CN112733296A - GRNN-based milling error prediction and compensation method for hybrid robot - Google Patents

GRNN-based milling error prediction and compensation method for hybrid robot Download PDF

Info

Publication number
CN112733296A
CN112733296A CN202110037335.0A CN202110037335A CN112733296A CN 112733296 A CN112733296 A CN 112733296A CN 202110037335 A CN202110037335 A CN 202110037335A CN 112733296 A CN112733296 A CN 112733296A
Authority
CN
China
Prior art keywords
error
workpiece
point
value
processing
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.)
Pending
Application number
CN202110037335.0A
Other languages
Chinese (zh)
Inventor
秦旭达
朱文富
李皓
李士鹏
赵庆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN202110037335.0A priority Critical patent/CN112733296A/en
Publication of CN112733296A publication Critical patent/CN112733296A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J11/00—Manipulators not otherwise provided for
    • B25J11/005—Manipulators for mechanical processing tasks
    • B25J11/0055—Cutting
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J19/00—Accessories fitted to manipulators, e.g. for monitoring, for viewing; Safety devices combined with or specially adapted for use in connection with manipulators
    • B25J19/0095—Means or methods for testing manipulators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Mechanical Engineering (AREA)
  • Robotics (AREA)
  • Computer Hardware Design (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Numerical Control (AREA)

Abstract

The invention discloses a method for predicting and compensating milling errors of a series-parallel robot, which comprises the following steps: acquiring milling error data of the hybrid robot, and defining the deviation of a theoretical coordinate value of the shape of the milling track and an actual measured value as a processing error; and establishing an error prediction model for milling the workpiece by the hybrid robot based on the Matlab and the generalized regression neural network, obtaining an optimal error prediction model through multiple training, and performing error prediction on a theoretical tool path of milling the workpiece generated by UG software by using the optimal neural network model to realize error compensation on the position coordinate of each processing point of the workpiece. By adopting the method, the pre-compensation of the machining error is realized, and the compensation precision is improved.

Description

GRNN-based milling error prediction and compensation method for hybrid robot
Technical Field
The invention belongs to the technical field of milling of hybrid robots, and particularly relates to a method for predicting and compensating milling errors of a hybrid robot.
Background
The parallel-serial robot is a new mechanical processing configuration device developed on the basis of a parallel robot and a serial robot. Most of the existing parallel-serial robots are realized by serially connecting serial turners on the basis of parallel mechanisms. Compared with the traditional serial configuration equipment and the traditional parallel configuration equipment, the parallel-serial robot has the advantages of simple structure, good rigidity, high positioning precision, good dynamic response, relatively large working space and small occupied area, and is suitable for large-scale machining processes. The machining precision is one of the important indexes in the machining process, various errors are difficult to avoid in the machining process, the requirement on the machining precision is guaranteed, and the problem that the hybrid robot is required to be solved urgently when being machined and landed is solved.
Compared with a numerical control machine tool, the hybrid robot is small in rigidity, and the rigidity changes along with the pose, so that the machining precision is difficult to guarantee. The factors influencing the machining errors of the robot are many, and mainly include the manufacturing and mounting errors of the robot, the position and attitude errors of the robot, the positioning errors, the tool errors, the clamp errors and the like. The error compensation comprises a hardware compensation method and software compensation, wherein the hardware compensation method has high cost, high operation difficulty and low cost, is not suitable for large-scale processing, and is simple and convenient to operate and has the highest comprehensive benefit.
At present, the method for predicting and compensating the machining error of the hybrid robot mainly focuses on geometric error modeling of the robot and self calibration of the robot, the geometric error modeling needs a complicated mathematical derivation process, time and labor are wasted, an actual cutting process is not involved, and the machining precision is difficult to guarantee.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a GRNN-based milling error prediction and compensation method for a hybrid robot, which realizes pre-compensation of a processing error and improves the compensation precision.
The invention discloses a method for predicting and compensating milling errors of a hybrid robot, which comprises the following steps of:
acquiring milling error data of a hybrid robot, and defining the deviation of a theoretical coordinate value of the shape of a milling track and an actual measured value as a processing error; the specific process is as follows:
firstly, establishing a workpiece CAD model in UG software, and planning each outline shape for milling the workpiece;
secondly, theoretical measuring points on each contour shape for processing the workpiece are planned by using a point set function in UG software;
thirdly, adding rough milling and finish milling procedures, and setting processing parameters required for processing the workpiece in UG software;
fourthly, generating a tool position track for processing the workpiece by using a UG processing function module, generating an NC code by using a post processor and inputting the NC code into the hybrid robot;
fifthly, milling the workpiece by the hybrid robot according to the NC code of the tool position track generated in the fourth step, recording the tool setting value of the workpiece machining position of the numerical control machining system of the hybrid robot, detaching the milled workpiece, placing the workpiece on a three-coordinate measuring instrument for coordinate measurement, measuring the coordinate of an actual point corresponding to a theoretical measurement point, calculating to obtain a difference value between the actual point and the theoretical measurement point as error data between the actual point and the theoretical measurement point, and storing the difference value in a txt document;
step two, establishing an error prediction model for milling a workpiece by a hybrid robot by adopting Matlab based on a generalized regression neural network, wherein GRNN input data is a three-dimensional coordinate value of a theoretical measurement point of the surface profile of the workpiece under a machine tool coordinate system, and an output value is an error value which is obtained in the step one and corresponds to the three-dimensional coordinate value of the theoretical measurement point; obtaining an optimal error prediction model through multiple times of training, wherein the smooth factor value of the generalized regression neural network is obtained by adopting a 10-fold cross validation method;
and thirdly, predicting errors of a theoretical tool location track generated by milling the workpiece through UG software by using the optimal neural network model, and offsetting the theoretical tool location track by an error value along the processing surface of the workpiece by using an off-line tool location track compensation method and an iteration method for multiple iterations to realize error compensation of the position coordinates of each processing point of the workpiece.
By adopting the method, the pre-compensation of the machining error is realized, and the compensation precision is improved.
Drawings
FIG. 1 is a flow chart of a hybrid robot milling error prediction and compensation method of the present invention;
FIG. 2 is a diagram of a processing trajectory shape of a processing error acquisition experiment according to an embodiment of the present invention;
FIG. 3 is a drawing of theoretical measurement points of machining error according to an embodiment of the present invention;
FIG. 4 is a tool position trajectory diagram of an experimental tool position for processing error acquisition according to an embodiment of the present invention;
FIG. 5 is a diagram of a hybrid robot used in a processing error acquisition experiment according to an embodiment of the present invention;
FIG. 6 is a Generalized Regression Neural Network (GRNN) architecture diagram as employed by the present invention;
FIG. 7 illustrates a model mean square error for different smoothing factors according to an embodiment of the present invention;
FIG. 8 is a partial test data fit of a prediction model training sample created using an optimal smoothing factor according to an embodiment of the present invention;
FIG. 9 is a graph of tool position trajectories before and after compensation according to an embodiment of the present invention;
FIG. 10 is a comparison of different depth circle diameters before and after compensation according to an embodiment of the present invention;
FIG. 11 is a comparison graph of linear machining accuracy before and after compensation according to an embodiment of the present invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention, the technical principles of the present invention will be further described below with reference to the accompanying drawings in the embodiments of the present invention.
As shown in fig. 1, fig. 2, fig. 3, fig. 4, fig. 5, and fig. 6, the present invention provides a method for predicting and compensating a milling error of a hybrid robot, including the following steps:
acquiring milling error data of a hybrid robot, and defining the deviation of a theoretical coordinate value of the shape of a milling track and an actual measured value as a processing error; the specific process is as follows:
firstly, establishing a workpiece CAD model in UG software, and planning each outline shape of the workpiece subjected to milling, wherein the outline shape can comprise an open rectangular groove, a circular groove and a square groove as shown in FIG. 2;
secondly, as shown in fig. 3, theoretical measurement points on each contour shape of the workpiece are processed by utilizing a point set function plan in UG software;
thirdly, adding rough milling and finish milling procedures, and setting processing parameters required for processing the workpiece in UG software, such as the radius of a cutter, the rotating speed of a main shaft, the feeding speed and the like;
fourthly, as shown in fig. 4, a UG processing function module is used for generating a tool position track for processing a workpiece, an NC code is generated through a post processor and is input into the parallel-serial robot;
and fifthly, as shown in fig. 5, milling the workpiece by the hybrid robot according to the NC code of the tool position trajectory generated in the fourth step, recording the tool setting value of the workpiece position processed by the numerical control processing system of the hybrid robot, detaching the milled workpiece, placing the workpiece on a Coordinate Measuring Machine (CMM) for coordinate measurement, measuring the coordinates of an actual point corresponding to a theoretical measurement point, calculating to obtain a difference value between the actual point and the theoretical measurement point as error data between the actual point and the theoretical measurement point, and storing the difference value in a txt document.
Step two, as shown in fig. 6, an error prediction model for milling the workpiece by the hybrid robot is established by using Matlab based on a Generalized Regression Neural Network (GRNN), wherein GRNN input data is a three-dimensional coordinate value of a theoretical measurement point of the workpiece surface profile in a machine tool coordinate system, and an output value is an error value corresponding to the three-dimensional coordinate value of the theoretical measurement point obtained in step one. And obtaining an optimal error prediction model through multiple training, wherein the smooth factor value of the generalized regression neural network is obtained by adopting a 10-fold cross validation method (10-fold cross-validation).
As an embodiment of the present invention, the modeling of the error prediction model for milling the workpiece by the hybrid robot includes the following steps:
firstly, normalizing data samples; converting a three-dimensional coordinate value of a theoretical measuring point in a workpiece coordinate system into a three-dimensional coordinate value of a theoretical measuring point in a machine tool coordinate system by adopting Matlab, normalizing the three-dimensional coordinate value of the theoretical measuring point and an error value corresponding to the three-dimensional coordinate value by utilizing a min-max function carried by Matlab to obtain normalized sample data, wherein the value range of the normalized data is [ -1,1], and the influence on training precision due to different orders of magnitude is avoided;
secondly, determining a training sample and a test sample; and carrying out random processing on the normalized sample data according to the ratio of 4: 1, dividing the sample into two parts, wherein one part is used as a training sample, the rest is used as a testing sample, the three-dimensional coordinate value of the normalized theoretical measuring point is used as an input variable, and an error value corresponding to the input variable is used as an output variable;
and thirdly, determining the optimal smooth factor by adopting 10-fold cross validation. Dividing training sample data into 10 parts and carrying out random processing, wherein 9 parts are used for training a neural network, 1 part is used for verifying the neural network, 10-fold cross verification is carried out, a smooth factor which enables the predicted output value and the actual error value Mean Square Error (MSE) to be minimum corresponds is recorded as an optimal smooth factor, and FIG. 7 shows model mean square errors corresponding to different smooth factors;
fourthly, training a GRNN neural network by using the optimal input value and the output value corresponding to the optimal smoothing factor, and taking the obtained model as an error prediction model, wherein a test data fitting graph of a prediction model training sample part built by using the optimal smoothing factor is shown in FIG. 8;
fifthly, testing an error prediction model;
step 101, testing the performance of the error prediction model by adopting the test sample in the second step, and calculating a decision coefficient R2Root mean square error RRMSEAverage relative error MMREAnd mean absoluteError MMAEEvaluating the performance of the error prediction model;
Figure BDA0002894810920000041
Figure BDA0002894810920000042
Figure BDA0002894810920000043
Figure BDA0002894810920000051
in the formula, xiRepresenting error values which are obtained in the first step and respectively correspond to the three-dimensional coordinate values of the theoretical measurement points;
Figure BDA0002894810920000052
representing error prediction values which are output by the GRNN error prediction model and correspond to three-dimensional coordinate values of all theoretical measurement points;
Figure BDA0002894810920000053
representing a value obtained by normalizing the error value corresponding to the three-dimensional coordinate value of each theoretical measurement point.
Step 102, repeating the third step to the fourth step and the operation of step 101-102 for a plurality of times, usually 10 times;
103, selecting a model with the largest decision coefficient and the smallest root mean square error, average relative error and average absolute error as an optimal error prediction model;
the principle of the step is as follows:
the method is established on the basis of nonparametric regression, performs Parzen nonparametric estimation by taking sample data as a posterior condition, calculates network output according to a maximum probability principle, and has the advantages of high nonlinear approximation performance by taking a radial basis network as a basis, more convenient training, strong nonlinear mapping capability, flexible network structure, high fault tolerance and robustness compared with a radial basis neural network; GRNN is somewhat similar in structure to RBF networks, having a total of four layers: divided into an input layer, a mode layer, a summation layer and an output layer. GRNN input data is three-dimensional theoretical coordinate values of the surface profile of the machined test piece, and output is a corresponding error value. The quality of a prediction result of the neural network model is directly related to the size of the sample data volume, the data volume is too small, the training effect is poor, and the prediction precision is too low; too much data, the training time will increase, and the computational burden of the computer will be increased, even the overfitting situation will occur. Therefore, selecting an appropriate number of training samples is a prerequisite for establishing a good neural network.
In the generalized neural network, the smooth factor determines the good performance of the model. When a larger value is taken, the prediction data is approximately close to the average value of the dependent variable of the sample, and the trend of the fitted curve is relatively gentle; when a smaller value is taken, the predicted data value is close to the value of the training sample, the fitting curve even has an over-learning condition, and the prediction effect is very poor. To obtain the optimal value of the slip factor, 10 fold cross-validation (10-fold cross-validation) was used. The 10-fold cross validation (10-fold cross validation) is that a data set is divided into ten parts, 9 parts of the data set are trained and 1 part of the data set is validated in turn, and the average value of 10 times of results is used as the estimation of the accuracy of the algorithm.
Thirdly, error prediction is carried out on a theoretical tool location track generated by milling the workpiece through UG software by adopting the optimal neural network model, then an off-line tool location track compensation method and an iteration method are adopted, the theoretical tool location track is biased by an error value along the processing surface of the workpiece, and multiple iterations are carried out, so that error compensation of position coordinates of each processing point of the workpiece is realized, and the method specifically comprises the following steps:
firstly, reading a theoretical tool path file (cls format) generated by UG) Extracting the position coordinates of the processing point in the workpiece coordinate system, converting the position coordinates of the processing point into the coordinates of the tool location point in the machine tool coordinate system, and obtaining the coordinates of the ith tool location point after conversion as (x)i,yi,zi) Wherein i belongs to (1, n), n is the number of cutter location points, and the tolerance epsilon is set according to the actual processing requirement;
secondly, starting an initialization cycle, reading a first tool location point coordinate for processing the workpiece when i is 1;
thirdly, predicting errors E corresponding to x, y and z coordinates of the current tool location point by adopting an optimal error prediction modelx,Ey,EzCalculating the normal error delta of the tool location pointiWherein, in the step (A),
Figure BDA0002894810920000061
the fourth step, judge whether deltaiIf yes, compensating the current tool location point, adding the predicted error of the current tool location point to the coordinate of the current tool location point, returning to the second step, and repeating the second step to the fourth step to calculate the coordinate of the next tool location point until the compensation of all the tool location point coordinates is finished;
if not, subtracting the predicted error of the current tool location point from the coordinate of the current tool location point, executing the third step to obtain a new normal error, and then executing the fourth step until delta is metiIf the value is less than epsilon, stopping the operation;
fifthly, arranging all the compensated tool location point coordinates obtained, converting the compensated tool location point coordinates into processing point position coordinates in a workpiece coordinate system, creating an empty tool location track file, and writing the processing point position coordinates into the tool location track file, wherein a tool location track before and after compensation is shown in a figure 9;
sixthly, processing the tool path file by using a UG postprocessor to generate a processing NC code;
and seventhly, machining the workpiece by the hybrid robot by using the compensated NC code.
The principle of the step is as follows: predicting an actual machining track by using theoretical tool location data (CL), further modifying numerical control machining (NC) codes of a robot, and performing error compensation before machining; the off-line tool path compensation method is an effective method for reducing machining errors. The error compensation is realized by not changing the processing parameters, but an error source is introduced by modifying the tool position track to offset the processing error. Before machining, the error of the tool position track generated by CAM software is predicted by a trained neural network model, and the theoretical tool position track is offset by an error amount along the machined surface to realize error compensation.
And verifying the machining error after compensation through an experiment, performing a milling experiment by using the compensated NC code, and measuring and comparing the compensated machining error with the machining error which is not compensated. As shown in FIG. 10, the diameters of the circles with different depths after compensation are obviously better than those before compensation, and as shown in FIG. 11, the precision of the processing straight line after compensation is greatly improved compared with that before compensation, thus proving the compensation effectiveness of the method of the invention.
The above description is only exemplary of the present invention and should not be taken as limiting, any modifications, equivalents, improvements and the like which are within the basic principles of the present invention are intended to be included within the scope of the present invention.

Claims (3)

1. A method for predicting and compensating milling errors of a hybrid robot is characterized by comprising the following steps:
acquiring milling error data of a hybrid robot, and defining the deviation of a theoretical coordinate value of the shape of a milling track and an actual measured value as a processing error; the specific process is as follows:
firstly, establishing a workpiece CAD model in UG software, and planning each outline shape for milling the workpiece;
secondly, theoretical measuring points on each contour shape for processing the workpiece are planned by using a point set function in UG software;
thirdly, adding rough milling and finish milling procedures, and setting processing parameters required for processing the workpiece in UG software;
fourthly, generating a tool position track for processing the workpiece by using a UG processing function module, generating an NC code by using a post processor and inputting the NC code into the hybrid robot;
fifthly, milling the workpiece by the hybrid robot according to the NC code of the tool position track generated in the fourth step, recording the tool setting value of the workpiece machining position of the numerical control machining system of the hybrid robot, detaching the milled workpiece, placing the workpiece on a three-coordinate measuring instrument for coordinate measurement, measuring the coordinate of an actual point corresponding to a theoretical measurement point, calculating to obtain a difference value between the actual point and the theoretical measurement point as error data between the actual point and the theoretical measurement point, and storing the difference value in a txt document;
step two, establishing an error prediction model for milling a workpiece by a hybrid robot by adopting Matlab based on a generalized regression neural network, wherein GRNN input data is a three-dimensional coordinate value of a theoretical measurement point of the surface profile of the workpiece under a machine tool coordinate system, and an output value is an error value which is obtained in the step one and corresponds to the three-dimensional coordinate value of the theoretical measurement point; obtaining an optimal error prediction model through multiple times of training, wherein the smooth factor value of the generalized regression neural network is obtained by adopting a 10-fold cross validation method;
and thirdly, predicting errors of a theoretical tool location track generated by milling the workpiece through UG software by using the optimal neural network model, and offsetting the theoretical tool location track by an error value along the processing surface of the workpiece by using an off-line tool location track compensation method and an iteration method for multiple iterations to realize error compensation of the position coordinates of each processing point of the workpiece.
2. The hybrid robot milling machining error prediction and compensation method of claim 1, wherein: the modeling of the error prediction model for milling the workpiece by the hybrid robot comprises the following steps:
firstly, normalizing data samples; converting a three-dimensional coordinate value of a theoretical measuring point in a workpiece coordinate system into a three-dimensional coordinate value of a theoretical measuring point in a machine tool coordinate system by adopting Matlab, normalizing the three-dimensional coordinate value of the theoretical measuring point and an error value corresponding to the three-dimensional coordinate value by utilizing a min-max function carried by Matlab to obtain normalized sample data, wherein the value range of the normalized data is [ -1,1], and the influence on training precision due to different orders of magnitude is avoided;
secondly, determining a training sample and a test sample; and carrying out random processing on the normalized sample data according to the ratio of 4: 1, dividing the sample into two parts, wherein one part is used as a training sample, the rest is used as a testing sample, the three-dimensional coordinate value of the normalized theoretical measuring point is used as an input variable, and an error value corresponding to the input variable is used as an output variable;
thirdly, determining an optimal smooth factor by adopting 10-fold cross validation: dividing training sample data into 10 parts and carrying out random processing, wherein 9 parts are used for training a neural network, 1 part is used for verifying the neural network, 10-fold cross verification is carried out for 10 times, and a smooth factor corresponding to the minimum mean square error between a predicted output value and an actual error value is recorded as an optimal smooth factor;
fourthly, training a GRNN neural network by using the optimal input value and the output value corresponding to the optimal smooth factor, and taking the obtained model as an error prediction model;
fifthly, testing an error prediction model;
step 101, testing the performance of the error prediction model by adopting the test sample in the second step, and calculating a decision coefficient R2Root mean square error RRMSEAverage relative error MMREAnd the mean absolute error MMAEEvaluating the performance of the error prediction model;
Figure FDA0002894810910000021
Figure FDA0002894810910000022
Figure FDA0002894810910000023
Figure FDA0002894810910000024
in the formula, xiRepresenting error values which are obtained in the first step and respectively correspond to the three-dimensional coordinate values of the theoretical measurement points;
Figure FDA0002894810910000025
representing error prediction values which are output by the GRNN error prediction model and correspond to three-dimensional coordinate values of all theoretical measurement points;
Figure FDA0002894810910000026
representing a value obtained by normalizing the error value corresponding to the three-dimensional coordinate value of each theoretical measurement point;
step 102, repeating the third step to the fourth step and the operation of step 101-102 for a plurality of times, usually 10 times;
and 103, selecting a model with the largest decision coefficient and the smallest root mean square error, average relative error and average absolute error as an optimal error prediction model.
3. The hybrid robot milling error prediction and compensation method according to claim 1 or 2, wherein: the third step comprises the following steps:
firstly, reading a theoretical tool location track file generated by UG, extracting a position coordinate of a processing point under a workpiece coordinate system, converting the position coordinate of the processing point into a tool location point coordinate under a machine tool coordinate system, and obtaining the ith tool location point coordinate (x) after conversioni,yi,zi) Wherein i belongs to (1, n), n is the number of cutter location points, and the tolerance epsilon is set according to the actual processing requirement;
secondly, starting an initialization cycle, reading a first tool location point coordinate for processing the workpiece when i is 1;
thirdly, predicting errors E corresponding to x, y and z coordinates of the current tool location point by adopting an optimal error prediction modelx,Ey,EzCalculating the normal error delta of the tool location pointiWherein, in the step (A),
Figure FDA0002894810910000031
the fourth step, judge whether deltaiIf yes, compensating the current tool location point, adding the predicted error of the current tool location point to the coordinate of the current tool location point, returning to the second step, and repeating the second step to the fourth step to calculate the coordinate of the next tool location point until the compensation of all the tool location point coordinates is finished;
if not, subtracting the predicted error of the current tool location point from the coordinate of the current tool location point, executing the third step to obtain a new normal error, and then executing the fourth step until delta is metiIf the value is less than epsilon, stopping the operation;
fifthly, arranging all the compensated tool location point coordinates obtained, converting the compensated tool location point coordinates into processing point position coordinates in a workpiece coordinate system, creating an empty tool location track file, and writing the processing point position coordinates into the tool location track file;
sixthly, processing the tool path file by using a UG postprocessor to generate a processing NC code;
and seventhly, machining the workpiece by the hybrid robot by using the compensated NC code.
CN202110037335.0A 2021-01-12 2021-01-12 GRNN-based milling error prediction and compensation method for hybrid robot Pending CN112733296A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110037335.0A CN112733296A (en) 2021-01-12 2021-01-12 GRNN-based milling error prediction and compensation method for hybrid robot

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110037335.0A CN112733296A (en) 2021-01-12 2021-01-12 GRNN-based milling error prediction and compensation method for hybrid robot

Publications (1)

Publication Number Publication Date
CN112733296A true CN112733296A (en) 2021-04-30

Family

ID=75590528

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110037335.0A Pending CN112733296A (en) 2021-01-12 2021-01-12 GRNN-based milling error prediction and compensation method for hybrid robot

Country Status (1)

Country Link
CN (1) CN112733296A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113102882A (en) * 2021-06-16 2021-07-13 杭州景业智能科技股份有限公司 Geometric error compensation model training method and geometric error compensation method
CN113712696A (en) * 2021-08-27 2021-11-30 佛山市安齿生物科技有限公司 Finish machining method for planting fixed bridge connector
CN114074390A (en) * 2021-11-23 2022-02-22 苏州博宇科技有限公司 Processing system and method for plastic mold electrode automation
CN114237157A (en) * 2021-12-20 2022-03-25 华中科技大学 Machine learning modeling method and system of data-driven machine tool feeding servo system
CN114472989A (en) * 2022-02-10 2022-05-13 中国工程物理研究院机械制造工艺研究所 Geometric error compensation method for surface microstructure machining, machining method and machining device
CN114750146A (en) * 2022-03-10 2022-07-15 南京航空航天大学 Robot milling track precision compensation method
CN116643020A (en) * 2023-05-26 2023-08-25 江苏师范大学 Method for improving surface quality of milling workpiece based on high-density water-based foam cutting fluid

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104759942A (en) * 2015-04-22 2015-07-08 华中科技大学 Online milling deformation measurement and complementation machining method for thin-walled part
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
US20200166906A1 (en) * 2018-11-28 2020-05-28 Industrial Technology Research Institute Simulation method for milling by use of dynamic position error

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104759942A (en) * 2015-04-22 2015-07-08 华中科技大学 Online milling deformation measurement and complementation machining method for thin-walled part
US20200166906A1 (en) * 2018-11-28 2020-05-28 Industrial Technology Research Institute Simulation method for milling by use of dynamic position error
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

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
冯慧君 等: "基于GRNN的冻融期土壤蒸发预测", 《水电能源科学》 *
李士鹏 等: "基于螺旋铣孔柔性切削力建模的孔径误差补偿", 《天津大学学报(自然科学与工程技术版)》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113102882A (en) * 2021-06-16 2021-07-13 杭州景业智能科技股份有限公司 Geometric error compensation model training method and geometric error compensation method
CN113712696A (en) * 2021-08-27 2021-11-30 佛山市安齿生物科技有限公司 Finish machining method for planting fixed bridge connector
CN114074390A (en) * 2021-11-23 2022-02-22 苏州博宇科技有限公司 Processing system and method for plastic mold electrode automation
CN114074390B (en) * 2021-11-23 2024-04-26 苏州博宇科技有限公司 Machining system and method for automation of plastic mold electrode
CN114237157A (en) * 2021-12-20 2022-03-25 华中科技大学 Machine learning modeling method and system of data-driven machine tool feeding servo system
CN114237157B (en) * 2021-12-20 2022-12-16 华中科技大学 Machine learning modeling method and system of data-driven machine tool feeding servo system
CN114472989A (en) * 2022-02-10 2022-05-13 中国工程物理研究院机械制造工艺研究所 Geometric error compensation method for surface microstructure machining, machining method and machining device
CN114750146A (en) * 2022-03-10 2022-07-15 南京航空航天大学 Robot milling track precision compensation method
CN114750146B (en) * 2022-03-10 2023-10-24 南京航空航天大学 Robot milling track precision compensation method
CN116643020A (en) * 2023-05-26 2023-08-25 江苏师范大学 Method for improving surface quality of milling workpiece based on high-density water-based foam cutting fluid

Similar Documents

Publication Publication Date Title
CN112733296A (en) GRNN-based milling error prediction and compensation method for hybrid robot
CN110385720B (en) Robot positioning error compensation method based on deep neural network
Ratchev et al. An advanced FEA based force induced error compensation strategy in milling
US11656614B2 (en) In-process digital twinning
CN108803487B (en) Point position contour error prediction method for side milling surface of part
TWI414376B (en) A five axis flank milling system for machining curved surface and the toolpath planning method thereof
CN111590581A (en) Positioning compensation method and device for robot
CN110069041B (en) Workpiece machining method and system based on-machine measurement
CN102785129A (en) On-line detection method of curved surface machining precision of complex part
CN110286650A (en) A kind of blank based on numerical control macroprogram is in machine fast aligning method
CN112926152B (en) Digital twin-driven thin-wall part clamping force precise control and optimization method
CN113910001B (en) Numerical control machine tool space error identification method
CN110686655A (en) Positioning method, device and system for three-dimensional curved surface workpiece
CN115963779A (en) Digital twinning
Ye et al. Geometric parameter optimization in multi-axis machining
CN109202539B (en) Online detection method for composite material weak-rigidity special-shaped structure
EP4121829A1 (en) Morphic manufacturing
JP2012176477A (en) Trajectory tracking system and method of machining robot
CN114310897B (en) Pose optimization and motion fairing calculation method, system and application of robot measurement
Barari et al. Evaluation of geometric deviations in sculptured surfaces using probability density estimation
Cho et al. Machining error compensation using radial basis function network based on CAD/CAM/CAI integration concept
CN112296753B (en) Iteration-based on-machine measurement workpiece clamping position alignment method
Marinescu et al. Adaptive dimensional control based on in-cycle geometry monitoring and programming for CNC turning center
Young et al. Implementation of a variable DH parameter model for robot calibration using an FCMAC learning algorithm
CN117348518B (en) Numerical control machine tool geometric error identification method and device based on improved nine-line method

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210430