CN114750146A - Robot milling track precision compensation method - Google Patents
Robot milling track precision compensation method Download PDFInfo
- Publication number
- CN114750146A CN114750146A CN202210231140.4A CN202210231140A CN114750146A CN 114750146 A CN114750146 A CN 114750146A CN 202210231140 A CN202210231140 A CN 202210231140A CN 114750146 A CN114750146 A CN 114750146A
- Authority
- CN
- China
- Prior art keywords
- milling
- robot
- data
- theoretical
- track
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000003801 milling Methods 0.000 title claims abstract description 143
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 16
- 244000062645 predators Species 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000012360 testing method Methods 0.000 claims description 14
- 238000002474 experimental method Methods 0.000 claims description 7
- 238000005520 cutting process Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 4
- 238000003754 machining Methods 0.000 description 11
- 238000013528 artificial neural network Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 101001001429 Homo sapiens Inositol monophosphatase 1 Proteins 0.000 description 1
- 102100035679 Inositol monophosphatase 1 Human genes 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000012636 effector Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011999 immunoperoxidase monolayer assay Methods 0.000 description 1
- 238000011900 installation process Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- GHZKGHQGPXBWSN-UHFFFAOYSA-N methyl(propan-2-yloxy)phosphinic acid Chemical compound CC(C)OP(C)(O)=O GHZKGHQGPXBWSN-UHFFFAOYSA-N 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1669—Programme controls characterised by programming, planning systems for manipulators characterised by special application, e.g. multi-arm co-operation, assembly, grasping
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Manipulator (AREA)
Abstract
The invention provides a robot milling track precision compensation method, which comprises the following steps: training a milling force prediction model based on a deep feedforward neural network model; determining a milling range of a robot, planning a preset number of sampling points in the range and collecting theoretical pose data of the sampling points; acquiring actual milling force and actual position coordinate data of the robot at a sampling point, and comparing the actual position coordinate data with the theoretical position coordinate data to obtain actual positioning error data; training a milling positioning error prediction model based on a deep feedforward neural network model; and correcting the coordinates of the target point based on the milling force prediction model and the milling positioning error prediction model. The method comprehensively considers the influence of external force and theoretical pose on the track precision of the robot, and meanwhile, in the process of training the model, the number of nodes of the hidden layer is determined by adopting an ocean predator algorithm and a grid search method, so that the track precision of the robot in milling processing is obviously improved.
Description
Technical Field
The invention belongs to the technical field of robot machining, and particularly relates to a robot milling track precision compensation method.
Background
In recent years, milling of large complex parts by using industrial robots has been increasingly and widely applied to the aerospace high-end manufacturing field. The industrial robot has good motion flexibility, strong man-machine interaction capacity, low manufacturing cost and small space requirement, can quickly adjust the working state, and realizes intelligent and flexible production of products.
In the machining process, the robot end effector completes interpolation motion of a machining track, and therefore the track precision of the robot directly influences the machining precision of a product. However, generally, since the industrial robot inevitably introduces errors in the manufacturing and installation processes, the actual kinematic model is different from the nominal kinematic model, and the trajectory precision of the robot is usually low. In addition, in the process of machining by the robot, besides the motion error of the robot body, the deformation error caused by the machining load is also one of the main factors influencing the machining track precision and the machining surface quality of the robot. In engineering applications, particularly in heavy machining operations, errors due to the effects of the working load tend to be in the order of millimeters. Therefore, the research on the track precision compensation of the industrial robot under the processing condition has practical and profound significance.
The patent of publication No. CN104535027A performs parameter identification on the actual kinematic error of the robot, and solves the actual pose of the robot through inverse kinematic solution to complete the positioning accuracy compensation of the robot, but this method ignores non-geometric factors, and is complex in modeling and limited in compensation effect. The patent publication CN102230783A adopts an inverse distance weighting method to predict and compensate the positioning error of the robot, but in this method, the sampling step size has a great influence on the compensation accuracy and efficiency, and the influence of the attitude on the error is not considered. Patent publication No. CN110385720B, Chinese Journal of Aeroniatics, 2021, 35 (2): 346-360 articles establish a robot positioning error compensation model by using a deep feedforward neural network, and the method ignores the influence of an external force on the accuracy of the robot end.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a robot milling track precision compensation method for improving the milling track precision of a robot.
The technical scheme provided by the invention is as follows:
on the one hand, the invention discloses a robot milling track precision compensation method, which comprises the following steps:
Making a first sample data set, selecting milling parameters, designing a preset number of milling experiments, collecting milling forces applied to the tail end of the robot, wherein the milling parameters and the corresponding milling forces form a group of sample data, preprocessing the first sample data set, dividing the preprocessed sample data set into a first training set and a first test set according to a preset proportion, training a milling force prediction model by using the first training set based on a deep feedforward neural network model,
determining a milling processing range of a robot, planning a preset number of sampling points in the range and collecting theoretical pose data of the sampling points, wherein the theoretical pose data comprise theoretical position coordinate data and theoretical attitude angle data;
controlling the robot to traverse the sampling point, acquiring actual milling force and actual position coordinate data of the robot at the sampling point, and comparing the actual position coordinate data with the theoretical position coordinate data to obtain actual positioning error data;
making a second sample data set, wherein the actual milling force, theoretical pose data and actual positioning error data of each sampling point are a group of sample data, preprocessing the second sample data set, dividing the second sample data set into a second training set and a second test set according to a preset proportion, and training a milling positioning error prediction model based on a deep feed-forward neural network model by using the second training set;
Obtaining milling parameters according to the milling task, inputting the milling parameters into a milling force prediction model to obtain a milling force prediction value;
acquiring a track to be processed according to a milling task, dispersing the track to be processed into a target point set consisting of the start and end positions of each line segment, and acquiring theoretical pose data of the target point;
and inputting the milling force predicted value and the theoretical pose data of the target point into a milling positioning error prediction model to obtain a milling positioning error predicted value, reversely compensating the positioning error predicted value into the theoretical pose data of the target point to obtain the coordinates of the corrected target point, sequentially sending the coordinates of the corrected target point to a control system of the robot, controlling the robot to move to the corrected target point, and executing a milling task.
Further, the milling parameters comprise spindle rotation speed, feed speed and cutting depth.
Further, milling experiments were performed using a single factor method.
Further, the milling force is measured using a six-dimensional force sensor.
Further, the milling force is gaussian filtered and averaged.
Further, the preprocessing adopts a normalization method.
Further, the deep feedforward neural network model determines the number of hidden layer nodes by adopting an ocean predator algorithm and a grid search method.
Further, the determining the milling range specifically includes:
and measuring and establishing a robot base coordinate system and a milling tool coordinate system by using a laser tracker, and determining a pose interval of the milling tool coordinate system under the robot base coordinate system according to the milling task.
Further, the error compensation is calculated according to the following formula:
Pmodified＝Epredict+Ptarget＝[x+Δx,y+Δy,z+Δz,a,b,c]
wherein E ispredict＝[Δx,Δy,Δz]Indicating the predicted value of the positioning error of the target point, Ptarget＝[x,y,z,a,b,c]And representing the theoretical pose of the target point.
In another aspect, the present invention discloses a computer-readable storage medium for storing computer-executable instructions, which, when executed, implement any one of the possible implementation methods for compensating the precision of the milling trajectory of the robot described in the first aspect.
The method for compensating the milling track precision of the robot comprehensively considers the influence of the external force and the theoretical pose on the track precision of the robot when sample data sets of a training milling force prediction model and a milling positioning error prediction model are manufactured, and meanwhile, in the process of training the models, the number of nodes of a hidden layer is determined by adopting an ocean predator algorithm and a grid search method, so that the prediction precision and efficiency of the models are further improved, the universality of the models is good, the applicability is wide, and the track precision of the robot in milling processing is remarkably improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings which are required to be used in the technical solution description will be briefly introduced below, it is obvious that the exemplary embodiments of the present invention and the description thereof are only used for explaining the present invention and do not constitute an unnecessary limitation of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without inventive labor. In the drawings:
FIG. 1 is a schematic flow chart of example 1 of the present invention;
FIG. 2 is a schematic diagram of a training process for optimizing a deep neural network by using an improved marine predator algorithm in embodiment 1 of the invention;
FIG. 3 is a schematic view of a milling test platform of an industrial robot used in embodiment 1 of the present invention;
FIG. 4 is a schematic view of a load device in embodiment 1 of the present invention;
FIG. 5 is a schematic diagram of the improved ocean predator algorithm optimized deep neural network structure in embodiment 1 of the invention;
FIG. 6 is a schematic diagram showing a comparison result of comprehensive positioning errors before and after precision compensation of milling tracks of different models in embodiment 1 of the present invention;
FIG. 7 is a diagram showing a comparison result of comprehensive errors before and after precision compensation of a linear milling path of a robot in embodiment 1 of the present invention;
Description of the reference numerals:
1-an industrial robot; 2-a laser tracker target ball; 3-milling the cutter; 4-milling the platform; 5-laser tracker.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
Example 1
Referring to fig. 1, the present embodiment provides a method for compensating precision of a milling trajectory of a robot, including the following steps:
this embodiment is based on the industrial robot that fig. 3 shows mills test platform implementation, industrial robot's mill test platform is including installing industrial robot 1, laser tracker target ball 2 on ground, milling cutter 3 that industrial robot 1's arm end set up, mills platform 4 and laser tracker 5.
S1: carrying out a milling experiment by adopting a single-factor method and measuring milling force in the milling process;
s1.1: selecting three milling parameters of spindle rotation speed, feed speed and cutting depth as variables to carry out single factor experiment;
Specifically, the milling direction is set to the x-axis of the robot-based coordinate system. Firstly, respectively fixing the rotation speed and the feed speed of a main shaft of the robot at 5000r/min and 200mm/min, generating 70 groups of cutting depth data within the range of 0.5 mm-3 mm, respectively fixing the rotation speed and the cutting depth of the main shaft at 5000r/min and 2.5mm, generating 70 groups of feed speed data within the range of 80 mm/min-300 mm/min, respectively fixing the feed speed and the cutting depth at 200mm/min and 2.5mm, and generating 80 groups of main shaft rotation speed data within the range of 3000 r/min-8000 r/min. Inputting the 220 groups of data into a robot operating system for milling test, and measuring the milling force under corresponding parameters in the milling process by using a six-dimensional force sensor.
S1.2: and performing Gaussian filtering on the milling force measured by the six-dimensional force sensor and then performing average processing.
S2: manufacturing a sample data set of a milling force prediction model of the training robot;
specifically, the milling parameters and the measured milling force are normalized by the maximum and minimum values, and the normalization formula is as follows:
wherein x isiIs a parameter value of a certain dimension, xmaxAnd xminRespectively representing the maximum and minimum values of this parameter,
is a normalized value;
Further, the milling parameters and the corresponding milling forces are a set of sample data, and the sample data set is divided into a training set and a test set according to a ratio of 19: 1.
S3: training a robot milling force prediction model by adopting a training set;
specifically, a deep feedforward neural network model is adopted as a robot milling force prediction model, and the input layer of the model is a robot milling parameter [ n, f, a ]p]The output layer of the model is the milling force (F) borne by the milling cutter at the tail end of the robotx,Fy,Fz]；
S31: randomly obtaining an initial weight and a threshold of the deep feedforward neural network model;
s32: optimizing a deep feed-forward neural network model by using an ocean predator algorithm (IMPA) and a grid search method to obtain the number of hidden layer nodes which enable the prediction effect of the network model to be optimal;
specifically, firstly carrying out rough optimal parameter search on required parameters by using an ocean predator algorithm to eliminate most low-precision intervals, then taking the parameters searched by the ocean predator algorithm as a search center, carrying out fine optimization on small step length among cells for the second time by using a grid search method, carrying out k-fold cross validation, and selecting the group of parameters with highest accuracy as optimal parameters.
S33: training the network model by using a training set of sample data;
Specifically, the output expression of the feed-forward neural network of the l-th layer is as follows:
al＝fl(Wl·al-1+bl)
wherein, WlIs a weight matrix between layer l-1 and layer l, blIs the threshold vector for layer l; the f (-) function is the ReLU activation function, which has the formula:
further, the loss function of the deep feedforward neural network model is Root Mean Square Error (RMSE), which is formulated as:
wherein, yiIn order to expect the output for the model,
n is the number of training samples of the model;
the loss function is used for reflecting the degree of inconsistency between the actual output value and the expected output value of the model, and the smaller the loss function is, the higher the model precision is. Deep feedforward neural networks are usually trained by an error Back Propagation (BP) algorithm, and the parameters W of each layer of neurons are continuously adjusted through multiple forward and backward calculationslAnd blTo minimize the loss function. And after each parameter is updated, recalculating the loss function value, determining whether to continuously update the parameters, and finishing the model training when the model precision reaches the preset requirement.
S34: and (3) taking the milling parameters of the test set as the input of the trained robot milling force prediction model to obtain a corresponding milling force prediction value, selecting a network model structure with the highest prediction precision and storing model data thereof to serve as the robot milling force prediction model.
S4: determining a milling range of the robot, planning a preset number of sampling points in the range and collecting theoretical pose data of the sampling points;
specifically, determining the milling space range of the robot refers to determining a pose interval of a milling tool coordinate system under a base coordinate system according to a processing task after measuring by a laser tracker and establishing the base coordinate system and the milling tool coordinate system of the robot;
in the embodiment, a cuboid area of 600mm × 1200mm × 800mm is used as a test space, and theoretical pose data of 4000 groups of sampling points are randomly generated, wherein the theoretical pose data comprise theoretical position coordinate data and theoretical attitude angle data;
s5: controlling the robot to traverse the sampling point, collecting actual milling force and actual position coordinate data of the robot at the sampling point, and comparing the actual position coordinate data with the theoretical position coordinate data to obtain actual positioning error data;
specifically, a world coordinate system, a robot base coordinate system, a flange plate coordinate system and a tool coordinate system are sequentially established according to a laser tracker manual, and a laser tracker and a working coordinate system of a robot are combined together;
inputting the theoretical pose data of the sampling points into a robot control system, applying loads with different sizes and directions to the tail end of the robot by using a load device shown in fig. 4, controlling the robot to sequentially move to the positions of the sampling points, measuring actual position coordinates by using a laser tracker, comparing the actual position coordinates with the theoretical position coordinates to obtain actual positioning errors, and measuring the stress of the robot on each sampling point by using a six-dimensional force sensor.
S6: making a sample data set of a positioning error prediction model of the training robot;
specifically, stress data, theoretical pose data and corresponding actual positioning errors of sampling points are used as a group of sample data, and after all the sample data are preprocessed, the sample data are divided into a training set and a test set according to a preset proportion;
the pretreatment and the scale division are the same as S2.
S7: training a robot positioning error prediction model by adopting a training set;
specifically, the stress magnitude and the theoretical pose [ F ] of the sampling point are determinedx,Fy,Fz,x,y,z,a,b,c]As input to the model, the actual positioning error vector [ Δ x, Δ y, Δ z]As the output of the model, secondly, training the model by using training set data, and storing the model structure which enables the prediction result to be optimal;
the specific steps of training are the same as S3.1-S3.4;
the robot milling force prediction model and the robot positioning error prediction model obtained in the step S3 jointly form an improved ocean predator algorithm optimized deep neural network (IMPA-DNN), and the structure of the improved ocean predator algorithm optimized deep neural network refers to fig. 5.
S8: determining a robot milling parameter according to a milling task, and inputting the parameter into a robot milling force prediction model to obtain a milling force prediction value;
specifically, milling parameters are determined according to the requirements of the milling task, the milling parameters are input into a robot milling force prediction model trained in S3, and the output result is subjected to inverse normalization processing to obtain predicted values of milling forces in X, Y and Z directions, which are subjected to the milling of the robot tail end under the parameters. If the task requires that the milling direction is not along the X direction of the robot base coordinate system, but a certain included angle exists, the predicted F is xAnd FyOrthogonal decomposition is performed along the X and Y axes of the sensor coordinates.
S9: dispersing the track to be processed into a point set consisting of the initial and final positions of each line segment and obtaining theoretical pose data of the points;
specifically, the linear track of the robot can be equally divided into a plurality of interpolation line segments according to a given step length, so that the linear track can be approximately processed into a point set consisting of the start and end positions of each line segment and executed through a linear motion mode;
and determining a machining track according to the milling task, and dispersing the machining track into a set of a plurality of points. And inputting the theoretical poses of the points and the milling force predicted value obtained in the step S8 into a robot positioning error prediction model trained in the step S7, and outputting a positioning error predicted value.
S10: carrying out robot milling track precision compensation;
specifically, the predicted positioning error value is subjected to inverse normalization processing to obtain a predicted positioning error value E of the target pointpredict＝[Δx,Δy,Δz]Will beidEerpCompensating reversely to target theoretical pose Ptarget＝[x,y,z,a,b,c]To obtain the compensated coordinates P of the target pointmodified＝Epredict+Ptarget＝[x+Δx,y+Δy,z+Δz,a,b,c]. The compensated coordinate PmodifiedInputting a control system of the robot to control the robot to move to a target point position; and measuring the actual coordinate of the track, and comparing the actual coordinate with a theoretical value to obtain a compensated track error.
Further, in order to embody the accuracy and superiority of model prediction in the present invention, in this embodiment, for example, the positioning error of the predicted sampling point is taken as an example, the positioning error prediction results of the deep neural network model (DNN), the particle swarm optimization deep neural network model (PSO-DNN), the genetic particle swarm optimization deep neural network (GPSO-DNN) and the improved marine predator optimization deep neural network (IMPA-DNN) in the present invention are compared, 3800 sets of samples are used to train the above five models respectively, and 200 sets of test data are substituted into the trained models, and the prediction accuracies of the above models are shown in fig. 6. As can be seen from Table 1, the uncompensated positioning error is within 0.704mm, the prediction error of the PSO-DNN model is within 0.169mm, the prediction error of the GPSO-DNN model is within 0.141mm, the prediction error of the DNN model is within 0.218mm, and the prediction error of the IMPA-DNN model in the invention is within 0.058 mm. It can be seen that the maximum value, the minimum value, the average value and the standard deviation of the prediction error of the IMPA-DNN model are all smaller than those of PSO-DNN and GPSO-DNN. The corresponding errors of the DNN and MPA-DNN models indicate that the prediction accuracy of the inventive IMPA-DNN model is the highest.
TABLE 1 comparison of error data (mm) of compensation results of different models
Further, in order to verify that the robot trajectory error compensation method based on the improved ocean predator algorithm optimized deep neural network model (IMPA-DNN) is effective and feasible, in the embodiment of the invention, a linear trajectory is directly milled on the robot milling test platform shown in FIG. 2, the trajectory error of the linear trajectory is measured by using the laser tracker, then the trajectory is subjected to discrete processing, error prediction and compensation according to S8-S10, the robot is controlled to perform a milling experiment according to the corrected coordinates, and the compensated trajectory error is measured by using the laser tracker. The comparison result of the comprehensive errors of the milling trajectories of the robots before and after compensation is shown in fig. 7, and it can be seen that the trajectory error compensation method can compensate the trajectory error of the robot from 0.893mm to 0.217mm, the trajectory error is reduced by 75.6%, and the feasibility of the robot milling trajectory error compensation method based on the IMPA-DNN model is verified.
In conclusion, the invention predicts the milling force and the tail end error of the robot by using the deep neural network model and carries out the track compensation without parameter identification and establishing a kinematic model, thereby obviously improving the milling track precision of the robot, and having simple method, low cost and good universality.
In addition, an embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium may store a program, and when the program is executed, the program includes some or all of the steps of any one of the robot milling trajectory precision compensation methods described in the above method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.
The integrated unit, if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned memory comprises: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, and various media capable of storing program codes.
Those skilled in the art will appreciate that all or part of the steps of the methods of the above embodiments may be implemented by a program, which is stored in a computer-readable memory, the memory including: flash memory disks, read-only memory, random access memory, magnetic or optical disks, and the like.
The above-mentioned embodiments, objects, technical solutions and advantages of the present application are further described in detail, it should be understood that the above-mentioned embodiments are only examples of the present application, and are not intended to limit the scope of the present application, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present application should be included in the scope of the present application.
Claims (10)
1. A robot milling track precision compensation method is characterized by comprising the following steps:
making a first sample data set, selecting milling parameters, designing a preset number of milling experiments, collecting milling forces applied to the tail end of the robot, wherein the milling parameters and the corresponding milling forces form a group of sample data, preprocessing the first sample data set, dividing the preprocessed sample data set into a first training set and a first test set according to a preset proportion, training a milling force prediction model by using the first training set based on a deep feedforward neural network model,
Determining a milling processing range of a robot, planning a preset number of sampling points in the range and collecting theoretical pose data of the sampling points, wherein the theoretical pose data comprise theoretical position coordinate data and theoretical attitude angle data;
controlling the robot to traverse the sampling point, acquiring actual milling force and actual position coordinate data of the robot at the sampling point, and comparing the actual position coordinate data with the theoretical position coordinate data to obtain actual positioning error data;
making a second sample data set, wherein the actual milling force, theoretical pose data and actual positioning error data of each sampling point are a group of sample data, preprocessing the second sample data set, dividing the second sample data set into a second training set and a second test set according to a preset proportion, and training a milling positioning error prediction model based on a deep feed-forward neural network model by using the second training set;
obtaining milling parameters according to the milling task, inputting the milling parameters into a milling force prediction model to obtain a milling force prediction value;
acquiring a track to be processed according to the milling task, dispersing the track to be processed into a target point set consisting of the start and end positions of each line segment, and acquiring theoretical pose data of the target point;
And inputting the milling force predicted value and the theoretical pose data of the target point into a milling positioning error prediction model to obtain a milling positioning error predicted value, reversely compensating the positioning error predicted value into the theoretical pose data of the target point, obtaining the coordinates of the corrected target point, sequentially sending the coordinates of the corrected target point to a control system of the robot, controlling the robot to move to the corrected target point, and executing a milling task.
2. The method for compensating the precision of the milling track of the robot according to claim 1, wherein the milling parameters comprise spindle rotation speed, feed speed and cutting depth.
3. The robot milling trajectory precision compensation method according to claim 1, characterized in that a single factor method is adopted for the milling experiment.
4. The method for compensating the precision of the milling track of the robot according to claim 1, wherein the milling force is measured by using a six-dimensional force sensor.
5. The method for compensating the precision of the milling track of the robot as claimed in claim 1, wherein the milling force is gaussian filtered and averaged.
6. The robot milling trajectory precision compensation method according to claim 1, wherein the preprocessing adopts a normalization method.
7. The precision compensation method for the milling track of the robot as claimed in claim 1, wherein the depth feed-forward neural network model determines the number of nodes of the hidden layer by using a marine predator algorithm and a grid search method.
8. The method for compensating for the precision of the milling trajectory of the robot according to claim 1, wherein the determining the milling range specifically comprises:
and measuring and establishing a robot base coordinate system and a milling tool coordinate system by using a laser tracker, and determining a pose interval of the milling tool coordinate system under the robot base coordinate system according to the milling task.
9. The precision compensation method for the milling track of the robot as claimed in claim 1, wherein the error compensation is calculated according to the following formula:
Pmodified＝Epredict+Ptarget＝[x+Δx,y+Δy,z+Δz,a,b,c]
wherein E ispredict＝[Δx,Δy,Δz]Representing the predicted value of the positioning error of the target point, Ptarget＝[x,y,z,a,b,c]Representing the theoretical pose of the target point.
10. A computer-readable storage medium having stored thereon computer-executable instructions that, when executed, implement the robot milling trajectory precision compensation method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210231140.4A CN114750146B (en) | 2022-03-10 | 2022-03-10 | Robot milling track precision compensation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210231140.4A CN114750146B (en) | 2022-03-10 | 2022-03-10 | Robot milling track precision compensation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114750146A true CN114750146A (en) | 2022-07-15 |
| CN114750146B CN114750146B (en) | 2023-10-24 |
Family
ID=82324762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210231140.4A Active CN114750146B (en) | 2022-03-10 | 2022-03-10 | Robot milling track precision compensation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114750146B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115237054A (en) * | 2022-09-21 | 2022-10-25 | 中科航迈数控软件(深圳)有限公司 | Spindle driving motor control method, device, terminal and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109746928A (en) * | 2019-01-30 | 2019-05-14 | 南京航空航天大学 | A kind of the intelligent flexible production line and its operation method of removable dual robot milling automatic |
| CN110000787A (en) * | 2019-04-12 | 2019-07-12 | 清华大学深圳研究生院 | A kind of control method of super redundant mechanical arm |
| CN110017790A (en) * | 2019-03-15 | 2019-07-16 | 南京航空航天大学 | A kind of generation of curved surface scanning track and optimization method based on measurement accuracy |
| CN110385720A (en) * | 2019-07-26 | 2019-10-29 | 南京航空航天大学 | A kind of robot localization error compensating method based on deep neural network |
| CN112733296A (en) * | 2021-01-12 | 2021-04-30 | 天津大学 | GRNN-based milling error prediction and compensation method for hybrid robot |
-
2022
- 2022-03-10 CN CN202210231140.4A patent/CN114750146B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109746928A (en) * | 2019-01-30 | 2019-05-14 | 南京航空航天大学 | A kind of the intelligent flexible production line and its operation method of removable dual robot milling automatic |
| CN110017790A (en) * | 2019-03-15 | 2019-07-16 | 南京航空航天大学 | A kind of generation of curved surface scanning track and optimization method based on measurement accuracy |
| CN110000787A (en) * | 2019-04-12 | 2019-07-12 | 清华大学深圳研究生院 | A kind of control method of super redundant mechanical arm |
| CN110385720A (en) * | 2019-07-26 | 2019-10-29 | 南京航空航天大学 | A kind of robot localization error compensating method based on deep neural network |
| CN112733296A (en) * | 2021-01-12 | 2021-04-30 | 天津大学 | GRNN-based milling error prediction and compensation method for hybrid robot |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115237054A (en) * | 2022-09-21 | 2022-10-25 | 中科航迈数控软件(深圳)有限公司 | Spindle driving motor control method, device, terminal and storage medium |
| CN115237054B (en) * | 2022-09-21 | 2022-12-16 | 中科航迈数控软件(深圳)有限公司 | Main shaft driving motor control method, device, terminal and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114750146B (en) | 2023-10-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110385720B (en) | Robot positioning error compensation method based on deep neural network | |
| CN111590594B (en) | Robot trajectory tracking control method based on visual guidance | |
| CN111702762B (en) | Industrial robot operation attitude optimization method | |
| Chen et al. | Robotic grinding of a blisk with two degrees of freedom contact force control | |
| Song et al. | Precision robotic deburring based on force control for arbitrarily shaped workpiece using CAD model matching | |
| CN109782601B (en) | Design method of self-adaptive neural network synchronous robust controller of coordinated mechanical arm | |
| CN111958611B (en) | Milling robot posture optimization method and device considering minimum contour error | |
| CN112828359B (en) | Robot milling attitude planning method and system based on multiple constraints of potential field method | |
| Lakshminarayanan et al. | An adaptive framework for robotic polishing based on impedance control | |
| CN102183205A (en) | Method for matching optimal assembly poses of large-sized parts | |
| CN110134062B (en) | Multi-axis numerical control machine tool machining path optimization method based on reinforcement learning | |
| CN115113582B (en) | Part contour error-oriented five-axis position loop gain off-line correction method | |
| US20090214312A1 (en) | Method for optimized milling close to the final contour | |
| CN112428274B (en) | Space motion planning method of multi-degree-of-freedom robot | |
| CN114131611B (en) | Off-line compensation method, system and terminal for joint errors of robot gravity pose decomposition | |
| CN113985807B (en) | Cladding head anti-collision obstacle avoidance control system and method for laser cladding | |
| CN114750146B (en) | Robot milling track precision compensation method | |
| CN111452033A (en) | Double NURBS curve milling trajectory planning method for industrial robot | |
| CN108608425B (en) | Off-line programming method and system for milling of six-axis industrial robot | |
| Zhang et al. | Online optimization method of controller parameters for robot constant force grinding based on deep reinforcement learning rainbow | |
| CN114660940A (en) | Grinding constant force control method based on deep reinforcement learning PPO algorithm | |
| CN113110290B (en) | Large-scale thin-wall part mirror image machining method based on force feedback control | |
| CN110977478B (en) | Mobile dual-robot machining system and method for drilling and milling weak rigid support | |
| CN112327938A (en) | Robot near-zero following error control method based on data driving | |
| CN116690557A (en) | Method and device for controlling humanoid three-dimensional scanning motion based on point cloud |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |