polishing efficiency and quality.

[0005] In order to overcome the above shortcomings, it is necessary to study a new robotic grinding technology that comprehensively considers multiple aspects such as workpiece morphology detection, grinding parameter optimization, quality evaluation feedback and defect repair grinding to achieve intelligent grinding of complex curved workpieces.

[0006] In view of this, the present application proposes an adaptive force-controlled grinding system and method for complex surface processing robots. Summary of the Invention

[0007] To achieve the above objectives, the present application provides an adaptive force-controlled grinding system and method for complex surface processing robots. The specific technical solutions are as follows:

[0008] An adaptive force-controlled grinding method for a complex surface processing robot, comprising:

[0009] Acquiring three-dimensional point cloud data of a workpiece to be processed, meshing the point cloud data to obtain a mesh model, and calculating characteristic parameters of a normal vector and curvature of a surface of the workpiece to be processed in the mesh model;

[0010] Based on the characteristic parameters, the working area of ​​the workpiece to be processed is calculated, and the motion path points are generated in each working area to generate the working trajectory of the workpiece to be processed;

[0011] Calculating the target force, feed rate, depth of action, and operation rhythm parameters of the operation area according to the normal vector and curvature of the workpiece to be processed;

[0012] The robot is controlled to move along the planned operating trajectory, and the applied force is monitored in real time through a force sensor. An adaptive impedance control algorithm is used to dynamically adjust the stiffness and damping parameters of the robot during operation based on the feedback information of the applied force.

[0013] Obtain image data of the processed workpiece surface, extract texture and roughness features from the image data, and build a deep learning model to identify processing defects; when processing defects are identified, readjust the operation path and plan it.

[0014] Preferably, the workpiece to be processed is scanned at multiple angles to obtain point cloud data of the workpiece to be processed;

[0015] The Poisson surface reconstruction algorithm is used to mesh the point cloud and generate a triangular mesh model of the workpiece to be processed;

[0016] Calculate the characteristic parameters of the normal vector and curvature in the mesh model of the workpiece to be processed.

[0017] Preferably, a curvature threshold and a normal vector angle threshold are set, and for each point on the mesh model, the Gaussian curvature and normal vector of the point are calculated:

[0018] If the curvature of a point is greater than the curvature threshold, the point is marked as a high curvature point. For each high curvature point, the angle between the normal vector of the high curvature point and the positive direction of the Z axis is calculated;

[0019] If the normal vector angle is greater than the normal vector angle threshold, the high curvature point is on a large slope surface;

[0020] Mark the points that meet both the curvature threshold and the normal vector angle threshold as points to be processed;

[0021] Taking each point to be processed as a seed point, the region growing algorithm is used to extract the area to be processed;

[0022] The area of ​​the extracted area to be operated is calculated. If it is greater than the set area threshold, the area to be operated is determined to be a valid area to be operated. Otherwise, the area is eliminated.

[0023] Preferably, the number of sampling points inside the area to be operated is calculated;

[0024] The Poisson disk sampling algorithm is used to generate uniformly distributed sampling points in the area to be operated;

[0025] In the Poisson disk sampling algorithm, the area to be operated, the number of sampling points and the minimum sampling distance are input, and the operation path point set is output.

[0026] Preferably, the Poisson disk sampling algorithm steps include:

[0027] Step a: Randomly select an initial point in the area to be operated and add it to the path point set;

[0028] Step b: With the initial point as the center and the radius set, a sampling disk is generated in the area to be operated;

[0029] Step c: Select a random point in the sampling disk and calculate the distance between the random point and all points in the path point set; if the distance between the random point and any point is less than the set radius, reject the random point and return to step c; otherwise, accept the random point and add it to the path point set;

[0030] Step d: Repeat steps b and c until no new point is accepted within the set number of times;

[0031] Step e: When all sampling points are screened, sampling ends and the path point set is output.

[0032] Preferably, the points in the path point set are numbered and sorted to obtain an ordered sequence of operation path points;

[0033] Between adjacent path points, cubic spline interpolation is used to generate a smooth operation path curve;

[0034] Connect the operation path curves of all the areas to be operated in sequence to obtain a complete operation path.

[0035] Preferably, the operation parameters are calculated according to the normal vector and curvature of the area to be operated, and the operation parameters include: grinding force, feed speed, grinding depth and grinding time;

[0036] The grinding actuator moves along the planned working path and monitors the grinding force in real time through a force sensor. An adaptive impedance control algorithm is used to dynamically adjust the stiffness and damping parameters of the grinding actuator based on feedback from the grinding force.

[0037] The expected impedance model is constructed to calculate the expected trajectory acceleration of the grinding actuator. The expected trajectory acceleration is input into the robot's motion controller to perform impedance control on the grinding actuator.

[0038] Preferably, image data of the surface of the processed workpiece is obtained, and an industrial camera is used to image the surface of the processed workpiece to obtain surface image data;

[0039] Extract the texture and roughness features from the image data of the finished workpiece surface, and use the gray-level co-occurrence matrix (GLCM) to extract the texture features of the image;

[0040] The surface image of the workpiece is annotated after processing, the defects on the surface image of the workpiece are annotated, and a polished workpiece dataset is constructed for training the defect recognition model.

[0041] Preferably, a defect recognition model is constructed based on a convolutional neural network (CNN) model, the extracted image feature vector is used as the input of the CNN model, the defect category is used as the output, and the CNN model is trained using the constructed polished workpiece dataset;

[0042] Use the trained CNN model to identify defects on the surface image of the newly processed workpiece. If grinding defects are identified, the location and range information of the defect area are extracted.

[0043] Map the defect area back to the original workpiece model to obtain the three-dimensional position of the defect on the workpiece;

[0044] Replan the operation path in the defect area based on the defect location and scope.

[0045] An adaptive force-controlled grinding system for a complex surface machining robot, which is used to implement the adaptive force-controlled grinding method for a complex surface machining robot, includes a data acquisition module, a trajectory calculation module, an operation parameter calculation module, an impedance control module, and a defect detection module;

[0046] A data acquisition module acquires three-dimensional point cloud data of the workpiece to be processed, performs gridding processing on the point cloud data to obtain a grid model, and calculates characteristic parameters of the normal vector and curvature of the surface of the workpiece to be processed in the grid model;

[0047] The trajectory calculation module calculates the working area of ​​the workpiece to be processed based on the characteristic parameters, generates motion path points in each working area, and generates the working trajectory of the workpiece to be processed;

[0048] An operation parameter calculation module calculates the operation parameters of the target force, feed speed, action depth and operation rhythm of the operation area according to the normal vector and curvature of the workpiece to be processed;

[0049] The impedance control module controls the robot's movement according to the planned operation trajectory, monitors the applied force in real time through a force sensor, and uses an adaptive impedance control algorithm to dynamically adjust the robot's stiffness and damping parameters during operation based on the applied force feedback information.

[0050] The defect detection module obtains image data of the processed workpiece surface, extracts texture and roughness features in the image data, and builds a deep learning model to identify processing defects. When a processing defect is identified, the operation path is readjusted and planned.

[0051] Beneficial effects of this application: This application obtains three-dimensional point cloud data of the workpiece and performs grid processing, calculates normal vectors and curvature characteristic parameters, provides accurate geometric information for subsequent operation path planning and parameter optimization, and improves grinding accuracy.

[0052] This application calculates the grinding area based on curvature characteristic parameters and generates operation path points, automatically generates an optimized grinding trajectory, reduces the workload of manual programming, and improves grinding efficiency.

[0053] This application calculates grinding parameters such as grinding force, feed speed, grinding depth and grinding time based on normal vector and curvature characteristics, realizes adaptive optimization of grinding parameters, and ensures the stability and consistency of grinding quality.

[0054] This application adopts an adaptive impedance control algorithm to dynamically adjust the stiffness and damping parameters of the grinding execution end according to the grinding force feedback information monitored in real time by the force sensor, thereby realizing real-time adaptive control of the grinding process and improving the grinding quality and efficiency.

[0055] This application obtains image data of the processed workpiece surface, extracts texture and roughness features, and constructs a deep learning model to identify polishing defects, thereby realizing intelligent evaluation and feedback of polishing quality. When defects are identified, the operation path is automatically replanned, thereby improving the polishing quality and degree of automation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flow chart of an adaptive force-controlled grinding method for a complex surface machining robot provided in this application;

[0057] Figure 2 This is a structural diagram of an adaptive force-controlled grinding system for complex surface processing robots provided in this application. DETAILED DESCRIPTION

[0058] For a better understanding of the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely descriptions of exemplary embodiments of the present application and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0059] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The accompanying drawings are for illustration only and are not drawn strictly to scale. As used herein, the terms "substantially," "approximately," and similar terms are used to indicate approximations, not degrees, and are intended to illustrate inherent deviations in measurements or calculations that would be recognized by a person of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these steps would occur in actual operation, unless otherwise specified or inferred from the context.

[0060] It should also be understood that expressions such as "comprises," "including," "having," "includes," and / or "comprising" are open rather than closed expressions in this specification, indicating the presence of the stated features, elements, and / or components, but do not exclude the presence of one or more other features, elements, components, and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application." And, the term "exemplary" is intended to refer to an example or illustration.

[0061] Unless otherwise defined, all words used herein (including engineering terms and scientific and technological terms) have the same meaning as commonly understood by those skilled in the art to which this application belongs. It should also be understood that, unless otherwise specified in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as they do in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0062] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0063] Example 1

[0064] Reference Figure 1 , which is the first embodiment of the present application, provides an adaptive force-controlled grinding method for a complex surface processing robot.

[0065] S1: Acquire three-dimensional point cloud data of a workpiece to be processed, perform gridding processing on the point cloud data to obtain a grid model, and calculate characteristic parameters of the normal vector and curvature of the surface of the workpiece to be processed in the grid model.

[0066] A structured light 3D scanner is used to scan the workpiece at multiple angles to obtain point cloud data of the workpiece. The point cloud data is preprocessed, and a statistical filtering method is used to remove noise and outliers, and then the point cloud data is smoothed.

[0067] The Poisson surface reconstruction algorithm is used to mesh the point cloud and generate a triangular mesh model of the workpiece to be processed.

[0068] Calculate the normal vectors in the mesh model of the workpiece to be processed and curvature K i characteristic parameters.

[0069] For each point p in the point cloud i , find p i The nearest neighbor point set Represents point p i The set of points closest to the nearest point, calculate The center point coordinates Calculate the covariance matrix C: Where C represents the neighborhood point set The covariance matrix of k is the number of neighborhood points, p j are the coordinates of the neighborhood points.

[0070] Perform eigenvalue decomposition on the covariance matrix C Take the unit eigenvector corresponding to the minimum eigenvalue As p i Normal vector λ j and denote the eigenvalue and unit eigenvector of the covariance matrix C, respectively. Represents point p i The normal vector of .

[0071] Calculate each neighborhood point Normal vector Represents the neighborhood point p j Normal vector of ; calculate the mean of normal vectors Represents the normal vector of the neighborhood point The arithmetic mean of the normal vector covariance matrix ∑ i : Among them, ∑ i Represents the normal vector of the neighborhood point The covariance matrix of ∑ i Do eigenvalue decomposition Take the minimum eigenvalue λ0 as p i The curvature K i ;λ j and Represents the normal vector covariance matrix ∑ i The eigenvalues ​​and unit eigenvectors of K i Represents point p i curvature.

[0072] By acquiring the workpiece's 3D point cloud data and performing gridding processing, a mesh model of the workpiece is generated, providing the necessary geometric information for subsequent path planning and parameter optimization. Calculating the mesh model's normal vectors and curvature characteristic parameters also provides a basis for subsequent grinding area division and grinding parameter calculation, improving grinding accuracy and efficiency.

[0073] S2: Based on the characteristic parameters, the working area of ​​the workpiece to be processed is calculated, and a motion path point is generated in each working area to generate the working trajectory of the workpiece to be processed.

[0074] Set the curvature threshold K th and normal vector angle threshold θ th , for each point p on the grid model i , calculate the Gaussian curvature K of the point i and normal vector

[0075] If point p i The curvature K i Greater than the curvature threshold K th , that is, K i >K th , then p i Marked as high curvature point, for each high curvature point p i , calculate its normal vector Angle θ with the positive direction of the Z axis i : in, is the unit vector in the positive direction of the Z axis.

[0076] If the normal vector angle θ i Greater than the normal vector angle threshold θ th , that is, θ i >θ th , it means that the point is on a surface with a large slope.

[0077] Mark the point that satisfies both the curvature threshold and the normal vector angle threshold as the point to be processed v i .

[0078] Each point to be processed v i As the seed point, the region growing algorithm is used to extract the area to be operated: set the normal vector angle difference threshold Δθ th , for the point v to be processed i , check the point to be processed v iPoint v in one-hop neighborhood j : Calculate v i and v j The normal vector angle difference is:

[0079] If Δθ ij <Δθ th , description v j With v i The normal vector angles are similar, so v j Merge v i The area where the work is to be done;

[0080] Recursive check v j One-hop neighborhood of , until the region no longer expands;

[0081] The extracted area to be worked R k , calculate its area A k If A k Greater than the set area threshold A th , then determine R k It is a valid area to be worked on, otherwise the area will be eliminated.

[0082] Assume that the radius of the grinding disc at the grinding execution end is r and the area of ​​the grinding disc is A. r =πr 2 , waiting area R k The area is A k The goal of path planning is to generate a series of grinding points in the area to be worked so that the sum of the coverage areas of the grinding disc at these points is approximately equal to the area of ​​the area to be worked.

[0083] Calculate the area to be worked R k The number of internal sampling points N k : Where η∈(0,1] is the grinding disc overlap coefficient, which is usually set to η=0.5~0.8, indicating that a certain overlap between the grinding discs is allowed to ensure complete grinding coverage; Represents the ceiling function.

[0084] The Poisson disk sampling algorithm is used to select the area to be operated R. k Internally generated N k Evenly distributed sampling points Input: Area to be worked R k , the number of sampling points N k , minimum sampling distance Output: Operation path point set

[0085] Step a: In R k Randomly select an initial point Add path point set P k ;

[0086] Step b: is the center of the circle, d is the radius, and R k A sampling disk is generated inside;

[0087] Step c: Randomly select a point in the sampling disk calculate With P k The distance between all points in If the distance from any point is less than d, reject Return to step c; otherwise, accept Add it to P k ;

[0088] Step d: Repeat steps b and c until no new point is accepted for K consecutive times (K is a set value and a constant);

[0089] Step e: If the number of random points selected is less than the number of sampling points, return to step a; otherwise, the sampling ends and the path point set P is output. k .

[0090] For the path point set P k Points in The points are numbered and sorted from one end to the other to obtain an ordered sequence of operation path points.

[0091] At adjacent waypoints and Between, cubic spline interpolation is used to generate a smooth operation path curve

[0092]

[0093] in, are cubic spline basis functions, satisfying: The tangent vector at t = 0 and t = 1 is and p i +1 k The path tangent vector at is continuous; The curvature at t = 0 and t = 1 is and p i +1 k The path curvature is continuous.

[0094] All the areas to be worked R k Operation path curve Connect them in sequence to get the complete operation path.

[0095] Based on the curvature characteristic parameters of the workpiece mesh model, the area requiring grinding is automatically divided, avoiding unnecessary grinding and improving grinding efficiency. Furthermore, within each grinding area, algorithms such as Poisson disk sampling and cubic spline interpolation generate a uniform and smooth working path, optimizing the grinding trajectory, improving grinding quality, and reducing quality issues caused by missed and over-grinding.

[0096] S3: Calculate the target force, feed speed, action depth and operation parameters of the operation area according to the normal vector and curvature of the workpiece to be processed.

[0097] According to the area to be worked R k Normal vector and curvature K k , calculate the job parameters.

[0098] Grinding force F k :F k =F0+αK k ; Among them, F0 is the reference grinding force, and α is the grinding force adjustment coefficient.

[0099] Feed speed v k :v k =v0-βK k ; Wherein, v0 is the reference feed speed and β is the feed speed adjustment coefficient.

[0100] Grinding depth d k :d k =d0+γK k ; Among them, d0 is the reference grinding depth and γ is the grinding depth adjustment coefficient.

[0101] Grinding time t k : Among them, A k The area to be operated R k area.

[0102] Based on the normal vector and curvature characteristics of the work area, key process parameters such as grinding force, feed rate, grinding depth, and grinding time are adaptively calculated, optimizing grinding parameters and improving grinding efficiency and quality. Different grinding parameters are used for areas with different curvatures and inclinations, avoiding under- or over-grinding caused by improper parameter selection and ensuring consistent and stable grinding.

[0103] S4: Control the robot to move along the planned working trajectory, monitor the applied force in real time through force sensors, and use an adaptive impedance control algorithm to dynamically adjust the stiffness and damping parameters of the robot during operation based on the feedback information of the applied force.

[0104] The grinding actuator moves along the planned working path s(t) and monitors the grinding force F in real time through a force sensor. An adaptive impedance control algorithm is used to dynamically adjust the stiffness and damping parameters of the grinding actuator based on the feedback information of the grinding force F.

[0105] Stiffness parameter K p :K p =K p0 +λ p (FF k ); where K p0 is the base stiffness parameter, λ p Adjust the gain for stiffness.

[0106] Damping parameter K d :K d =K d0 +λ d (FF k ); where K d0 is the reference damping parameter, λ d Adjust the gain for damping.

[0107] The expected impedance model of the polishing execution end is: Where M is the virtual inertia of the grinding execution end, e is the grinding trajectory tracking error, that is, e = s(t) - x(t), x(t) is the actual trajectory of the robot, represents the velocity error obtained by taking the derivative with respect to time; It represents the acceleration error obtained by taking the second-order derivative with respect to time.

[0108] Calculate the expected trajectory acceleration of the grinding execution end Indicates the second-order derivative of the operation path with respect to time t; the desired trajectory acceleration Input the robot's motion controller to achieve impedance control at the grinding execution end.

[0109] The grinding actuator moves along a planned trajectory, using a force sensor to monitor the grinding force in real time and using an adaptive impedance control algorithm to dynamically adjust the robot's stiffness and damping parameters. This enables adaptive control of the grinding process and improves grinding quality. Stiffness and damping are adaptively adjusted based on feedback from the grinding force. When the grinding force is too high, stiffness is automatically reduced and damping is increased to avoid over-grinding. When the grinding force is too low, stiffness is automatically increased to ensure the removal rate, enhancing both the adaptability and robustness of the grinding process.

[0110] S5: Obtain image data of the processed workpiece surface, extract texture and roughness features from the image data, and build a deep learning model to identify processing defects; when processing defects are identified, readjust the operation path and plan it.

[0111] Acquire image data of the processed workpiece surface, use an industrial camera to image the processed workpiece surface and acquire surface image data; preprocess the acquired image data, including image denoising, enhancement and correction; extract texture and roughness features from the processed workpiece surface image data, use the gray-level co-occurrence matrix (GLCM) to extract the texture features of the image, and calculate the GLCM matrix P(h,m,b,θ) of the image, where h and m are gray levels, b is the pixel spacing, and θ is the direction.

[0112] Texture features including contrast, correlation, and energy are calculated based on the GLCM matrix. Roughness features of the image are extracted using wavelet transform, and the image is subjected to wavelet decomposition to obtain wavelet coefficients at different scales. The statistical features of the mean, variance, and skewness of the wavelet coefficients at each scale are calculated as roughness features. The extracted texture features and roughness features are combined to form a feature vector of the image.

[0113] A deep learning model is constructed to identify polishing defects; the surface image of the processed workpiece is annotated, and the defects therein are annotated to form a training data set; a defect recognition model is constructed using a convolutional neural network (CNN) model, and a CNN model network structure is constructed. The CNN model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer.

[0114] The extracted image feature vector is used as the input of CNN and the defect category is used as the output to train the CNN model. The model is evaluated using methods such as cross-validation to optimize the model structure and hyperparameters.

[0115] Use the trained CNN model to identify defects on the surface image of the newly processed workpiece. If a grinding defect is identified, the location and range information of the defect area is extracted; the defect area is mapped back to the original workpiece model to obtain the three-dimensional position of the defect on the workpiece; based on the defect location and range, the operation path of the area is replanned; at the same time, the density of grinding points in the defect area is increased and the grinding spacing is reduced; based on the severity of the defect, the grinding force of the area is increased and the feed speed is reduced; the grinding trajectory of the defect area is regenerated and spliced ​​with the original grinding trajectory; the grinding execution end re-grinds the defect area according to the re-planned operation path.

[0116] By using machine vision to capture images of the finished workpiece surface, extracting texture and roughness features, and combining this with a deep learning model, the system enables intelligent assessment of polished surface quality and defect identification, increasing the automation of polishing quality inspection. Based on the identified defect information, the work path is replanned and the defective area is re-polished, enabling automatic defect repair, improving polishing quality, and reducing manual rework and quality issues. This method also provides a foundation for closed-loop optimization of the polishing process. By using defect identification results as feedback, polishing parameters and paths are optimized, continuously improving polishing results.

[0117] Example 2

[0118] Reference Figure 2 , which is the second embodiment of the present application, provides an adaptive force-controlled grinding system for complex surface processing robots.

[0119] The system includes a data acquisition module, a trajectory calculation module, an operation parameter calculation module, an impedance control module and a defect detection module.

[0120] The data acquisition module acquires three-dimensional point cloud data of the workpiece to be processed, grids the point cloud data to obtain a grid model, and calculates characteristic parameters of the normal vector and curvature of the surface of the workpiece to be processed in the grid model.

[0121] The trajectory calculation module calculates the working area of ​​the workpiece to be processed based on the characteristic parameters, generates motion path points in each working area, and generates the working trajectory of the workpiece to be processed.

[0122] The operation parameter calculation module calculates the operation parameters of the target force, feed speed, action depth and operation rhythm of the operation area according to the normal vector and curvature of the workpiece to be processed.

[0123] The impedance control module controls the robot to move according to the planned operation trajectory, monitors the applied force in real time through a force sensor, and adopts an adaptive impedance control algorithm to dynamically adjust the stiffness and damping parameters of the robot during operation based on the feedback information of the applied force.

[0124] The defect detection module obtains image data of the processed workpiece surface, extracts texture and roughness features in the image data, and builds a deep learning model to identify processing defects. When a processing defect is identified, the operation path is readjusted and planned.

[0125] The above sequence of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the sequence specifically described above unless otherwise specifically stated.

[0126] In addition, in some embodiments, the present application may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers a recording medium storing a program for executing the method according to the present application.

[0127] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0128] The above-described specific embodiments further illustrate the purpose, technical solutions, and beneficial effects of this application. It should be understood that the above description is merely a specific embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. An adaptive force-controlled grinding method for complex surface processing robots, characterized in that: include: Obtaining three-dimensional point cloud data of the workpiece to be processed, meshing the point cloud data to obtain a mesh model, and calculating characteristic parameters of the normal vector and curvature of the workpiece to be processed in the mesh model; Based on the characteristic parameters, the working area of ​​the workpiece to be processed is calculated, and the motion path points are generated in each working area to generate the working trajectory of the workpiece to be processed; Calculating the target force, feed rate, depth of action, and operation rhythm parameters of the operation area according to the normal vector and curvature of the workpiece to be processed; The robot is controlled to move along the planned operating trajectory, and the applied force is monitored in real time through a force sensor. An adaptive impedance control algorithm is used to dynamically adjust the stiffness and damping parameters of the robot during operation based on the feedback information of the applied force. Obtain image data of the processed workpiece surface, extract texture and roughness features from the image data, and build a deep learning model to identify processing defects; when processing defects are identified, readjust the operation path and plan it.

2. The adaptive force-controlled grinding method for complex surface processing robots according to claim 1, characterized in that: Scan the workpiece to be processed at multiple angles to obtain point cloud data of the workpiece to be processed; The Poisson surface reconstruction algorithm is used to mesh the point cloud and generate a triangular mesh model of the workpiece to be processed; Calculate the characteristic parameters of the normal vector and curvature in the mesh model of the workpiece to be processed.

3. The adaptive force-controlled grinding method for complex surface processing robots according to claim 2, characterized in that: Set the curvature threshold and normal vector angle threshold, and calculate the Gaussian curvature and normal vector of each point on the mesh model: If the curvature of a point is greater than the curvature threshold, the point is marked as a high curvature point. For each high curvature point, the angle between the normal vector of the high curvature point and the positive direction of the Z axis is calculated; If the normal vector angle is greater than the normal vector angle threshold, the high curvature point is on a large slope surface; Mark the points that meet both the curvature threshold and the normal vector angle threshold as points to be processed; Taking each point to be processed as a seed point, the region growing algorithm is used to extract the area to be processed; The area of ​​the extracted area to be operated is calculated. If it is greater than the set area threshold, the area to be operated is determined to be a valid area to be operated. Otherwise, the area is eliminated.

4. The adaptive force-controlled grinding method for complex surface processing robots according to claim 3, characterized in that: Calculate the number of sampling points inside the area to be operated; The Poisson disk sampling algorithm is used to generate uniformly distributed sampling points in the area to be operated; In the Poisson disk sampling algorithm, the area to be operated, the number of sampling points and the minimum sampling distance are input, and the operation path point set is output.

5. The adaptive force-controlled grinding method for complex surface processing robots according to claim 4, characterized in that: The steps of the Poisson disk sampling algorithm include: Step a: Randomly select an initial point in the area to be operated and add it to the path point set; Step b: With the initial point as the center and the radius set, a sampling disk is generated in the area to be operated; Step c: Select a random point in the sampling disk and calculate the distance between the random point and all points in the path point set; if the distance between the random point and any point is less than the set radius, reject the random point and return to step c; otherwise, accept the random point and add it to the path point set; Step d: Repeat steps b and c until no new point is accepted within the set number of times; Step e: When all sampling points are screened, sampling ends and the path point set is output.

6. The adaptive force-controlled grinding method for complex surface processing robots according to claim 5, characterized in that: Number and sort the points in the path point set to obtain an ordered sequence of operation path points; Between adjacent path points, cubic spline interpolation is used to generate a smooth operation path curve; Connect the operation path curves of all the areas to be operated in sequence to obtain a complete operation path.

7. The adaptive force-controlled grinding method for complex surface processing robots according to claim 6, characterized in that: Calculating operation parameters according to the normal vector and curvature of the area to be operated, wherein the operation parameters include: grinding force, feed speed, grinding depth and grinding time; The grinding actuator moves along the planned working path and monitors the grinding force in real time through a force sensor. An adaptive impedance control algorithm is used to dynamically adjust the stiffness and damping parameters of the grinding actuator based on feedback from the grinding force. The expected impedance model is constructed to calculate the expected trajectory acceleration of the grinding actuator. The expected trajectory acceleration is input into the robot's motion controller to perform impedance control on the grinding actuator.

8. The adaptive force-controlled grinding method for complex surface processing robots according to claim 7, characterized in that: Acquire surface image data of the processed workpiece, use an industrial camera to image the surface of the processed workpiece, and acquire surface image data; Extract the texture and roughness features from the image data of the finished workpiece surface, and use the gray-level co-occurrence matrix (GLCM) to extract the texture features of the image; The surface image of the workpiece is annotated after processing, the defects on the surface image of the workpiece are annotated, and a polished workpiece dataset is constructed for training the defect recognition model.

9. The adaptive force-controlled grinding method for complex surface processing robots according to claim 8, characterized in that: A defect recognition model is constructed based on the convolutional neural network (CNN) model. The extracted image feature vector is used as the input of the CNN model, and the defect category is used as the output. The CNN model is trained using the constructed polished workpiece dataset. Use the trained CNN model to identify defects on the surface image of the newly processed workpiece. If grinding defects are identified, the location and range information of the defect area are extracted. Map the defect area back to the original workpiece model to obtain the three-dimensional position of the defect on the workpiece; Replan the operation path in the defect area based on the defect location and scope.

10. An adaptive force-controlled grinding system for a complex surface machining robot, which is used to implement the adaptive force-controlled grinding method for a complex surface machining robot according to any one of claims 1 to 9, characterized in that: Including data acquisition module, trajectory calculation module, operation parameter calculation module, impedance control module and defect detection module; A data acquisition module acquires three-dimensional point cloud data of the workpiece to be processed, performs gridding processing on the point cloud data to obtain a grid model, and calculates characteristic parameters of the normal vector and curvature of the surface of the workpiece to be processed in the grid model; The trajectory calculation module calculates the working area of ​​the workpiece to be processed based on the characteristic parameters, generates motion path points in each working area, and generates the working trajectory of the workpiece to be processed; An operation parameter calculation module calculates the operation parameters of the target force, feed speed, action depth and operation rhythm of the operation area according to the normal vector and curvature of the workpiece to be processed; The impedance control module controls the robot's movement according to the planned operation trajectory, monitors the applied force in real time through a force sensor, and uses an adaptive impedance control algorithm to dynamically adjust the robot's stiffness and damping parameters during operation based on the applied force feedback information. The defect detection module obtains image data of the processed workpiece surface, extracts texture and roughness features in the image data, and builds a deep learning model to identify processing defects. When a processing defect is identified, the operation path is readjusted and planned.

Images