Polishing robot intelligent polishing head of rigid-flexible coupling structure and wall surface detection method

Abstract

The application provides a polishing robot intelligent polishing head with rigid-flexible coupling structure and a wall surface detection method, which is composed of a base, a distance sensor fixing seat, a pressure sensor one fixing seat, a pressure sensor one, a motor fixing seat, a motor, a motor output shaft connecting sheet, a damping compression spring, a polishing sheet, a damping force measuring unit and a distance sensor, the damping force measuring unit is composed of a compression spring group, a compression spring slide, a connecting plate, a sensor fixing seat, a connecting bolt, a pressure sensor array and a universal wheel, the distance sensor fixing seat and the pressure sensor one fixing seat are fixed on the base in sequence, the bottom of the pressure sensor one is fixed on the pressure sensor one fixing seat, the motor fixing seat is fixed on the top of the pressure sensor one, the bottom of the motor is fixed on the motor fixing seat, the motor output shaft connecting sheet is fixed on the motor output shaft, the bottom end of the damping compression spring is fixed on the motor output shaft connecting sheet, and the top end is fixed on the polishing sheet. The polishing head can measure the distance and the force to realize intelligent polishing of a large plane.

Description

Technical Field

[0001] This invention relates to the fields of intelligent manufacturing, intelligent robots, and bionic mechanisms, specifically to an intelligent grinding head for a grinding robot with a rigid-flexible coupling structure and a wall surface detection method. Background Technology

[0002] Wall sanding plays a crucial role in building construction, directly impacting the building's appearance and the smooth progress of subsequent processes. As the construction industry increasingly demands higher efficiency and quality, traditional manual sanding methods, due to their high labor intensity, low efficiency, and inconsistent quality, are gradually being replaced by intelligent construction robots. These robots have achieved significant results in improving work consistency and reducing worker fatigue. However, the adaptability of sanding heads to wall surfaces still faces numerous technical challenges.

[0003] Currently, most sanding robots on the market use sanding heads that lack sufficient flexibility when in contact with walls. This lack of flexibility prevents the sanding head from adaptively adjusting its contact state when facing minor unevenness or material inhomogeneity on the wall, thus affecting the stability of the sanding effect. The lack of sufficient flexibility limits the sanding head's ability to adapt to complex wall curves, leading to poor contact and consequently over-cutting and under-cutting. Over-cutting occurs when the angle and pressure applied by the sanding head are improper, removing too much material and causing unevenness on the wall surface; while under-cutting fails to effectively remove the target material, leaving uneven bumps on the surface. This not only affects the flatness and smoothness of the wall surface but may also require subsequent repair measures, increasing construction costs and time.

[0004] For example, patent CN 116372909A proposes a proportional damping control method for a grinding robot based on force sensors. This robot can collect wall height data through sensors and control the grinding head to adjust its position, thereby reducing the impact force on the workpiece at the moment of contact. Finally, it accurately controls the tool to grind the workpiece according to the desired force. This design mainly relies on sensors to provide feedback control to the wall. It lacks a rigid-flexible coupling structure and cannot achieve flexible following on walls with complex curvatures, resulting in a serious control lag effect.

[0005] On the other hand, patent CN 116652740A relates to a spring-loaded large-area grinding mechanism, grinding robot and grinding method, which overcomes the technical defects of the prior art when grinding workpieces with weld seams, which are prone to damage to weld seams, incomplete grinding and uneven grinding. However, the flexible component structure of this solution is simple, and it has the limitation of rigidity and flexibility imbalance when facing large areas, multi-varying areas and complex curvature walls, and cannot adapt to grinding large curved surfaces.

[0006] Furthermore, existing grinding head designs typically lack a rigid-flexible coupling structure, failing to provide flexible responsiveness while ensuring sufficient strength and stability. While rigid structures perform well in high-precision grinding, they are prone to vibration and unnecessary wear when dealing with irregular and complex wall surfaces due to their lack of flexibility. Therefore, effectively integrating rigidity and flexibility in grinding head design to ensure both durability and adaptability to wall variations is crucial for improving the performance of grinding robots.

[0007] To address the aforementioned problems, this invention proposes an intelligent grinding head for a grinding robot based on a rigid-flexible coupling structure, aiming to fundamentally improve the contact quality between the grinding head and the wall surface. This grinding head integrates a rigid frame and flexible components, maintaining structural stability while adapting to the undulations and shape changes of the wall surface. By introducing flexible damping elements and a spring array, the ability to mitigate vibrations and adjust dynamic contact forces is greatly enhanced, effectively protecting the wall surface while ensuring excellent grinding results during operation. This innovative design aims to provide a more uniform surface finish, reducing over-cutting and under-cutting phenomena, and offering a more efficient and reliable solution for the construction industry. Summary of the Invention

[0008] To solve the above-mentioned technical problems, this invention proposes a smart grinding head for a grinding robot with a rigid-flexible coupling structure and a wall detection method, which consists of a base, a distance sensor mounting base, a pressure sensor mounting base, a pressure sensor, a motor mounting base, a motor, a motor output shaft connecting plate, a shock-absorbing compression spring, a grinding disc, a shock-absorbing force measuring unit, and a distance sensor.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0010] The intelligent grinding head of the grinding robot with rigid-flexible coupling structure consists of a base, a distance sensor mounting base, a pressure sensor mounting base, a pressure sensor, a motor mounting base, a motor, a motor output shaft connecting piece, a shock-absorbing compression spring, a grinding disc, a shock-absorbing force measuring unit, and a distance sensor.

[0011] The distance sensor mounting base and the pressure sensor mounting base are sequentially fixed above the base; the bottom of the pressure sensor is fixed on the pressure sensor mounting base; the motor mounting base is fixed on the top of the pressure sensor; the bottom of the motor is fixed on the motor mounting base; the motor output shaft connecting piece is fixed on the motor output shaft; the bottom end of the shock-absorbing compression spring is fixed on the motor output shaft connecting piece; the top end of the shock-absorbing compression spring is fixed on the grinding disc.

[0012] The shock absorption force measuring unit consists of a compression spring assembly, a compression spring slide rod, a connecting plate, a sensor mounting base, connecting bolts, a pressure sensor array, and casters.

[0013] Each spring in the compression spring assembly is equipped with a spring slide rod. The bottom end of the compression spring assembly is fixed to the motor mounting base, and the top end of the compression spring assembly is fixed to the connecting plate. The sensor mounting base is fixed to the connecting plate by the connecting bolts. The pressure sensor array is fixed to the sensor mounting base, and the caster wheel is fixed to the pressure sensor array. The pressure sensor array includes at least two pressure sensors distributed circumferentially.

[0014] The ranging sensor is fixed on the ranging sensor mounting base.

[0015] As a further improvement to the grinding head of the present invention, the distance measuring sensor has three components: a first distance measuring sensor, a second distance measuring sensor, and a third distance measuring sensor, which are arranged in a circular pattern.

[0016] The intelligent grinding head of this invention detects the distance between the grinding head and the wall surface through multiple distance sensors, thereby adjusting the distance between the grinding head and the wall surface during the grinding process; the intelligent grinding head also measures the distributed pressure between the grinding disc and the wall surface through multiple pressure sensors, thereby detecting changes in the unevenness of the wall surface and assisting in the intelligent control of the grinding head.

[0017] This invention provides a wall surface detection method for a smart grinding head of a grinding robot with a rigid-flexible coupling structure. The specific steps are as follows:

[0018] S1: Reference plane identification and establishment;

[0019] The identification and establishment of the reference plane includes the transformation between the wall and the robot coordinate system, and the optimization calculation of the reference plane based on the z-component values ​​of the wall.

[0020] S2: Wall surface unevenness detection;

[0021] The wall surface protrusion detection algorithm requires a multi-sensor information fusion strategy to collect data from a range sensor and pressure sensor array in real time; the range sensor provides three-dimensional point cloud data of the wall surface. Real-time detection of the distance between the grinding head and the wall {d R ,d L ,d B Changes; pressure sensor array detects the distribution of contact force on the wall surface.

[0022] The wall surface protrusion detection algorithm needs to establish a mapping relationship between distance measurement data and force data to realize local wall surface morphology reconstruction.

[0023] S3: Grinding path planning and force control;

[0024] Based on the wall surface unevenness detection results, a wall surface height field map is constructed; gradient analysis is performed on the height field map to identify the boundaries of uneven regions.

[0025] S4: Real-time wall monitoring and dynamic adjustment;

[0026] During the polishing process, distance measurement data is collected in real time. and stress data Calculate the current real-time distance sensor's position on the wall (z). i Components and target z i 0 deviation e i Based on deviation e i Dynamically adjust the grinding head posture R and contact force

[0027] A force-position hybrid contact model was established to achieve a dynamic grinding effect. The model is as follows:

[0028]

[0029] in: For the force at contact node i, and e represents the inertia, damping, and stiffness of node i, respectively. i =x i -x0 represents the generalized position of node i. and These represent its velocity and acceleration, respectively.

[0030] S5: Data Analysis and Quality Assessment;

[0031] Set evaluation indicators for wall surface flatness, including average roughness R. a Maximum height difference R z Local flatness R local Overall flatness R global Scan and inspect the polished area to generate a quality report;

[0032] Establish a grinding quality database and collect data on the combination of parameters P = {F} during the process. d ,T d ,v d ,Θ x ,Θ y}, corresponding to the generated quality result Q={R a R max R local R global}, establish a parameter-quality dataset D = {(P i Qi )};

[0033] Then, the parameters are continuously optimized and refined through machine learning;

[0034] It enables visualization of the polishing process and provides a heat map of polishing quality to intuitively reflect the polishing effect;

[0035] S6: Determine the reference plane between the grinding head and the wall surface using the position information;

[0036] S7: Construct a wall point cloud model based on distance sensor data;

[0037] S8: Establish the mapping relationship between distance measurement data and force data to realize the reconstruction of local wall surface morphology;

[0038] S9: Identify the boundaries of concave and convex regions based on the relationship between the wall surface S(x,y,z) and the reference plane P(x,y,z);

[0039] S10: Eliminate noise effects and extract wall surface features through data filtering and feature extraction algorithms;

[0040] S11: Based on the uneven features of the wall surface, a hybrid contact model combining force and position is used to achieve dynamic polishing.

[0041] As a further improvement to the detection method of the present invention, step S1 is specifically as follows:

[0042] (1) Use a distance sensor to collect a measurement point cloud P = {p1, p2, ..., p} on the wall surface. n}, the three-dimensional spatial coordinates of each point are p i =(x i ,y i ,z i The wall surface is parameterized by x and y coordinates. The z component can be expressed as z = f(x, y). The coordinates of the measured points on the wall surface can be expressed as δ(x, y) = [x, y, z]. T ;

[0043] (2) The wall surface is parametrically defined as S * :={p∈R 3 |p=δ(x,y)};The process of finding the reference plane is transformed into an optimization problem of z-component values. The expression for minimizing the z-component values ​​of the wall is:

[0044] minmize:

[0045] subjectto:

[0046]

[0047] Among them, f i(x i ,y i ) represents the z-direction value in the wall coordinate system, f p (x p ,y p () represents the optimal reference z-component; by minimizing the difference between the target and other z-components, the simulated annealing algorithm can optimize the function to find the optimal solution f. p (x p ,y p );

[0048] (3) Define the reference plane as δ p (x,y)=[x p y p f p (x,y)] T The normal vector is n p =[00f(x p ,y p )] T The reference point is p. p =[x p y p 0] T The desired reference coordinate system is defined as O. p (x p ,y p ,z p The Z-axis of the reference coordinate system is perpendicular to the wall and points in the direction of the grinding head;

[0049] (4) The reference plane must be aligned with the robot's base coordinate system O. v (x v ,y v ,z v Establish the transformation relationship to obtain the position of the wall point P on the reference plane. The coordinate transformation matrix T is as follows:

[0050]

[0051] Where: P v ∈R 3 This indicates the position of the point in the robot's base coordinate system.

[0052] As a further improvement to the detection method of the present invention, step S2 establishes a mapping relationship between distance measurement data and force data to achieve local wall surface morphology reconstruction. The specific steps are as follows:

[0053] (1) Based on the wall data from the distance sensor Establish a real-time local fitting plane B, and calculate the normal vector of plane B as follows: The normal vector of the locally fitted plane B can be expressed as n Bp =(n Bx ,nBy ,n Bz );

[0054] (2) Establish a coordinate system O on the reference plane A. p (x p ,y p ,z p ), where the z-axis is the normal vector to plane A. A = (0,0,1) alignment; to make n A With n Bp Alignment, the end effector must be around O p Rotation along the x, y, and z axes, the transformation matrix is:

[0055] n ′ Bp =R z (γ)·R y (β)·R x (α)·n A s

[0056] Where: the rotation angles of plane A around the z-axis, y-axis, and x-axis are respectively defined as follows: β=arctan(n By / n Bx ), α=arctan(n By / n Bz ),. Normal vector n B ′ p Equivalent to n Bp ;

[0057] (3) The unevenness of the local wall surface is determined only by the rotational response of plane A around the y and x axes, without considering the rotation angle of plane A around the z axis. Further, the contact forces between γ, α and the wall surface are established. Relationship, that is, establishing and F L ext The mapping relationship between them is as follows:

[0058]

[0059] Among them: contact force and There is a coupling relationship; the initial rotation angle deviations of the x-axis and y-axis are e respectively. α =α d -α0 and e β =β d -β0.

[0060] As a further improvement to the detection method of the present invention, the specific steps of step S3, the boundary of the concave and convex region, are as follows:

[0061] (1) The relationship between the wall surface S(x,y,z) and the reference plane P(x,y,z) is projected onto the O-xy plane of the reference coordinate system. Further, the wall surface S(x,y,z) and the reference plane P(x,y,z) are simplified to a curve S(x,z) and a straight line P(x,z), where V si The tangent line to curve S, e si Indicates the normal line;

[0062] (2) This path passes through the interval [0, s end The data is discretized into N segments, resulting in N+1 grid points:

[0063] 0 = s0, s1, s2...s N-1 ,s N :=s end

[0064] (3) Perform a mesh search along the path in the convex and concave regions to find boundary points and determine the points that satisfy the following constraints:

[0065] Subjectto:

[0066]

[0067] Where: Δ represents the upper tangent point of the curved surface and the plane, □ represents the lower tangent point, and ◇ and ○ represent the intersection points at the bottom and top, respectively. Path segments Δ-◇, Δ-Δ, ○-◇, and ○-Δ are convex regions. Path segments ◇-□, ◇-○, □-○, and □-□ represent concave regions;

[0068] Based on the distribution characteristics of uneven areas, the optimal polishing path is planned; for areas with different degrees of unevenness, the polishing intensity and polishing time are adaptively adjusted.

[0069] As a further improvement to the detection method of the present invention, step S5 then continuously optimizes the polishing parameters through machine learning. The learning method steps are as follows:

[0070] (1) Model building -- Supervised learning methods are used, such as random forest regression or support vector regression (SVR), to build a correlation model between parameters P and quality Q, Q = f ML (P) The model is trained using historical data and its prediction accuracy is tested using a validation set.

[0071] (2) Parameter optimization -- Applying Bayesian optimization algorithm to obtain the optimal parameter P * =argminL(f ML (P)), where L is the loss function that measures the difference between the predicted value and the target value of the machine learning model;

[0072] (3) Online model update -- After each refinement, new data is added to D and incremental learning is performed to dynamically optimize the model.

[0073] As a further improvement to the detection method of the present invention, step S5 visualizes the polishing process and provides a heat map of polishing quality to intuitively reflect the polishing effect. The specific method steps are as follows:

[0074] (1) Using 3D modeling technology, the wall shape is rendered based on the real-time collected height field data F(x,y,t). The wall model can support free scaling, rotation and translation operations.

[0075] (2) Generate a heat map based on the height field data: The height difference is represented by a color gradient, such as red-blue, that is, the higher the height, the redder the area, and the lower the height, the bluer the area; at the same time, the current grinding intensity and contact pressure distribution are displayed.

[0076] (3) After each polishing is completed, the polished area is scanned and inspected to generate a quality report. The report includes flatness index, unevenness distribution map, and force control data record.

[0077] The technical solution of the present invention has the following beneficial effects:

[0078] (1) The grinding head of the inventor has the characteristics of a combination of rigid and flexible structure, which can passively adapt to the concave and convex changes of the wall surface, prevent the grinding head from being damaged by rigid collision with the wall surface and the mechanical arm that drives the grinding head, and reduce the difficulty of force control of the mechanical arm grinding.

[0079] (2) The wall detection method proposed in this invention realizes the automatic identification and establishment of the reference plane, providing a unified working coordinate system for the grinding robot, making the grinding operation more accurate and reliable.

[0080] (3) This invention significantly improves the accuracy of wall surface concavity and convexity detection by multi-source sensor information fusion, and can simultaneously sense macroscopic and microscopic concavity and convexity areas, providing comprehensive wall surface status information for intelligent polishing.

[0081] (4) The wall detection method of the present invention supports real-time monitoring and dynamic adjustment, enabling the sanding robot to adaptively adjust the sanding strategy according to the wall condition, thereby significantly improving the sanding quality and efficiency.

[0082] (5) This invention establishes a wall sanding quality assessment system, and continuously optimizes sanding parameters through data analysis and machine learning, thereby realizing the intelligentization and visualization of the sanding process. Attached Figure Description

[0083] Figure 1 This is a three-dimensional schematic diagram of the intelligent grinding head in an example of the present invention;

[0084] Figure 2 This is a three-dimensional exploded view of the intelligent grinding head in an example of the present invention;

[0085] Figure 3 This is a front view of the intelligent grinding head in an example of the present invention;

[0086] Figure 4 This is a rear view of the intelligent grinding head in an example of the present invention;

[0087] Figure 5 This is a right view of the intelligent grinding head in an example of the present invention;

[0088] Figure 6 This is a schematic diagram of the vibration damping and force measuring unit of the intelligent grinding head in an example of the present invention;

[0089] Figure 7 This is a flowchart of the wall reference plane identification and establishment process in an example of the present invention;

[0090] Figure 8 This is a schematic diagram of the wall surface unevenness detection method in an example of the present invention;

[0091] Figure 9 This is a schematic diagram of the wall surface unevenness detection principle based on a pressure sensor array in an example of the present invention;

[0092] Figure 10 This is a flowchart illustrating the real-time monitoring and dynamic adjustment of the grinding head on the wall surface in an example of the present invention.

[0093] Figure 11 This is a color gradient diagram of the wall height difference in an example of the present invention;

[0094] Figure label:

[0095] 1. Base; 2. Distance sensor mounting base; 3. Pressure sensor mounting base (1); 4. Pressure sensor (1); 5. Motor mounting base; 6. Motor; 7. Motor output shaft connecting piece; 8. Shock-absorbing compression spring; 9. Grinding disc; 10. Shock-absorbing force measuring unit; 10-1. Compression spring assembly; 10-2. Compression spring slide rod; 10-3. Connecting plate; 10-4. Sensor mounting base; 10-5. Connecting bolt; 10-6. Pressure sensor array; 10-7. Caster wheel; 11-1. First distance sensor; 11-2. Second distance sensor; 11-3. Third distance sensor. Detailed Implementation

[0096] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0097] See attached document Figure 1 Appendix Figure 2 Appendix Figure 3The intelligent grinding head consists of a base 1, a distance sensor mounting base 2, a pressure sensor mounting base 3, a pressure sensor 4, a motor mounting base 5, a motor 6, a motor output shaft connecting piece 7, a shock-absorbing compression spring 8, a grinding disc 9, a shock-absorbing force measuring unit 10, and a distance sensor.

[0098] See attached document Figure 1 Appendix Figure 2 Appendix Figure 6 The shock-absorbing force measuring unit 10 consists of a compression spring assembly 10-1, a compression spring slide rod 10-2, a connecting plate 10-3, a sensor fixing seat 10-4, a connecting bolt 10-5, a pressure sensor array 10-6, and a caster wheel 10-7.

[0099] See attached document Figure 1 Appendix Figure 2 Appendix Figure 3 Appendix Figure 4 Appendix Figure 5 The ranging sensor mounting base 2 is fixed to the base 1 with screws, and the pressure sensor 4 is fixed to the base 1 with screws; the bottom of the pressure sensor 4 is fixed to the pressure sensor mounting base 3; the motor mounting base 5 is fixed to the top of the pressure sensor 4; the bottom of the motor 6 is fixed to the motor mounting base 5; the motor output shaft connecting piece 6 is fixed to the output shaft of the motor 6; the bottom end of the shock-absorbing compression spring 8 is fixed to the motor output shaft connecting piece 7; and the top end of the shock-absorbing compression spring 8 is fixed to the grinding disc 9.

[0100] See attached document Figure 1 Appendix Figure 2 Appendix Figure 6 The compression spring assembly 10-1 includes 6 compression springs. The bottom end of the compression spring is fixed to the motor mounting base 5, and the top end of the compression spring is fixed to the connecting plate 10-3. The sensor mounting base 10-4 is fixed to the connecting plate 10-3 by the connecting bolt 10-5. The pressure sensor array includes 3 pressure sensors evenly arranged in a circle. The bottom end of the pressure sensor is fixed to the sensor mounting base 10-4. The universal wheels 10-7 include 3, and the 3 universal wheels are respectively fixed to the top ends of the 3 pressure sensors.

[0101] See attached document Figure 1 and attached Figure 2 The distance measuring sensor consists of three sensors: a first distance measuring sensor 11-1, a second distance measuring sensor 11-2, and a third distance measuring sensor 11-3, which are fixed at equal angles on the distance measuring sensor mounting base 2.

[0102] See attached document Figure 7The reference plane identification and establishment method of the present invention specifically includes the following steps:

[0103] Step 1: Initialize and calibrate the ranging sensor to ensure its measurement accuracy and range; establish the transformation relationship between the sensor coordinate system and the robot base coordinate system; set the ranging data sampling frequency and data preprocessing parameters.

[0104] Step 2: Multi-point scanning of the wall surface

[0105] Control the grinding robot to perform a grid scan within a preset area; collect no fewer than 100 ranging points and record the three-dimensional coordinates (x, y, z) of each point. i ,y i ,z i ); outlier detection and removal are performed on the collected point cloud data.

[0106] Step 3: Reference plane fitting

[0107] Based on the optimization of the z-component values ​​of the wall, the optimal solution f is obtained. p (x p ,y p Based on the equation ax + by + cz + d = 0, calculate the fitting error. If the error exceeds the threshold, increase the sampling points and refit. Determine the optimal reference plane δ. p (x,y)=[x p y p f p (x,y)] T And calculate the normal vector n = (0, 0, f p (x,y)).

[0108] Step 4: Establishing the coordinate system

[0109] The origin is defined as the point on the reference plane projected from the robot's current position; the X-axis is defined as parallel to the water surface, the Y-axis as perpendicular to the X-axis and within the reference plane, and the Z-axis as along the direction of the reference plane's normal vector; a coordinate system O is established from the robot's base coordinate system. v (x v ,y v ,z v To the reference plane coordinate system O p (x p ,y p ,z p The transformation matrix T of ).

[0110] See attached document Figure 8 and attached Figure 9 The wall surface unevenness detection method of the present invention specifically includes the following steps:

[0111] Step 1: Multi-source sensor information acquisition

[0112] The distance sensor collects distance data from the wall at a frequency of 25Hz; the pressure sensor array collects contact force data at a frequency of 100Hz; and records the position and attitude information of the grinding head in the robot's base coordinate system.

[0113] Step 2: Point Cloud Model Construction

[0114] Convert the ranging data to a reference plane coordinate system O. p (x p ,y p ,z p The three-dimensional point cloud is obtained by interpolating the robot's motion; the point cloud density is increased by interpolating the robot's motion; and the point cloud noise is eliminated by applying a bilateral filtering algorithm.

[0115] Step 3: Extraction of wall morphological features

[0116] Calculate the height deviation Δh between the point cloud data and the reference plane. i Set a threshold τ for determining concavity / convexity; when |⊿h i When |>τ, mark it as a concave or convex point; perform region clustering on the concave and convex points to identify continuous concave and convex regions; calculate the area, depth / height, and boundary contour of each concave or convex region.

[0117] Step 4: Pressure Distribution Analysis

[0118] Spatial interpolation is performed on the pressure sensor array data to generate a pressure distribution map; abnormal pressure areas are identified and compared with the point cloud unevenness area for verification; a quantitative relationship model between pressure value and surface unevenness is established; for areas that cannot be directly contacted by the pressure sensor, inference is made through the pressure gradient of adjacent areas.

[0119] Step 5: Multi-source information fusion

[0120] Design a multi-sensor data fusion algorithm based on Kalman filtering; align distance measurement data and pressure data in time and space; improve the accuracy and robustness of concavity and convexity detection through information complementarity; generate a comprehensive wall concavity and convexity map for subsequent polishing planning.

[0121] See attached document Figure 10 The wall surface real-time monitoring and dynamic adjustment method of the present invention specifically includes the following steps:

[0122] Step 1: Real-time Data Acquisition and Processing

[0123] During the polishing process, distance and pressure data are continuously collected; the wall surface status information is updated in real time through a sliding window algorithm; and the difference between the current wall surface status and the target flatness is calculated.

[0124] Step Two: Dynamically Adjusting the Polishing Strategy

[0125] Adjust the following sanding parameters based on the differences in wall surface condition:

[0126] 1. Grinding head posture: By controlling the posture of the robot's end effector, the grinding head is kept at the optimal contact angle with the local wall surface;

[0127] 2. Contact pressure: Increase pressure on raised areas and decrease pressure on recessed areas;

[0128] 3. Polishing time: The polishing time is dynamically allocated according to the unevenness of the area;

[0129] 4. Grinding trajectory: Automatically generates spiral or cross grinding trajectories for complex uneven areas.

[0130] Step 3: Real-time assessment of polishing quality

[0131] Define the wall surface flatness index R a (average roughness) and R z (Maximum height difference); Evaluate the current R of the polished area through rapid scanning. a and R z The value is compared with the preset target value to determine whether further polishing is needed; a heat map of polishing quality is generated to visually display the polishing effect.

[0132] Step 4: Self-learning optimization

[0133] Record the relationship between different wall materials, initial state and final sanding effect; establish a mapping between wall state and optimal sanding parameters through support vector regression algorithm; continuously optimize sanding strategy based on historical data to improve sanding efficiency and quality.

[0134] This invention also provides a specific method for force detection of the unevenness of a wall surface, including:

[0135] Step 1: Force distribution sampling

[0136] Control the grinding head to slide along a predetermined path with a fixed pressure; record the force distribution data F(xy) of the pressure sensor array at different positions; generate a force distribution map of the wall surface.

[0137] Step 2: Force Gradient Analysis

[0138] Calculate the spatial gradient of force distribution Areas with large gradients correspond to locations where the wall surface undergoes drastic changes in unevenness; the direction of the gradient indicates whether it is a depression or a convexity.

[0139] Step 3: Locating the concave and convex positions

[0140] The wall surface S(x,y,z) and the reference plane P(x,y,z) are simplified to a curve S(x,z) and a straight line P(x,z), where V siThe tangent line to curve S, e si Represent the normal; mark the intersection of P(x,z) and S(x,z) as the concave / convex boundary point;

[0141] The location and extent of the concave and convex regions are determined by closing the curves at the boundary points; the center point and extreme points of each concave and convex region are calculated.

[0142] Step 4: Concavity / Concavity Depth Estimation

[0143] Establish the mapping relationship between force and the depth of concavity / concavity: d(x,y)=α·F(x,y)+β; where d(x,y) is the depth of concavity / concavity, and α and β are empirical parameters; determine the values ​​of α and β through calibration experiments; generate as follows Figure 11 The diagram shows the depth of the wall surface's unevenness.

[0144] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A rigid-flexible coupling structure for a grinding robot intelligent grinding head, comprising a base (1), a distance sensor mounting base (2), a pressure sensor mounting base (3), a pressure sensor (4), a motor mounting base (5), a motor (6), a motor output shaft connecting piece (7), a shock-absorbing compression spring (8), a grinding disc (9), a shock-absorbing force measuring unit (10), and a distance sensor, characterized in that... ； The distance sensor mounting base (2) and the pressure sensor mounting base (3) are fixed to the base (1) in sequence; the bottom of the pressure sensor (4) is fixed to the pressure sensor mounting base (3); the motor mounting base (5) is fixed to the top of the pressure sensor (4); the bottom of the motor (6) is fixed to the motor mounting base (5); the motor output shaft connecting piece (7) is fixed to the output shaft of the motor (6); the bottom end of the shock-absorbing compression spring (8) is fixed to the motor output shaft connecting piece (7); the top end of the shock-absorbing compression spring (8) is fixed to the grinding plate (9); The shock-absorbing force measuring unit (10) consists of a compression spring assembly (10-1), a compression spring slide rod (10-2), a connecting plate (10-3), a sensor mounting base (10-4), connecting bolts (10-5), a pressure sensor array (10-6), and a caster wheel (10-7). Each spring in the compression spring assembly (10-1) is equipped with a spring slide rod (10-2). The bottom end of the compression spring assembly (10-1) is fixed on the motor mounting base (5), and the top end of the compression spring assembly (10-1) is fixed on the connecting plate (10-3). The sensor mounting base (10-4) is fixed on the connecting plate (10-3) by the connecting bolt (10-5). The pressure sensor array (10-6) is fixed on the sensor mounting base (10-4), and the universal wheel (10-7) is fixed on the pressure sensor array (10-6). The pressure sensor array (10-6) includes at least two pressure sensors distributed circumferentially. The ranging sensor is fixed on the ranging sensor mounting base (2).

2. The intelligent grinding head for a grinding robot with a rigid-flexible coupling structure according to claim 1, characterized in that: The distance measuring sensors consist of three sensors: a first distance measuring sensor (11-1), a second distance measuring sensor (11-2), and a third distance measuring sensor (11-3), which are arranged in a circular pattern.

3. A wall surface detection method based on the intelligent grinding head of a grinding robot with a rigid-flexible coupling structure according to any one of claims 1-2, characterized in that, The specific steps are as follows: S1: Reference plane identification and establishment; The identification and establishment of the reference plane includes the transformation between the wall and the robot coordinate system, and the optimization calculation of the reference plane based on the z-component values ​​of the wall. S2: Wall surface unevenness detection; The wall surface protrusion detection algorithm requires a multi-sensor information fusion strategy to collect data from a range sensor and pressure sensor array in real time; the range sensor provides three-dimensional point cloud data of the wall surface. Real-time detection of the distance between the grinding head and the wall {d R ,d L ,d B Changes; pressure sensor array detects the distribution of contact force on the wall surface. The wall surface protrusion detection algorithm needs to establish a mapping relationship between distance measurement data and force data to realize local wall surface morphology reconstruction. S3: Grinding path planning and force control; Based on the wall surface unevenness detection results, a wall surface height field map is constructed; gradient analysis is performed on the height field map to identify the boundaries of uneven regions. S4: Real-time wall monitoring and dynamic adjustment; During the polishing process, distance measurement data is collected in real time. and stress data Calculate the current real-time distance sensor's position on the wall (z). i Components and target z i 0 deviation e i Based on deviation e i Dynamically adjust the grinding head posture R and contact force A force-position hybrid contact model was established to achieve a dynamic grinding effect. The model is as follows: in: For the force at contact node i, and e represents the inertia, damping, and stiffness of node i, respectively. i =x i -x0 represents the generalized position of node i. and These represent its velocity and acceleration, respectively. S5: Data Analysis and Quality Assessment; Set evaluation indicators for wall surface flatness, including average roughness R. a Maximum height difference R z Local flatness R local Overall flatness R gl obal; scans and inspects the polished area to generate a quality report; Establish a grinding quality database and collect data on the combination of parameters P = {F} during the process. d ,T d ,v d ,Θ x ,Θ y }, corresponding to the generated quality result Q={R a ,R max ,R local ,R global }, establish a parameter-quality dataset D = {(P i Q i )}; Then, the parameters are continuously optimized and refined through machine learning; It enables visualization of the polishing process and provides a heat map of polishing quality to intuitively reflect the polishing effect; S6: Determine the reference plane between the grinding head and the wall surface using the position information; S7: Construct a wall point cloud model based on distance sensor data; S8: Establish the mapping relationship between distance measurement data and force data to realize the reconstruction of local wall surface morphology; S9: Identify the boundaries of concave and convex regions based on the relationship between the wall surface S(x,y,z) and the reference plane P(x,y,z); S10: Eliminate noise effects and extract wall surface features through data filtering and feature extraction algorithms; S11: Based on the uneven features of the wall surface, a hybrid contact model combining force and position is used to achieve dynamic polishing.

4. The wall surface detection method for the intelligent grinding head of the grinding robot with a rigid-flexible coupling structure according to claim 3, characterized in that, The specific steps of step S1 are as follows: (1) Use a distance measuring sensor to collect a measurement point cloud P = {p1, p2, ..., p} on the wall surface. n }, where the three-dimensional spatial coordinates of each point are p i =(x i ,y i ,z i The wall surface is parameterized by x and y coordinates. The z component can be expressed as z = f(x, y). The coordinates of the measured points on the wall surface can be expressed as δ(x, y) = [x, y, z]. T ; (2) The wall surface is parametrically defined as S * :={p∈R 3 |p=δ(x,y)};The process of finding the reference plane is transformed into an optimization problem of z-component values. The expression for minimizing the z-component values ​​of the wall is: Among them, f i (x i ,y i ) represents the z-direction value in the wall coordinate system, f p (x p ,y p () represents the optimal reference z-component; by minimizing the difference between the target and other z-components, the simulated annealing algorithm can optimize the function to find the optimal solution f. p (x p ,y p ); (3) Define the reference plane as δ p (x,y)=[x p y p f p (x,y)] T The normal vector is n p =[00f(x p ,y p )] T The reference point is p. p =[x p y p 0] T The desired reference coordinate system is defined as O. p (x p ,y p ,z p The Z-axis of the reference coordinate system is perpendicular to the wall and points in the direction of the grinding head; (4) The reference plane must be aligned with the robot's base coordinate system O. v (x v ,y v ,z v Establish the transformation relationship to obtain the position of the wall point P on the reference plane. The coordinate transformation matrix T is as follows: Where: P v ∈R 3 This indicates the position of the point in the robot's base coordinate system.

5. The wall surface detection method for the intelligent grinding head of the grinding robot with a rigid-flexible coupling structure according to claim 3, characterized in that, Step S2 establishes a mapping relationship between distance measurement data and force data to achieve local wall surface morphology reconstruction. The specific steps are as follows: (1) Based on the wall data from the distance sensor Establish a real-time local fitting plane B, and calculate the normal vector of plane B as follows: The normal vector of the locally fitted plane B can be expressed as n Bp =(n Bx ,n By ,n Bz ); (2) Establish a coordinate system O on the reference plane A. p (x p ,y p ,z p ), where the z-axis is the normal vector to plane A. A = (0,0,1) alignment; to make n A With n Bp Alignment, the end effector must be around O p Rotation along the x, y, and z axes, the transformation matrix is: n′ Bp =R z (c)·R y (b)·R x (a)·n A s Where: the rotation angles of plane A around the z-axis, y-axis, and x-axis are respectively defined as follows: β=arctan(n By / n Bx ), α=arctan(n By / n Bz ),. Normal vector n′ Bp Equivalent to n Bp ; (3) The unevenness of the local wall surface is determined only by the rotational response of plane A around the y and x axes, without considering the rotation angle of plane A around the z axis. Further, the contact forces between γ, α and the wall surface are established. Relationship, that is, establishing and ext The mapping relationship between them is as follows: Among them: contact force and There is a coupling relationship; the initial rotation angle deviations of the x-axis and y-axis are e respectively. α =α d -α0 and e β =β d -β0.

6. The wall surface detection method for the intelligent grinding head of the grinding robot with a rigid-flexible coupling structure according to claim 3, characterized in that, The specific steps for defining the boundary of the concave-convex region in step S3 are as follows: (1) The relationship between the wall surface S(x,y,z) and the reference plane P(x,y,z) is projected onto the O-xy plane of the reference coordinate system. Further, the wall surface S(x,y,z) and the reference plane P(x,y,z) are simplified to a curve S(x,z) and a straight line P(x,z), where V si The tangent line to curve S, e si Indicates the normal line; (2) This path passes through the interval [0, s end The data is discretized into N segments, resulting in N+1 grid points: 0=:s0,s1,s2...s N-1 ,s N :=s end (3) Perform a mesh search along the path in the convex and concave regions to find boundary points and determine the points that satisfy the following constraints: Subjectto: Where: Δ represents the upper tangent point of the curved surface and the plane, □ represents the lower tangent point, ◇ and ○ represent the intersection points below and above, respectively; path segments Δ-◇, Δ-Δ, ○-◇ and ○-Δ are convex regions, and path segments ◇-□, ◇-○, □-○ and □-□ represent concave regions; Based on the distribution characteristics of uneven areas, the optimal polishing path is planned; for areas with different degrees of unevenness, the polishing intensity and polishing time are adaptively adjusted.

7. The wall surface detection method for the intelligent grinding head of the grinding robot with a rigid-flexible coupling structure according to claim 3, characterized in that, Step S5 then uses machine learning to continuously optimize and refine the parameters. The learning method steps are as follows: (1) Model building -- Supervised learning methods are used, such as random forest regression or support vector regression (SVR), to build a correlation model between parameters P and quality Q, Q = f ML (P), the model is trained using historical data and the prediction accuracy is tested using a validation set; (2) Parameter optimization -- Applying Bayesian optimization algorithm to obtain the optimal parameter P * =argminL(f ML (P)), where L is the loss function that measures the difference between the predicted value and the target value of the machine learning model; (3) Online model update -- After each refinement, new data is added to D and incremental learning is performed to dynamically optimize the model.

8. The wall surface detection method for the intelligent grinding head of the grinding robot with a rigid-flexible coupling structure according to claim 3, characterized in that, Step S5 visualizes the polishing process, providing a heat map of polishing quality that intuitively reflects the polishing effect. The specific steps are as follows: (1) Using 3D modeling technology, the wall shape is rendered based on the real-time collected height field data F(x,y,t). The wall model can support free scaling, rotation and translation operations. (2) Generate a heat map based on the height field data: The height difference is represented by a color gradient, such as red-blue, that is, the higher the height, the redder the area, and the lower the height, the bluer the area; at the same time, the current grinding intensity and contact pressure distribution are displayed. (3) After each polishing is completed, the polished area is scanned and inspected to generate a quality report. The report includes flatness index, unevenness distribution map, and force control data record.

Images