Combined electric porcelain insulator automatic glazing track and flow control method

Abstract

The application provides a combined electric porcelain insulator automatic glazing track and flow control method, which comprises the following steps: obtaining a surface model of a porcelain piece to be glazed and dividing the surface into surface partitions, collecting joint static torque before glazing, and establishing a force arm response relationship between each surface partition and a robot joint; generating an initial glazing track and an initial flow instruction according to the surface model, combining the maximum stiffness direction of the end, the normal stiffness of the end and the glazing hose traction, and determining a target joint configuration sequence by using an ant colony algorithm; executing initial glazing and collecting joint static torque after glazing, inverting the new glaze quality of each surface partition according to the joint static torque before and after glazing and the force arm response relationship, and determining under-glazed partitions and compensation glaze amount. The application reduces the influence of forming shrinkage, local deformation and clamping deviation on track planning by establishing a measured surface model of a porcelain piece to be glazed and dividing the surface into partitions, and improves the track adaptability of complex umbrella skirts and transition areas.

Description

Technical Field

[0001] This invention relates to the field of industrial robot program control technology, and in particular to an automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot. Background Technology

[0002] Composite porcelain insulators are typically composed of multiple porcelain components with skirts, necks, and annular transition surfaces. To improve their insulation and weather resistance, a uniform glaze layer needs to be formed on the surface of the porcelain components. Existing robotic glazing methods usually pre-set the spraying path according to a design model or taught trajectory and perform glazing at a fixed speed and flow rate. However, dimensional deviations and positional shifts may occur in the porcelain components during forming, drying, and clamping processes, and shading can easily form at the base of the skirts, the lower surface of the skirts, and between adjacent skirts. If a fixed spray distance, fixed speed, or fixed flow rate is still used, underglazing is likely to occur in areas of curvature variation, and localized overglazing may occur in the trajectory overlap area.

[0003] Secondly, for robots with redundant degrees of freedom, the same nozzle pose can be achieved by multiple joint configurations, and different joint configurations correspond to different end-effector stiffnesses. When the robot is in a joint configuration with low stiffness, the jet reaction force and the traction force of the glaze supply hose can easily cause the nozzle to deviate from the target position, thereby changing the actual spray distance and spray width. Existing methods usually select the joint configuration only based on whether the robot can reach the target pose, making it difficult to take into account the influence of end-effector stiffness, trajectory continuity, and external disturbance forces.

[0004] Therefore, this invention proposes an automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing an automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot includes the following steps:

[0008] S1. Obtain the surface model of the ceramic piece to be glazed, divide it into multiple surface partitions, determine the target glazing quality of each surface partition, control the robot to maintain multiple reference detection postures before glazing, collect the static torque of the joints before glazing, and establish the lever arm response relationship.

[0009] S2. Generate the initial glazing trajectory and initial flow command based on the surface model, generate candidate joint configurations for the trajectory points, and determine the target joint configuration sequence using the ant colony algorithm based on the direction of maximum end stiffness, end normal stiffness and glazing hose traction.

[0010] S3. Control the robot to execute the initial glazing trajectory according to the target joint configuration sequence, and adjust the glaze flow rate according to the initial flow command. After the glazing is completed, establish the glazing reference state, and collect the static torque of the joint after glazing according to the reference detection posture.

[0011] S4. Based on the static torque of the joint before glazing, the static torque of the joint after glazing, and the lever arm response relationship, the quality of the newly added glaze in each surface zone is inverted and compared with the target glazing quality to determine the under-glazed zones and the amount of glaze compensation.

[0012] S5. Generate compensation glazing trajectory and compensation flow instructions based on the underglaze zone and compensation glaze amount. Use ant colony algorithm to determine the compensation target joint configuration sequence, execute compensation glazing, and update the cumulative new glaze mass based on the compensated joint static torque until the target glazing quality is reached or the compensation termination condition is met.

[0013] S1 specifically includes: acquiring the surface model of the ceramic piece to be glazed, establishing the coordinate system of the ceramic piece, dividing the surface model into geometric partitions and circumferential subdivisions, determining the area, center position, average surface normal, and target glazing quality of each surface partition, and forming a set of surface partitions; generating candidate reference detection postures based on the set of surface partitions, selecting reference detection postures that meet the identifiable conditions based on the candidate lever arm response matrix, controlling the robot to enter the reference detection posture, and collecting the joint static torque and detection reference data before glazing; calculating the lever arm response coefficient based on the set of surface partitions, the reference detection posture, and the robot's kinematic parameters, correcting it using a known mass calibration block, and forming the final lever arm response matrix and lever arm response relationship.

[0014] S2 specifically includes: determining the glazing sequence based on the surface model, surface partition set, and target glazing quality; generating an initial glazing trajectory and initial flow command including nozzle position, spray direction, spray distance, velocity, and mass flow rate; solving for candidate joint configurations for trajectory points, eliminating configurations that do not meet the constraints of joint range, collision, singular configuration, and glazing hose bending; determining the configuration cost based on the direction of maximum end stiffness, end normal stiffness, jet reaction, and glazing hose traction; establishing a search structure based on trajectory points and candidate joint configurations; using an ant colony algorithm to search for a sequence of target joint configurations that meet continuous motion constraints; and binding these sequences with the initial glazing trajectory, initial flow command, and timing information to form an initial glazing control sequence.

[0015] S3 specifically includes: controlling the robot to execute the initial glazing according to the initial glazing control sequence, compensating for the glazing supply response delay, and collecting actual joint angles, actual nozzle positions, actual mass flow rates, glazing supply pressure, and glazing supply hose traction force to form an initial glazing execution record; after the initial glazing is completed, eliminating residual glaze, restoring the glazing supply pressure, glazing supply hose traction force, and clamping force to the reference range of the pre-glazing detection benchmark data, and controlling the robot to enter each benchmark detection posture within the detection time window; collecting the joint static torque after glazing in each benchmark detection posture, and screening the posture repeatability, glazing benchmark state, and torque stability to form an effective set of joint static torques after glazing and an effective initial glazing execution record.

[0016] S4 specifically includes: based on the static torque of the joint before glazing, the static torque of the joint after glazing, and the detection benchmark data, deducting the background torque caused by the difference in joint temperature, glazing pressure, and glazing hose traction force, to obtain the corrected set of new torques and the set of new torque uncertainties; based on the corrected set of new torques, the final lever arm response matrix, the surface partition adjacency relationship, and the effective initial glazing execution record, performing weighted inversion under the conditions of non-negative new glaze quality and limited reference adhesion quality, to obtain the new glaze quality distribution, quality inversion uncertainty, inversion residual, and effective quality inversion marker; comparing the new glaze quality distribution with the corresponding target glazing quality, and combining the allowable glazing deviation and quality inversion uncertainty, to determine the under-glazing partition, compensation glaze amount, under-glazing cause marker, and non-compensation marker.

[0017] S5 specifically includes: determining the compensation area and compensation spraying transfer efficiency based on the under-glazing zone, compensation glaze amount, quality inversion uncertainty, and non-compensation markings; generating a set of candidate compensation glazing trajectories and a set of candidate compensation flow instructions; generating candidate compensation joint configurations for candidate compensation glazing trajectory points; constructing search nodes with trajectory points, candidate compensation joint configurations, and candidate compensation mass flow levels; and using an ant colony algorithm to determine the compensation glazing control sequence; executing the compensation glazing control sequence; collecting the static torque of the joints after compensation; updating the newly added glaze mass, cumulative newly added glaze mass, and compensation spraying transfer efficiency in this round of compensation; and ending compensation when the target glazing quality is reached or the compensation termination condition is met.

[0018] The beneficial effects of this invention are as follows:

[0019] This invention reduces the impact of molding shrinkage, local deformation, and clamping deviations on trajectory planning by establishing a measured surface model of the ceramic part to be glazed and dividing it into zones, thereby improving the trajectory adaptability of complex umbrella skirts and transition areas. By jointly determining the nozzle trajectory, running speed, and mass flow rate, the glaze output is matched with the surface area, spray width, and transfer efficiency, reducing underglazing in areas of curvature variation and overglazing in trajectory overlap areas.

[0020] This invention reduces the normal deviation of the nozzle during glazing by comparing the maximum end stiffness direction, end normal stiffness, jet reaction, and glaze supply hose traction of candidate joint configurations, thereby improving the stability of the spray distance and spray direction. By collecting the joint static torque under multiple benchmark detection postures before and after glazing, and combining this with the lever arm response relationship to invert the quality of newly added glaze in each surface zone, the influence of wet glaze reflection and surface obstruction on the detection results can be reduced, enabling quantitative identification of local glazing conditions.

[0021] This invention identifies under-glazed zones by combining target glazing quality, quality inversion uncertainty, and allowable glazing deviation. This avoids misjudging normal fluctuations as under-glazing and provides a quantitative basis for compensation areas and the amount of compensation glaze. By generating a boundary-constrained compensation glazing trajectory for the under-glazed zones and re-performing joint static torque detection and quality inversion after compensation, the risks of insufficient compensation, repeated glazing, and local over-glazing can be reduced, forming an automatic glazing closed loop. Attached Figure Description

[0022] Figure 1 This is a flowchart of an automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example: Figure 1 As shown, this embodiment provides an automatic glazing trajectory and flow control method for a combined electric porcelain insulator glazing robot, including the following steps:

[0025] S1. Obtain the surface model of the ceramic piece to be glazed, divide it into multiple surface partitions, determine the target glazing quality of each surface partition, control the robot to maintain multiple reference detection postures before glazing, collect the static torque of the joints before glazing, and establish the lever arm response relationship.

[0026] S2. Generate the initial glazing trajectory and initial flow command based on the surface model, generate candidate joint configurations for the trajectory points, and determine the target joint configuration sequence using the ant colony algorithm based on the direction of maximum end stiffness, end normal stiffness and glazing hose traction.

[0027] S3. Control the robot to execute the initial glazing trajectory according to the target joint configuration sequence, and adjust the glaze flow rate according to the initial flow command. After the glazing is completed, establish the glazing reference state, and collect the static torque of the joint after glazing according to the reference detection posture.

[0028] S4. Based on the static torque of the joint before glazing, the static torque of the joint after glazing, and the lever arm response relationship, the quality of the newly added glaze in each surface zone is inverted and compared with the target glazing quality to determine the under-glazed zones and the amount of glaze compensation.

[0029] S5. Generate compensation glazing trajectory and compensation flow instructions based on the underglaze zone and compensation glaze amount. Use ant colony algorithm to determine the compensation target joint configuration sequence, execute compensation glazing, and update the cumulative new glaze mass based on the compensated joint static torque until the target glazing quality is reached or the compensation termination condition is met.

[0030] S1 specifically includes the following sub-steps:

[0031] S110. Obtain the surface model of the ceramic piece to be glazed, establish the coordinate system of the ceramic piece, and form a set of surface partitions. The surface model should preferably be the 3D model in the product design file; when the ceramic piece to be glazed has firing shrinkage, clamping eccentricity, or discrete body size, a structured light scanner, laser contour sensor, or robot end-effector 3D vision device is used to collect the measured point cloud of the current ceramic piece to be glazed. After outlier removal, hole repair, and triangular mesh reconstruction, a measured surface model is formed.

[0032] The central axis of the ceramic part is determined based on the fitting results of the central hole, cylindrical neck, or multiple coaxial rotating surfaces. The intersection of the central axis and the mounting end face is taken as the origin of the ceramic part coordinate system. The direction of the central axis pointing from the mounting end face to the free end is taken as the positive axial direction. The direction corresponding to the ceramic part positioning groove, assembly notch, or fixture positioning pin is taken as the circumferential zero-degree direction. The directions of the remaining coordinate axes are determined according to the right-hand rule. The coordinate transformation relationship between the ceramic part coordinate system and the robot end effector coordinate system is obtained through fixture calibration, and the coordinates of each grid vertex are transformed to the ceramic part coordinate system.

[0033] The surface is first divided according to the geometric boundaries of the skirt edge, skirt root, neck, and annular transition surface, and then further divided according to the circumferential angle. Locations with abrupt curvature changes or where the angle between the normals of adjacent grid surfaces exceeds the normal change threshold are defined as partition boundaries. The normal change threshold is derived from the manual annotation results of the geometric boundaries of ceramic pieces of the same model and the corresponding grid normal statistics. A surface partition refers to a continuous surface area with a unique partition number, surface area, area-weighted center location, area-weighted average surface normal, axial range, circumferential range, and target glazing quality.

[0034] The target glazing quality for the i-th surface zone is determined by the following formula:

[0035]

[0036] in, This represents the target glaze mass for the i-th surface partition, in grams. This indicates the wet density of the current batch of glaze slurry, derived from a density test conducted on the glaze slurry in the glaze supply tank after thorough stirring before glazing. The unit is g / cm³. 3 ; This represents the surface area of ​​the i-th surface region, which is obtained by summing all the triangular mesh faces within that surface region, and the unit is cm. 2 ; This indicates the target wet glaze thickness specified in the glazing process document, in cm. The glaze absorption correction coefficient is derived from the pre-glaze weighing calibration results of blanks of the same material and moisture content.

[0037] If the process documentation directly specifies the target wet adhesion quality, then that quality is used directly. This results in a set of surface partitions.

[0038] S120. Generate candidate reference detection postures based on the surface partition set, and collect the joint static torque and pre-glazing detection reference data. The reference detection posture refers to the predetermined robot joint posture in which the robot maintains the clamping state of the ceramic piece to be glazed and is used to collect the joint static torque. The joint static torque refers to the joint load torque directly output by the joint torque sensor or calculated by combining the servo motor current with the motor torque constant and the reducer transmission ratio when the robot maintains the reference detection posture and the speed of each joint is lower than the static judgment threshold.

[0039] First, candidate baseline detection postures are generated within the robot's reachable space, and postures that cause collisions, joint positions or loads exceeding limits, fixture slippage, insufficient bending radius of the glaze supply hose, and may cause wet glaze slurry to migrate across zones are eliminated.

[0040] Based on the robot's nominal kinematic parameters, the candidate lever arm response matrix is ​​calculated for each candidate baseline detection posture. This matrix is ​​used only for pose selection. Let the number of surface partitions be N, the number of baseline detected poses be K, and the number of effective robot joints in each pose be... The selected posture should satisfy:

[0041]

[0042] Here, rank represents the rank of the matrix. An effective robot joint is a robot joint whose torque response caused by a unit increase in mass is greater than the joint torque detection resolution. The joint torque detection resolution is derived from the joint torque sensor calibration certificate or servo current repeated sampling test. If the conditions are not met, a reference detection posture is added, the clamping direction is adjusted, or adjacent surface partitions with similar responses are merged.

[0043] For example, with 12 surface partitions and 4 effective robot joints per pose, at least 3 poses are required; if the responses of 2 skirt partitions are approximately the same in 3 poses, then a tilting pose that can disrupt the symmetry relationship is added.

[0044] Each time, the robot enters the corresponding baseline detection posture along the same joint movement direction. After the velocity of each joint enters the static judgment range, continuous sampling is performed. Sampling points that deviate from the median by more than the outlier threshold are removed, and the remaining sampling points are averaged. The static judgment threshold, sampling duration, outlier threshold, and torque stability threshold are derived from repeated holding tests of the robot holding unglazed ceramic parts.

[0045] While collecting the static torque of the joints, the actual joint angles, the temperature of each effective robot joint, the glaze supply pressure, the traction force of the glaze supply hose, and the clamping force of the fixture are recorded to form the benchmark data for pre-glazing inspection and the set of static torques of the joints before glazing.

[0046] S130. Establish the final lever arm response matrix and lever arm response relationship based on the surface partition set, the reference detection posture, and the robot kinematic parameters. The lever arm response coefficient refers to the gravitational torque response generated on a specified effective robot joint under a specified reference detection posture when a certain surface partition increases by a unit mass.

[0047] The lever arm response coefficients corresponding to the k-th baseline detection posture, the j-th effective robot joint, and the i-th surface partition are determined according to the following formula:

[0048]

[0049] in, Indicates the lever arm response coefficient; This represents the position vector of the area-weighted center of the i-th surface partition in the robot base coordinate system under the k-th reference detection posture; This represents the position vector of the reference point on the j-th effective robot joint axis; The unit direction vector representing the joint axis of rotation; This represents the gravitational acceleration vector in the robot's base coordinate system; "×" indicates the cross product of vectors, and "·" indicates the dot product of vectors. The position vector is obtained by transforming the area-weighted center position of S110 through the robot's forward kinematics. The joint axis parameters are derived from the robot calibration file, and the gravity direction is derived from the gravity calibration after the robot is installed.

[0050] To eliminate fixture deformation, reducer friction, and actual joint axis deviation, a known mass calibration block was installed near the weighted center of the surface area of ​​a representative unglazed ceramic part. The difference in the joint static torque before and after installation was collected. This difference was divided by the mass of the calibration block to obtain the measured lever arm response coefficient, which was then used to correct the theoretical lever arm response coefficient. After removing the calibration block, the joint static torque was checked to confirm that it returned to the allowable fluctuation range.

[0051] Arrange all corrected lever arm response coefficients in a fixed order according to the baseline detection posture number, effective robot joint number, and surface partition number to form the final lever arm response matrix. The relationship between lever arm response and the final lever arm response matrix is ​​used for mass inversion after S420 and each subsequent compensation.

[0052] The S130 ultimately outputs a set of surface partitions, a set of reference detection postures, reference data for pre-glazing detection, a set of static torques of joints before glazing, a final lever arm response matrix, and lever arm response relationships, which are then used for subsequent trajectory generation, state reproduction, and quality inversion.

[0053] S2 specifically includes the following sub-steps:

[0054] S210. Generate the initial glazing trajectory and initial flow command based on the surface model, surface partition set, and target glazing quality. The initial glazing trajectory consists of trajectory points with unique time sequence numbers. Each trajectory point records the nozzle target position, spray direction, corresponding surface partition, surface normal, target spray distance, target running speed, target mass flow rate, and arrival time.

[0055] The glazing sequence is determined according to the axial position, surface adjacency, and occlusion relationship of the porcelain piece to be glazed, so that the surface areas that are easily occluded by adjacent umbrella skirts are glazed before the exposed surface areas; a continuous circumferential trajectory is generated on the neck and umbrella skirt body, the surface geodesic distance between adjacent trajectories is reduced at the root of the umbrella skirt, an independent edge trajectory is set at the edge of the umbrella skirt, and a trajectory is generated along the continuous direction of curvature on the annular transition surface.

[0056] The target position of the nozzle is obtained by offsetting any point on the trajectory centerline outward along the corresponding surface normal by the target spray distance, with the spray direction opposite to the corresponding surface normal direction. The target spray distance is derived from the calibration results of the glaze quality of standard ceramic tiles under different spray distances; the effective spray width is derived from the spray width calibration under the same nozzle, spray pressure, spray distance, and spray angle; the trajectory overlap rate is derived from the pre-glaze coverage test, and is determined by the condition that no thin glaze gaps or overlapping glaze nodules are formed between adjacent trajectories. The trajectory centerline is discrete according to the robot control cycle and surface curvature, and the spacing between trajectory points is reduced when the surface curvature increases.

[0057] The target glazing quality per unit area of ​​the i-th surface partition is determined by the following formula:

[0058]

[0059] in, This represents the target glaze mass per unit area of ​​the i-th surface partition, in g / cm³. 2 ; and These represent the target glazing quality and surface area determined by S110, respectively.

[0060] The initial mass flow rate of the l-th trajectory point is determined by the following formula:

[0061]

[0062] in, This represents the initial mass flow rate at the l-th trajectory point, in g / s. Indicates the effective spray width, in cm; The target running speed is expressed in cm / s. This indicates the coating transfer efficiency, which is the ratio of the difference in quality of the standard ceramic tile before and after glazing to the reduction in the weight of the glaze supply tank.

[0063] The target mass flow rate is allocated to the repeatedly covered area according to the expected number of coverages. By setting the metering pump speed or flow valve opening in stages and weighing the sprayed glaze within a specified time, a correspondence between the target mass flow rate and the actuator control quantity is established. The initial glazing trajectory and initial flow command are formed from all trajectory points, the target mass flow rate, and the actuator control quantity.

[0064] S220. Generate a set of candidate joint configurations for each trajectory point in the initial glazing trajectory, and calculate the direction of maximum end-effector stiffness, end-effector normal stiffness, and configuration cost. A candidate joint configuration refers to a set of robot joint angles that enable the nozzle to reach the target position and spray direction of the corresponding trajectory point, and satisfy the constraints of joint range, collision, singular configuration, and bending of the glazing hose.

[0065] Based on the robot kinematic calibration file, the nozzle tool center point calibration results, and the 3D models of the robot body, fixture, glazing hose, and ceramic piece to be glazed, all inverse kinematic solutions are solved for each trajectory point. Inverse kinematic solutions that are out of joint position limits, have collisions, nozzle pose errors exceeding allowable values, singular configuration margins below allowable values, insufficient bending radius of the glazing hose, or cannot be continuously transformed with adjacent trajectory points are eliminated.

[0066] The permissible values ​​for nozzle pose error are derived from calibration tests of spray distance and spray angle variations and glaze quality deviations. The permissible values ​​for singular configuration margins are derived from kinematic constraints provided by the robot manufacturer. The maximum end-effector stiffness direction refers to the spatial direction corresponding to the maximum eigenvalue of the robot's end-effector stiffness matrix; the end-effector normal stiffness refers to the external force required for the nozzle to generate a unit displacement along the normal to the corresponding surface when the robot is in the current candidate joint configuration. The robot joint equivalent stiffness is derived from manufacturer data or obtained through calibration tests by applying a known external force to the end-effector and measuring the displacement.

[0067] The end-effector compliance matrix is ​​calculated based on the robot's Jacobian matrix and the equivalent joint stiffness matrix. The end-effector normal stiffness of the l-th trajectory point and the q-th candidate joint configuration is determined according to the following formula:

[0068]

[0069] in, This indicates the end-effector normal stiffness, expressed in N / m. Represents the unit surface normal of the l-th trajectory point; This represents the end-effector translation compliance matrix for the corresponding candidate joint configuration; the superscript T indicates transpose.

[0070] The direction of the jet reaction is determined by the opposite direction of the nozzle's central axis and its magnitude is calibrated using a wrist force sensor. The traction direction and magnitude of the glaze supply hose are determined by the hose fixing point location, the current candidate joint configuration, and the hose tension sensor readings. Normalization is performed on the angle between the maximum end stiffness direction and the surface normal, the end normal stiffness, the singular configuration margin, the hose traction effect, and the proximity of the joint boundary to form a configuration cost value. The smaller the value, the more suitable the corresponding candidate joint configuration is for executing that trajectory point.

[0071] S230. An ant colony algorithm is used to search for the target joint configuration sequence from the candidate joint configuration set of each trajectory point, forming an initial glazing control sequence. Each trajectory point forms a search layer, and the candidate joint configurations of that trajectory point form search nodes. A feasible transition path is established only when two candidate joint configurations of adjacent trajectory points satisfy constraints on joint velocity, joint acceleration, joint jerk, collision, and bending of the glazing hose. Joint jerk refers to the rate of change of joint acceleration relative to time.

[0072] The path cost is determined by combining the configuration cost, joint displacement, end-effector normal stiffness variation, end-effector maximum stiffness direction variation, singular configuration margin variation, and enamel supply hose traction variation, and heuristic information is generated accordingly.

[0073] Specifically, let the total path cost from node i to node j be... Then heuristic information Set as the reciprocal of the path cost, i.e. The probability of an ant choosing to move from its current node i to a feasible subsequent node j. Calculate using the following formula:

[0074]

[0075] in, This represents the current pheromone concentration along the path from node i to node j; The pheromone heuristic factor has a value range of 1.0 to 2.0. is the expected heuristic factor, ranging from 2.0 to 5.0; allowed is the set of feasible subsequent nodes that satisfy all physical constraints such as joint velocity and collision. s represents any optional node in the set of feasible subsequent nodes allowed. and These represent the pheromone concentration and heuristic information on the path from node i to node s, respectively.

[0076] Initialize the pheromones for each feasible transformation path. Each ant selects the next node in sequence based on the pheromones and heuristic information. The sequence is terminated if a candidate sequence violates any physical constraint.

[0077] After each search round, pheromone evaporation is performed, and the pheromone level along the path is increased according to the total cost of the complete candidate joint configuration sequence. The number of ants, pheromone evaporation coefficient, maximum number of iterations, and convergence threshold are derived from offline search experiments on standard ceramic parts of the same model; they are recalibrated when the robot model, gripper position, or hose arrangement changes. The search stops when the maximum number of iterations is reached, or when the change in the optimal total cost is less than the convergence threshold for multiple consecutive rounds. The candidate joint configuration sequence with the lowest total cost and satisfying all constraints is determined as the target joint configuration sequence.

[0078] Finally, the nozzle target position, spray direction, target joint angle, target joint velocity, target joint acceleration, arrival time, initial mass flow rate, and actuator control quantity of each trajectory point are time-sequentially bound to form an initial glazing control sequence for direct execution by S310. Simultaneously, qualified candidate joint configurations that were not selected at each trajectory point are reserved as backup candidate joint configurations for S310 to perform constrained switching when the actual normal trajectory deviates or the end-effector normal stiffness exceeds the limit.

[0079] S3 specifically includes the following sub-steps:

[0080] S310. Control the robot to execute the initial glazing according to the initial glazing control sequence and generate an initial glazing execution record. The initial glazing execution record is a data set that records the robot's actual motion state, actual mass flow rate, actual output glaze quality, and glaze supply status in order of trajectory point number and time sequence.

[0081] Using the robot controller's system clock as a unified time reference, the arrival time of the trajectory point, the target joint configuration, the target mass flow rate, and the actuator control quantity are all set with the same timestamp. The glaze supply response delay is obtained through a standard spraying test, which is the time difference between the issuance of the flow command and the actual mass flow rate of the nozzle reaching the target value; during formal glazing, the flow command is sent ahead of the glaze supply response delay relative to the arrival time of the trajectory point.

[0082] For example, if the glaze supply response delay is 120ms and the trajectory running speed is 0.12m / s, the actual landing point will be delayed by 14.4mm if no delay compensation is performed. Therefore, the flow command should be sent 120ms before the nozzle reaches the corresponding trajectory point.

[0083] The robot moves according to the target joint configuration sequence, reads the actual joint angles according to the control cycle, and calculates the actual position and actual spray direction of the nozzle based on the forward kinematic parameters. Simultaneously, it collects the actual mass flow rate, glaze supply pressure, jet reaction force, and glaze supply hose traction force through mass flow sensors, glaze supply pressure sensors, wrist force sensors, and hose tension sensors. When no mass flow sensor is installed, the actual mass flow rate is calculated by subtracting the return mass from the mass change collected by the glaze tank weighing sensor.

[0084] The normal trajectory deviation of the l-th trajectory point is determined by the following formula:

[0085]

[0086] in, This represents the normal trajectory deviation of the l-th trajectory point, in mm; Represents the unit surface normal of the l-th trajectory point; and These represent the actual position vector of the nozzle and the target position vector of the nozzle, respectively.

[0087] The allowable value for normal trajectory deviation is derived from calibration tests between changes in spray distance and changes in the glazing quality of standard ceramic tiles. The controller also recalculates the maximum stiffness direction and end normal stiffness based on the actual joint angle; when the normal trajectory deviation continuously exceeds the allowable value, or the end normal stiffness is lower than the lower stiffness limit, the glazing of the corresponding surface area is paused, a backup candidate joint configuration is switched, or the operating speed is reduced.

[0088] The lower limit of stiffness is derived from a standard ceramic glazing test under the same jet reaction force and hose traction force. The actual output glaze mass is obtained by integrating the actual mass flow rate over time, and the above data and abnormal surface zoning marks are written into the initial glazing execution record.

[0089] S320: After initial glazing, eliminate residual glaze discharge, establish a glazing supply baseline state, and control the robot to enter each baseline detection posture within the detection time window. Residual glaze discharge refers to the continued output process caused by the remaining pressure in the pipeline and the residual glaze slurry in the nozzle after the glazing device receives the stop command. The glazing supply baseline state refers to the equipment state in which the flow valve state, metering pump state, glazing hose filling state, glazing pressure, and glazing hose traction force all meet the requirement of comparable static torque of the joints before and after glazing.

[0090] After the robot completes the last trajectory point, it moves the nozzle to the glaze slurry recovery position, closes the flow valve and stops the metering pump, or switches the glaze supply circuit to the reflux state; it continuously collects the actual mass flow rate and glaze supply pressure. When the actual mass flow rate is continuously lower than the residual flow rate threshold and the glaze supply pressure enters the reference range corresponding to the pre-glazing detection benchmark data, it is determined that the residual glaze discharge has ended.

[0091] The residual flow threshold and pressure reference range were derived from repeated shut-off tests of the same glazing device. Subsequently, the glazing circuit was adjusted so that the glazing pressure, glazing hose traction force, and clamping force were within the reference range corresponding to the pre-glazing test baseline data, and it was confirmed that the hose fixing point, clamping position, and robot end effector components remained unchanged.

[0092] For example, if the glazing pressure is 0.20 MPa during the pre-glazing test and remains 0.25 MPa after glazing, the static torque of the joint should not be collected immediately. The glazing pressure should first be restored to the reference range corresponding to 0.20 MPa.

[0093] The detection time window refers to the time range within which all benchmark detection postures are completed after the residual glaze has been removed, provided that the migration amount of wet glaze across surface zones does not exceed the allowable migration mass. The detection waiting time, posture tilt range, posture switching speed, holding time, and allowable migration mass are derived from the tilt holding test of the same batch of glaze and the standard ceramic piece firing defect verification test.

[0094] The robot enters each reference detection posture in sequence according to the posture order and joint movement direction determined by S120, and simultaneously records the actual joint angle, joint temperature, glaze supply pressure, glaze supply hose traction force and clamping force.

[0095] S330: Collect, filter, and output the set of effective static joint torques after glazing and the effective initial glazing execution record. The effective static joint torque after glazing refers to the static joint torque obtained when the reference detection posture, glazing reference state, robot thermal state, clamping state, and torque stability all meet the requirements.

[0096] In each baseline detection posture, the robot continuously collects the torque values ​​of each effective robot joint at the same sampling frequency and duration as S120. The difference between the actual joint angle and the corresponding angle before glazing must not exceed the allowable value for posture repetition. The speed of each joint should be lower than the static judgment threshold. The glazing pressure, glazing hose traction force, joint temperature, and clamping force should be within the reference range corresponding to the baseline detection data before glazing. The allowable value for posture repetition is derived from the repeated holding test of the robot in the same baseline detection posture.

[0097] Calculate the torque range for the sampled data of the j-th effective robot joint under the k-th baseline detection posture:

[0098]

[0099] in, This represents the torque range during the sampling period, expressed in N·m. Let represent the joint torque at the t-th sampling time, where max represents finding the maximum value and min represents finding the minimum value.

[0100] When the torque range is not greater than the stability threshold, sampling points that deviate from the median by more than the outlier threshold are removed, and the remaining sampling points are averaged to obtain the static torque of the joint after glazing. If the failure is due to joint instability, the holding time is extended; if the failure is due to unqualified glazing pressure or hose traction, the glazing reference state is re-established; if the failure is due to actual joint angle deviation, the reference detection posture is re-entered in the specified direction; if the failure is due to wet glaze migration, all posture data for this round are discarded.

[0101] After completing the last pose, return to the first baseline detection pose to collect the joint static torque again; if the difference between the two results exceeds the migration check threshold, re-detection is performed after the glaze stabilizes. The maximum number of re-acquisitions is derived from the robot's repeated hold experiment; if valid data cannot be obtained after reaching the maximum number of re-acquisitions, an anomaly flag is output and the process stops entering S410.

[0102] The valid data are arranged in the same order as the static torque set of the joint before glazing to form a valid static torque set of the joint after glazing; at the same time, the timestamp, trajectory point number and the continuity of the actual mass flow rate of the initial glazing execution record are checked to form a valid initial glazing execution record, and the two results are output to S410 and S420.

[0103] S4 specifically includes the following sub-steps:

[0104] S410. Based on the current set of detected joint static torques, the selected set of reference joint static torques, and the corresponding detection reference data, determine the corrected set of newly added torques and the set of newly added torque uncertainties. In the initial glazing inversion and the cumulative new glaze quality inversion, the set of reference joint static torques is the set of joint static torques before glazing obtained in S120; in the incremental inversion of this round of compensation, the set of reference joint static torques is the set of effective joint static torques obtained before the start of this round of compensation.

[0105] The corrected new torque refers to the torque change after subtracting the selected reference joint static torque from the current detected joint static torque, and deducting the background torque caused by differences in joint temperature, glaze supply pressure, and glaze supply hose traction force. All data are arranged in order of the reference detection posture number and the effective robot joint number used in the final lever arm response matrix.

[0106] The corrected torque for the j-th effective robot joint under the k-th baseline detection posture is determined by the following formula:

[0107]

[0108] in, This indicates the newly added torque after the correction, in N·m. and These represent the current static torque of the detected joint and the static torque of the selected reference joint, respectively. , and These represent the joint temperature difference, glaze supply pressure difference, and glaze supply hose traction difference between the two tests, respectively. , and These represent the temperature correction factor, pressure correction factor, and traction correction factor, respectively.

[0109] Each correction factor is derived from a single-factor calibration test of the robot holding an unglazed ceramic part. This involves changing the joint temperature, glaze supply pressure, or glaze supply hose traction force, recording the changes in the joint static torque under the same baseline detection posture, and then fitting the results. If the corresponding difference is within the negligible range determined by repeated calibration, the corresponding correction term is set to 0.

[0110] For example, if the current detection torque increases by 0.08 N·m relative to the reference torque, and the temperature change and hose traction change correspond to background increments of 0.03 N·m and 0.01 N·m respectively, then the corrected new torque is 0.04 N·m.

[0111] The newly added moment uncertainty is jointly determined by the current detection sampling standard deviation, the baseline detection sampling standard deviation, and the maximum fitting residual of the background correction relationship. Observations whose uncertainty exceeds the minimum detectable new mass response to stress moment are marked as low-confidence observations and have their weight reduced in S420. This outputs the corrected set of newly added moments, the set of newly added moment uncertainties, and the data sorting index.

[0112] S420. Based on the revised set of newly added torques, the set of newly added torque uncertainties, the final lever response matrix, the adjacency relationship of surface partitions, and the current effective glazing execution records, invert the quality of the newly added glaze slurry for each surface partition.

[0113] The current effective glazing execution record is the effective initial glazing execution record after the initial glazing, and the effective compensated glazing execution record after the u-th compensation; when performing cumulative inversion, all effective execution records of the initial glazing and all previous compensated glazing are used. The inversion of the newly added glaze slurry quality is performed under the conditions that the newly added glaze slurry quality in each surface zone is non-negative, the total adhered quality is limited, and the quality change of adjacent consecutive glazed zones is constrained, to solve for the zone mass distribution that can explain the newly added torque after correction.

[0114] The solution employs weighted nonnegativity constraints, with the objective function being to find the new glaze slurry quality vector that minimizes the residual. The specific mathematical optimization formula is as follows:

[0115]

[0116] in, This represents the newly added glaze slurry quality vector that needs to be optimized. This means finding the minimum value of the formula under the constraint that the mass of the newly added glaze slurry is non-negative (i.e., each component is greater than or equal to zero); The squared norm of a vector is denoted by 2. This represents the newly added glaze slurry quality vector; This represents the final lever arm response matrix formed by S130; This indicates the newly added torque vector after the correction; This represents the weight matrix, whose diagonal elements are determined according to the reciprocal of the newly added moment uncertainty; This represents the adjacency constraint matrix of the surface partitions, which is derived from the surface partition set of S110; This represents the adjacency constraint coefficient, derived from the standard ceramic part inversion calibration test with known partition quality.

[0117] No adjacency constraints are established between partitions that cross umbrella skirt edges, abrupt boundary changes in surface normal, or discontinuous glazing areas. Before solving, the condition number and minimum singular value of the final lever arm response matrix are calculated; the upper limit of the allowable condition number, the lower limit of the allowable minimum singular value, and the allowable value of the inversion residual are all determined by inversion verification experiments using known quality calibration blocks. If the condition number exceeds the limit or the minimum singular value is insufficient, a reference detection attitude is added, adjacent partitions with similar responses are merged, or an anomaly marker for quality inversion is output.

[0118] The reference adhesion quality is calculated based on the actual mass flow rate and corresponding spraying transfer efficiency in the current effective glazing execution record, and the difference between the sum of the newly added glaze mass in each zone and the reference adhesion quality is constrained to be within the allowable range of adhesion quality. After solving, the following steps are taken: Reconstruct the newly added moment and compare it with the corrected newly added moment; if the inversion residual exceeds the limit, the inversion is deemed invalid.

[0119] For each newly added moment after correction, while keeping the other moments unchanged, the uncertainty of the corresponding newly added moment is increased and decreased respectively, and the inversion is repeated. The maximum absolute deviation of each result from the baseline inversion result is determined as the mass inversion uncertainty. Finally, the newly added glaze slurry mass distribution, mass inversion uncertainty, inversion residual, and mass inversion valid markers are output.

[0120] S430. Based on the newly added glaze slurry quality distribution, quality inversion uncertainty, target glazing quality, allowable glazing deviation, and current effective glazing execution records, determine the set of under-glazing zones, the set of compensating glaze amounts, the set of under-glazing cause markers, and the markers for non-compliance.

[0121] The mass gap of the i-th surface partition is determined by the following formula:

[0122]

[0123] in, This represents the mass gap of the i-th surface partition, in grams. This represents the target glazing quality of the i-th surface partition; This indicates the quality of the newly added glaze slurry obtained through inversion.

[0124] The allowable deviation for glazing is derived from the product glazing process document, or calculated from the allowable deviation for the target wet glaze thickness, surface area, and wet density of the glaze slurry. Only when the quality inversion is marked as valid, and the quality gap is greater than the sum of the allowable deviation for glazing and the uncertainty of the quality inversion, is the corresponding surface area determined as an underglazed area.

[0125] For example, when the target glaze mass is 18g, the inverted new glaze mass is 17.2g, the allowable deviation of glaze application is 0.5g, and the mass inversion uncertainty is 0.4g, the mass shortfall of 0.8g does not exceed 0.9g and is not judged as under-glazed; when the inverted new glaze mass is 16.8g, the mass shortfall is 1.2g and is judged as under-glazed.

[0126] The quality gap is defined as the compensation glaze amount, which represents the actual quality of the wet glaze paste that still needs to be attached to the under-glazed zone.

[0127] Based on the current valid glazing execution records, the cause of insufficient glazing is marked as follows: insufficient flow rate is marked as insufficient flow rate; excessive normal trajectory deviation is marked as trajectory deviation; insufficient stiffness at the end of the normal trajectory is marked as insufficient stiffness; abnormal transfer efficiency is marked as abnormal stiffness when the reference adhesion quality is lower than the predetermined adhesion quality; and other cases are marked as undetermined. Surface zones with excessive inversion residuals, wet glaze migration, fixture slippage, or newly added glaze quality exceeding the target glazing quality are not allowed to be compensated.

[0128] Finally, the set of underglaze zones, the set of compensated glaze amounts, the uncertainty of quality inversion, the set of underglaze cause markers, and the markers indicating that compensation is not allowed are output to S510 and S520. The reference type and data version used in this inversion are also recorded to avoid mixing the initial reference with the current compensation reference in subsequent compensation cycles.

[0129] S5 specifically includes the following sub-steps:

[0130] S510. Based on the set of underglaze zones, the set of compensating glaze amounts, the uncertainty of quality inversion, the set of underglaze cause markers, the markers for disallowed compensation, and the surface model, generate a set of candidate compensating glaze application trajectories and a set of candidate compensating flow instructions.

[0131] First, surface zones marked with "compensation not allowed" are removed. For surface zones with invalid quality inversion, excessive inversion residuals, fixture slippage, wet glaze migration, over-glazing of the current zone, or those unable to form a compensation spray pattern without affecting adjacent normal zones, a manual verification mark is output, and they are not included in automatic compensation. The compensation area is then determined for the remaining under-glazed zones. The compensation area refers to the surface area within the under-glazed zone where automatic compensation glazing is permitted, after deducting the safety boundary distance.

[0132] The safety boundary distance is derived from the spray edge attenuation test under the same spray distance, spray pressure, and spray angle. Its value should ensure that the mass of glaze falling into the adjacent non-underglazed zone does not exceed the corresponding glaze allowable deviation. Larger compensation areas generate one or more equidistant tracks, narrow areas generate a central track along the long axis, and the root of the umbrella skirt and the annular transition surface generate tracks along the continuous curvature direction.

[0133] Each candidate compensation trajectory point records the underglaze zone number, location, surface normal, compensation spray distance, compensation running speed, expected coverage area, and risk of repeated coverage. When the minimum width of the underglaze area is less than the minimum effective spray width of the nozzle, replace with a narrow-width nozzle, adjust the spray angle, or set a "compensation not allowed" mark. For example, if the circumferential width of the underglaze area is 12mm, but the minimum effective spray width of the nozzle is 30mm, direct spraying along the center of the area is prohibited.

[0134] Compensation spray transfer efficiency refers to the proportion of the mass of glaze actually adhering to the compensation area during the compensation glazing process. Its initial value comes from the weighing test of standard ceramic pieces with different wet glaze thicknesses and waiting times after secondary spraying.

[0135] The compensation output quality of the i-th underglaze zone is determined by the following formula:

[0136]

[0137] in, This indicates the compensated output mass, expressed in grams (g). This indicates the amount of compensating glaze determined by S430, in grams. This indicates compensation for coating transfer efficiency.

[0138] The compensated output quality is distributed to each trajectory point according to the area of ​​the compensated application zone and the trajectory coverage ratio. Then, candidate compensated quality flow rates are determined based on the effective spray width, compensated operating speed, and the flow rate calibration relationship of the glazing device. The candidate compensated quality flow rate levels are classified according to the minimum stable flow rate, maximum allowable flow rate, and flow control resolution of the glazing device. Each level corresponds to a specific metering pump speed or flow valve opening. This forms a set of candidate compensated glazing trajectories and a set of candidate compensated flow rate commands.

[0139] S520. Based on the candidate set of compensating glazing trajectories, the candidate set of compensating flow commands, the set of markers for underglazing causes, and the uncertainty of quality inversion, the ant colony algorithm is used to determine the compensating glazing trajectory, the compensating flow command, and the sequence of compensating target joint configurations.

[0140] For each candidate compensation trajectory point, a set of candidate compensation joint configurations is generated according to the inverse kinematics solution, collision check, singular configuration check, and enamel supply hose bending check rules of S220. A multi-layer search structure is established based on the temporal sequence of the compensation trajectory points, and each search node consists of one candidate compensation trajectory point, one candidate compensation joint configuration, and one candidate compensation mass flow rate level.

[0141] A feasible conversion path is established only when adjacent nodes simultaneously meet the following conditions: continuous nozzle trajectory, joint velocity and acceleration not exceeding limits, joint jump meeting requirements, mass flow rate change not exceeding the response capability of the glazing device, end normal stiffness not lower than the lower limit of stiffness, the angle between the end maximum stiffness direction and the surface normal is within the allowable range, the glazing response delay can be compensated, and the expected adhesion quality of adjacent non-underglazing zones does not exceed the limit.

[0142] The evaluation weights are adjusted according to the cause of the underglaze: for insufficient flow type, the weight of flow matching is increased; for trajectory deviation type and insufficient stiffness type, the weights of normal trajectory deviation, end normal stiffness and end maximum stiffness direction are increased; for abnormal transfer efficiency type, the weights of compensation spray distance, spray angle and compensation spray transfer efficiency matching are increased.

[0143] The total cost of the candidate compensation control path is determined by the following formula:

[0144]

[0145] in, This represents the total cost of the candidate compensation control path; This represents the trajectory cost, which is formed by the trajectory length, the number of turning points, and the number of partition switching. This represents the flow cost incurred to compensate for mass flow rate deviation and flow rate change rate. This represents the end-effector normal stiffness, the direction of maximum end-effector stiffness, joint displacement, singular configuration margin, and the configuration cost resulting from hose traction variations. This indicates the risk of repeated glazing in adjacent non-underglazed zones; This indicates the overcompensation risk corresponding to the uncertainty in quality inversion; , , , and The corresponding weight is derived from the standard ceramic component compensation test.

[0146] To achieve a balance among the various optimization objectives, in an embodiment for a typical combined porcelain insulator, the specific normalized values ​​of each weight can be set as follows: Trajectory Cost Weight =0.20, Traffic Cost Weight =0.35, configuration cost weight =0.15, repeating the risk weight of glaze. =0.15, overcompensation risk weight =0.15.

[0147] Initialize feasible path pheromones, and each ant selects nodes layer by layer based on pheromones and heuristic information. After each round of search, pheromone evaporation is performed, and the pheromone is updated according to the total cost of the complete path. The search stops when the maximum number of iterations is reached, or when the change in the optimal total cost is less than the convergence threshold for multiple consecutive rounds. The trajectory points with the lowest total cost that satisfy all constraints, the compensation target joint configuration, the compensation mass flow rate, and the arrival time are time-series bound together to form a compensation glazing control sequence.

[0148] S530. Execute the compensation glazing control sequence, and update the quality of newly added glaze slurry and the cumulative newly added glaze slurry quality in this round of compensation through the static torque detection of the joints after compensation. The compensation cycle refers to one complete process from the generation of the compensation glazing control sequence, through compensation glazing, torque detection after compensation and inversion of the cumulative newly added glaze slurry quality, to the update of the under-glazed zone set and the compensation glaze amount set.

[0149] The robot performs compensatory glazing according to the unified clock and glaze supply response delay compensation method of S310. It simultaneously collects the actual joint angle, actual nozzle position, actual compensated mass flow rate, actual compensated output mass, normal trajectory deviation, end normal stiffness, end maximum stiffness direction, glaze supply pressure and glaze supply hose traction force to form a compensatory glazing execution record.

[0150] After compensation is completed, residual glaze is eliminated, a glaze supply baseline state is established, and the effective compensation set of joint static torques is obtained according to S320 and S330. The mass of newly added glaze slurry in this round of compensation is based on the effective set of joint static torques before the start of this round of compensation, and the cumulative mass of newly added glaze slurry is always based on the set of joint static torques before glazing in S120. Then, S410, S420, and S430 are executed in sequence to obtain the updated cumulative mass of newly added glaze slurry, mass inversion uncertainty, set of under-glazed zones, and set of compensated glaze amount.

[0151] Based on the ratio of the newly added glaze slurry quality in this round of compensation to the actual output quality in this round of compensation, the compensation spraying transfer efficiency of the corresponding under-glazed zone is updated, and the updated value is used in the next compensation cycle.

[0152] In the u-th compensation cycle, the amount of glaze compensation for the i-th under-glazed zone is determined according to the following formula:

[0153]

[0154] in, This indicates the amount of glaze to be compensated, in grams. This represents the cumulative increase in glaze mass of the i-th surface partition after the u-th compensation. This represents the compensation ratio coefficient, which is no greater than 1. It is determined based on the proportion of the quality inversion uncertainty to the current quality gap, the fluctuation of the compensation spraying transfer efficiency, and the risk of adjacent overglazing, according to the segmented correspondence table formed by multiple rounds of compensation tests on standard ceramic parts.

[0155] Glazing ends when the quality defects of all surface zones do not exceed the sum of the allowable glazing deviation and the uncertainty of quality inversion. Automatic compensation stops and a manual verification mark is output when the maximum number of compensation cycles is reached, the commanded compensation glaze amount is lower than the minimum executable compensation glaze amount, the quality defects fail to decrease according to the minimum reduction ratio in two consecutive rounds, any zone becomes over-glazed, or the quality inversion is continuously invalid. The maximum number of compensation cycles, the minimum executable compensation glaze amount, and the minimum reduction ratio are derived from the minimum stable flow rate test of the glazing device and the multi-round compensation test of the standard ceramic parts.

[0156] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0157] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0158] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0159] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatic glazing trajectory and flow control of a combined electric porcelain insulator glazing robot, characterized in that, Includes the following steps: S1. Obtain the surface model of the ceramic piece to be glazed, divide it into multiple surface partitions, determine the target glazing quality of each surface partition, control the robot to maintain multiple reference detection postures before glazing, collect the static torque of the joints before glazing, and establish the lever arm response relationship. S2. Generate the initial glazing trajectory and initial flow command based on the surface model, generate candidate joint configurations for the trajectory points, and determine the target joint configuration sequence using the ant colony algorithm based on the direction of maximum end stiffness, end normal stiffness and glazing hose traction. S3. Control the robot to execute the initial glazing trajectory according to the target joint configuration sequence, and adjust the glaze flow rate according to the initial flow command. After the glazing is completed, establish the glazing reference state, and collect the static torque of the joint after glazing according to the reference detection posture. S4. Based on the static torque of the joint before glazing, the static torque of the joint after glazing, and the lever arm response relationship, the quality of the newly added glaze in each surface zone is inverted and compared with the target glazing quality to determine the under-glazed zones and the amount of glaze compensation.

2. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 1, characterized in that, Also includes: S5. Generate compensation glazing trajectory and compensation flow instructions based on the underglaze zone and compensation glaze amount. Use ant colony algorithm to determine the compensation target joint configuration sequence, execute compensation glazing, and update the cumulative new glaze mass based on the compensated joint static torque until the target glazing quality is reached or the compensation termination condition is met.

3. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 1, characterized in that, S1 specifically includes: Obtain the surface model of the porcelain piece to be glazed, establish the porcelain coordinate system, divide the surface model into geometric partitions and circumferential subdivisions, determine the area, center position, average surface normal and target glazing quality of each surface partition, and form a set of surface partitions. Candidate reference detection postures are generated based on the surface partition set. Reference detection postures that meet the identifiable conditions are selected based on the candidate lever arm response matrix. The robot is then controlled to enter the reference detection posture, and the static torque of the joints and the detection reference data before glazing are collected.

4. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 3, characterized in that, Also includes: The lever arm response coefficient is calculated based on the surface partition set, the reference detection posture, and the robot's kinematic parameters. The known mass calibration block is used for correction to form the final lever arm response matrix and lever arm response relationship.

5. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 1, characterized in that, S2 specifically includes: The glazing sequence is determined based on the surface model, surface partition set, and target glazing quality, and an initial glazing trajectory and initial flow command are generated, including nozzle position, spray direction, spray distance, speed, and mass flow rate. For the trajectory points, candidate joint configurations are solved, and configurations that do not meet the constraints of joint range, collision, singular configuration and glaze supply hose bending are eliminated. The configuration cost is determined based on the direction of maximum end stiffness, end normal stiffness, jet reaction and glaze supply hose traction. A search structure is established using trajectory points and candidate joint configurations. An ant colony algorithm is used to search for a sequence of target joint configurations that satisfy continuous motion constraints. This sequence is then bound to the initial glazing trajectory, initial flow command, and timing information to form an initial glazing control sequence.

6. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 1, characterized in that, S3 specifically includes: The robot is controlled to perform initial glazing according to the initial glazing control sequence, compensate for glazing response delay, and collect actual joint angles, actual nozzle positions, actual mass flow rates, glazing pressure, and glazing hose traction force to form an initial glazing execution record. After the initial glazing is completed, residual glaze is eliminated, and the glazing pressure, glazing hose traction force, and clamping force are restored to the reference range of the pre-glazing test baseline data. The robot is then controlled to enter each baseline test posture within the test time window.

7. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 6, characterized in that, Also includes: Static torques of the joints after glazing are collected under various reference detection postures. The repeatability of the posture, the reference state of glazing, and the stability of the torque are screened to form a set of effective static torques of the joints after glazing and effective initial glazing execution records.

8. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 1, characterized in that, S4 specifically includes: Based on the static torque of the joint before glazing, the static torque of the joint after glazing, and the test benchmark data, the background torque caused by the difference in joint temperature, glazing pressure, and glazing hose traction force is deducted to obtain the set of new torques after correction and the set of new torque uncertainties. Based on the revised set of new torques, the final lever response matrix, the surface partition adjacency relationship and the effective initial glazing execution record, a weighted inversion is performed under the conditions of non-negative new glaze slurry quality and limited reference adhesion quality to obtain the new glaze slurry quality distribution, quality inversion uncertainty, inversion residual and effective quality inversion marker. The newly added glaze slurry quality distribution is compared with the corresponding target glazing quality. Combined with the allowable deviation of glazing and the uncertainty of quality inversion, the underglaze zone, the amount of glaze compensation, the marker of the cause of underglaze and the marker of no compensation are determined.

9. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 2, characterized in that, S5 specifically includes: Based on the underglaze zone, compensation glaze amount, quality inversion uncertainty and the mark that compensation is not allowed, the compensation area and compensation spraying transfer efficiency are determined, and a set of candidate compensation glazing trajectories and a set of candidate compensation flow instructions are generated. Candidate compensation joint configurations are generated for candidate compensation glazing trajectory points. Search nodes are constructed using trajectory points, candidate compensation joint configurations, and candidate compensation mass flow rates. Ant colony algorithm is used to determine the compensation glazing control sequence.

10. The automatic glazing trajectory and flow control method for the combined porcelain insulator glazing robot according to claim 9, characterized in that, It also includes: executing the compensation glazing control sequence, collecting the static torque of the joint after compensation, updating the quality of the newly added glaze slurry in this round of compensation, the cumulative quality of the newly added glaze slurry, and the compensation spraying transfer efficiency, and ending the compensation when the target glazing quality is reached or the compensation termination condition is met.

Images