CN109656229B - Construction method of robot end performance prediction model based on GA-RBF network - Google Patents
Construction method of robot end performance prediction model based on GA-RBF network Download PDFInfo
- Publication number
- CN109656229B CN109656229B CN201811501005.7A CN201811501005A CN109656229B CN 109656229 B CN109656229 B CN 109656229B CN 201811501005 A CN201811501005 A CN 201811501005A CN 109656229 B CN109656229 B CN 109656229B
- Authority
- CN
- China
- Prior art keywords
- data
- rbf network
- robot
- prediction model
- rbf
- 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.)
- Active
Links
- 238000010276 construction Methods 0.000 title claims abstract description 6
- 238000012360 testing method Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims description 33
- 230000002068 genetic effect Effects 0.000 claims description 18
- 238000010606 normalization Methods 0.000 claims description 16
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 238000004891 communication Methods 0.000 claims description 8
- 210000000349 chromosome Anatomy 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000003745 diagnosis Methods 0.000 description 7
- 238000001514 detection method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 230000001360 synchronised effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000002955 isolation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000011056 performance test Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 239000008186 active pharmaceutical agent Substances 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 210000000707 wrist Anatomy 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0221—Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Genetics & Genomics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Physiology (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
- Numerical Control (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a construction method of a robot tail end performance prediction model based on a GA-RBF network, which comprises the steps of constructing an axis joint acquisition hardware platform for acquiring tail end data of a robot, and taking an EtherCAT bus and a laser tracker as auxiliary tools for tail end test as input and output data acquisition modes for training the GA-RBF network respectively; and acquiring the position, the speed and the torque of each joint in real time through a bus to obtain data serving as the input of a GA-RBF network, and acquiring tail end data serving as the output of the GA-RBF network by using a laser tracking coordinate measuring system to train a robot tail end performance prediction model based on the GA-RBF network. The invention greatly improves the acquisition precision of the shaft joint servo pulse, greatly improves the application of a subsequent RBF network on terminal data and the precision of calculating terminal parameters by a shaft joint data DH model, and has more practical significance for high-precision data research.
Description
Technical Field
The invention relates to practical engineering application of an intelligent algorithm, in particular to application of an intelligent algorithm for optimizing a radial basis function (GA-RBF) neural network (GA-RBF) based on a genetic algorithm to a robot terminal performance test.
Background
For the intense competition of the robot industry (here, mainly speaking, a six-degree-of-freedom mechanical arm) nowadays, most robot manufacturers urgently need a complete set of robot testing schemes in order to improve their own core competitiveness. Related tests of the robot also come along, and the error precision of the tail end of the robot can reflect the integral performance of the robot best, so that the tail end error precision is always a hot spot problem of the robot test. There are many factors that contribute to robot tip error: including errors caused by structural parameters, errors caused by motion variables; and inevitable random errors such as inertia, dead weight, vibration of a connecting rod and the like. In addition, there are errors caused by the external environment, especially temperature and wear problems; programming problems in the control system and control algorithm variability. The existence of various error factors inevitably influences the working precision of the robot, and correlation exists between the error factors.
The test of the robot does not form a perfect and approved test index. For example, most of domestic robot companies stay at a relatively basic level for testing robots, and external high-precision distance measuring instruments are still generally adopted, for example, a laser tracker measures and analyzes absolute positioning precision and repeated positioning precision of the robots, so that relatively systematic robot testing specifications are not formed, and relatively systematic testing steps are not formed. Meanwhile, in actual work, the precision test by using the laser tracker cannot be carried out on each robot.
In view of the above-proposed practical engineering problems, better solutions are sought. Therefore, the invention provides a robot terminal parameter calculation method based on a GA-RBF network. And obtaining an optimized RBF network according to the basic function width, the central vector and the weight vector optimized by the genetic algorithm, and realizing the prediction of the robot terminal data. The invention provides a construction method for obtaining the RBF network.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to solve the defects in the prior art and provides a method for constructing a terminal performance prediction model of a robot based on a GA-RBF network.
The technical scheme is as follows: the invention discloses a method for constructing a terminal performance prediction model of a robot based on a GA-RBF network, which comprises the following steps:
(1) an axis joint acquisition hardware platform for acquiring data at the tail end of the robot is built, and an EtherCAT bus and a laser tracker are used as auxiliary tools for tail end test and are respectively used as input and output data acquisition modes for training a GA-RBF network;
(2) and acquiring the position, speed and torque feedback of each joint in real time through an EtherCAT bus to obtain data serving as the input of a GA-RBF network, and acquiring tail end data serving as the output of the GA-RBF network by using a laser tracking coordinate measuring system to train a robot tail end performance prediction model based on the GA-RBF network.
Further, the shaft joint acquisition hardware platform include servo driver, RC controller, demonstrator and host computer, servo driver include a plurality ofly, each servo driver one end is connected with each joint for gather position, speed, the torque data of each joint, the other end gathers to the RC controller through the EtherCAT bus, the RC controller still is connected with demonstrator and host computer.
Furthermore, the servo driver comprises a therCAT communication module.
Furthermore, the laser tracking coordinate measuring system comprises a laser tracker host, a tracking camera, an upper computer and a T-Mac sensor, wherein the laser tracker host is respectively connected with the tracking camera, the upper computer and the T-Mac sensor through communication cables, and the system is used for collecting the tail end position and the speed parameter data of the robot in real time.
Further, the T-Mac sensor is arranged in the center of an end plane of the end flange of the robot.
Further, the process of forming the robot end performance prediction model comprises the following steps:
(1) selecting sample data: selecting and analyzing according to sample data, taking at least 240 pairs of data as RBF network training samples, namely taking the positions of six axis joints acquired by an EtherCAT real-time data acquisition bus as input, correspondingly, acquiring the tail end position by a laser tracker as output, and forming tail end position parameter training sample data;
(2) normalization: all data are converted into a [0,1] interval for normalization, and the normalization method is realized by a linear conversion function, as shown in the formula:
in the formula: x'kIs the input value, x, of the normalized converted networkkIs the original sample data, xmax、xminRespectively the maximum value and the minimum value in the original sample data;
(3) GA (genetic algorithm) optimizes RBF network parameters: the method comprises the steps of initializing parameter setting, initial population and chromosome coding, and GA optimization program;
(4) selecting a test sample: as a verification for the established RBF model, at least 50 groups of data are selected from the previously collected training data, or at least 50 groups of data are tested in addition to be used as test data samples;
(5) reverse normalization: and (3) substituting the test sample data into the terminal data prediction model based on the RBF network obtained by training, performing inverse normalization to obtain terminal data of the predicted robot, comparing the terminal data with the terminal data of the actual robot to confirm the feasibility of the model, and retraining the RBF network according to the steps (1) to (3) when the deviation between the predicted data and the measured data is large.
Further, the genetic initialization parameter setting in the step (3) includes: the population size N is 30, the evolution generation P is 100, the cross probability Pc is 0.8, and the variation probability Pm is 0.15.
Further, the initial population and the chromosome code in the step (3) specifically include: first, according to the general formula of the number of hidden layer nodes:
wherein m is the number of nodes of the output layer, and n is the number of nodes of the input layer, and the structure of the network is determined; in the process of building a prediction model based on the terminal parameters of the RBF network, the pulse of the six-freedom-degree axial joint is used as an input, so that an input node n is set to be 6, the terminal position is a three-dimensional spatial position, so that an output node m is set to be 3, and a is set to be [1,10 & lt/EN & gt ]]Within the range, for rounding setting a to 3, then h to 6; the network is a 6-6-3 structure, and then the genetic RBF center vector is 6 multiplied by 6, the basis function width vector is 6 multiplied by 1, and the weight vector is 6 multiplied by 3.
Further, the GA optimization procedure in the step (3) specifically adopts a genetic algorithm to optimize RBF network parameters to obtain a most suitable basis function width vector B meeting the precision requirementpCenter vector CpAnd a weight vector Wp。
Has the advantages that: the invention has the following beneficial effects:
1. the real-time industrial Ethernet is changed into a data acquisition bus from the traditional control bus function, so that the method is significant, the acquisition precision of the shaft joint servo pulse is greatly improved, and the high-precision data research is closer to the practical significance for the application of subsequent RBF network on the prediction of the tail end data;
2. by comparing the end parameters calculated by the DH model with the end parameters calculated by the GA-RBF neural network, it can be known that: the consistency of the end data calculated by the RBF network and the data actually measured by the laser tracker is better, and the data are more suitable for being used as the basis of subsequent index calculation;
3. the diagnosis result of the robot fault simulation experiment based on the GA-RBF network shows that the error between the test output and the expected output is small, the error probability is small, and the robot fault detection requirement can be met.
Drawings
FIG. 1 is a block diagram of EtherCAT axial joint parameter acquisition in the present invention;
FIG. 2 is a block diagram of laser tracker end parameter acquisition in accordance with the present invention;
FIG. 3 is a diagram of the hardware architecture of the DS5 servo driver of the present invention;
FIG. 4 is a schematic diagram of data acquisition synchronization according to the present invention;
FIG. 5 is a flowchart of the GA-RBF based end parameter prediction model acquisition process of the present invention;
FIG. 6 shows a common failure of the robot body according to the present invention;
fig. 7 is a fault diagnosis flow chart of fault sample data based on the RBF neural network in the present invention.
Detailed Description
The technical solution of the present invention will be further described in detail with reference to the following specific examples.
The invention discloses a GA-RBF network-based robot end performance test method, which comprises the following steps:
(1) an axis joint acquisition hardware platform for acquiring data at the tail end of the robot is built, and an EtherCAT bus and a laser tracker are used as auxiliary tools for tail end test and are respectively used as input and output data acquisition modes for training a GA-RBF network;
(2) and acquiring the position, speed and torque feedback of each joint in real time through an EtherCAT bus to obtain data serving as the input of a GA-RBF network, and acquiring tail end data serving as the output of the GA-RBF network by using a laser tracking coordinate measuring system to train a robot tail end performance prediction model based on the GA-RBF network.
The method for testing the terminal performance of the robot based on the GA-RBF network is specifically described as follows:
1. construction of robot tail end data acquisition hardware platform
1.1EtherCAT axial Joint data acquisition
As shown in the figures 1 and 3, the RC control driver and the DS servo driver are developed into EtherCAT master and slave stations, so that the parameters of six-axis joints of the robot can be acquired in real time, and the accuracy of the tail end test of the robot is improved. And then, reading the acquired position and speed data through an RC (remote control) controller, and finally analyzing the acquired data so as to facilitate analysis.
FIG. 1 shows, the axis joint gathers hardware platform and includes servo driver, RC controller (robot control), demonstrator and host computer, servo driver include a plurality ofly, each servo driver one end is connected with each joint for gather position, speed, the torque data of each joint, the other end gathers to the RC controller through the EtherCAT bus, the RC controller still is connected with demonstrator and host computer.
As shown in fig. 3, the servo driver includes a main controller, a rectifier module, an IPM inverter, a power module, an optical coupling isolation circuit, an EtherCAT communication module, a phase current detection, an encoder, a hall sensor, and a joint permanent magnet synchronous motor, where the input end of the main controller is connected to the phase current detection, the encoder, and the hall sensor, and both the encoder and the hall sensor are connected to the joint permanent magnet synchronous motor; the output end of the main controller is connected with an IPM inverter through an optical coupling isolation circuit, and the IPM inverter is also connected with phase current detection and a joint permanent magnet synchronous motor; the main controller is further connected with a power module and an EtherCAT communication module, and the power module is respectively connected with a direct-current switching power supply and an alternating-current power supply.
1.2 laser tracker collects robot end data
The whole process of acquiring the tail end data of the robot by the laser tracker is shown in figure 2, the laser tracking coordinate measuring system comprises a laser tracker host, a tracking camera, an upper computer and a T-Mac sensor (mechanical tracking), the laser tracker host is respectively connected with the tracking camera, the upper computer and the T-Mac sensor through communication cables, and the system is used for acquiring the tail end position and the speed parameter data of the robot in real time.
A T-Mac sensor in the laser tracking system is arranged in the center of an end plane of a robot tail end flange, and a communication cable is connected with an upper computer in a link mode, so that parameters such as tail end position and speed of the robot can be acquired.
2. Processing of data
2.1 synchronicity of data
Since sample data, i.e. the real-time acquired individual axis servo pulses and the laser tracker tip position and velocity parameters, are acquired by two different systems. Fig. 4 is a schematic diagram of a method for ensuring data acquisition synchronization. Assuming that the laser tracker starts to acquire the end data at the time t1, the EtherCAT bus starts to acquire the axis joint data at the time t2, and the robot starts to move at the time t3, since the data acquired by the laser tracker and the EtherCAT before the time t3 are acquired in the static state of the robot, the data are both the start point data, and the data are invalid data in the experiment and are all discarded. The data collected after the robot starts to move are used as effective data, and the tail end or the shaft joint takes the moment when the robot starts to move as a collection starting point, so that the synchronism of the data collected by the laser tracker and the EtherCAT is ensured.
2.2 selection of data samples
The screening of the data not only ensures that intervals exist among the data, but also ensures that enough data samples exist, every 200 points in the experiment are selected as sample data, the final sample data is 480 points, and the sample data is represented by a set S:
S＝{x|x＝Q(200k),k∈[1,480]} (1)
110 sample points on the PA trace, using the data set SPAAnd (4) showing. The test sample data is represented by set T:
T＝{x|x＝SPA(2n),n∈[1,55]} (2)
the training samples are represented by data set E:
E＝S-T (3)
3. prediction based on GA-RBF end position parameters
The invention adopts a terminal parameter prediction model based on a GA-RBF neural network to calculate the terminal parameters of the robot, and the terminal parameters of the robot comprise two parts: an end position parameter and an end velocity parameter. The RBF network-based end parameter prediction model acquisition process is shown in FIG. 5.
3.1 data samples
And selecting and analyzing according to sample data, taking 240 pairs of data as RBF network training samples, namely taking the positions of six axis joints acquired by an EtherCAT real-time data acquisition bus as input, and correspondingly, acquiring the tail end position by a laser tracker as output to form tail end position parameter training sample data.
3.2 normalization
In the experiment of the invention, in order to prevent the loss of some characteristic relations with low orders of magnitude, the original sample data must be standardized, and all data are converted into a [0,1] interval, namely, the normalization processing is carried out. The normalization method of the sample data in the experiment is realized by a linear transfer function, as shown in the formula:
in the formula: x'kIs a normalized conversionInput value, x, of the post-networkkIs the original sample data, xmax、xminRespectively the maximum and minimum values in the original sample data.
3.3GA optimized RBF networks
3.3.1 initialization parameter settings
Before genetic algorithm optimization, relevant initialization parameters in a program need to be set. The parameters of the genetic algorithm are adjusted by experience and continuous simulation experiments to determine appropriate values. The genetic initialization parameters are set as in table 1:
TABLE 1 genetic initialization parameter settings
3.3.2 initial population and chromosome coding
First, according to the general formula of the number of hidden layer nodes:
(m is the number of nodes of the output layer, and n is the number of nodes of the input layer) the structure of the network is determined. In the process of building a prediction model based on the end position parameters of the RBF network, the pulse of the six-freedom-degree axial joint is used as an input, so that an input node n is set to be 6, the end position is a three-dimensional space position, so that an output node m is set to be 3, and a is set to be [1,10 & lt/EN & gt ]]Within the range, for rounding, if a is 3, h is 6. The network is a 6-6-3 structure, and then the RBF center vector is inherited 6 multiplied by 6, the basis function width vector is inherited 6 multiplied by 1, and the weight vector is inherited 6 multiplied by 3.
3.3.3GA optimization program
The algorithm searches for the optimal individual by continuously calculating the fitness value of the individual, executes genetic operator operation, and continuously updates the initial parameters of the RBF network to obtain the optimized network.
And in the genetic algorithm optimizing process, the evolution times are increased, the errors are reduced, when the evolution times reach 20 times, the errors are below 0.02, the errors of the individuals are sorted, and the reciprocal is taken as the optimal fitness value.
Optimizing RBF network parameters by adopting a genetic algorithm to obtain a most adaptive basis function width vector B meeting the precision requirementpCenter vector CpAnd a weight vector Wp。
3.4 test specimens
As a verification for the established RBF model, at least 50 groups of training data can be selected from the previously acquired training data, and at least 50 groups of data can be tested additionally to be used as test data samples.
3.5 reverse normalization
Test sample data is substituted into a trained terminal data prediction model based on the RBF network, and because output prediction results are normalized values and are not output actual values, in order to perform precision analysis on the predicted output of the RBF model, the output needs to be subjected to inverse normalization.
3.6 model validation
And normalizing to obtain the terminal data of the predicted robot, comparing the terminal data of the predicted robot with the terminal data of the actual robot to confirm the feasibility of the model, and retraining the RBF network according to the steps 3.1-3.5 when the deviation between the predicted data and the actual measured data is large.
4. Prediction of network terminal speed parameter based on GA-RBF
The establishment of the terminal speed prediction model based on the GA-RBF network is quite similar to that of the terminal position model. The network establishment procedure for end velocity prediction and the initial parameters of the genetic algorithm refer to the network for end position prediction, except that the RBF network structure becomes 6-6-1 since the output layer has only one node.
Robot fault diagnosis of GA-RBF neural network
The present invention is only discussed briefly to classify a single fault type. The robot is a huge system, such as a control system, a detection system, an execution system and a pneumatic system, and a plurality of parts exist under each system. Since the number of parts is large and complicated, the types of failures are also large, and a common failure of an industrial robot is shown in fig. 6.
A large amount of data of common faults of the robot are utilized, sample data of the faults are obtained after the data are screened and classified, and a general fault model of the robot based on the RBF neural network is trained, so that the fault type of the robot is judged for the robot with unqualified terminal performance indexes. The diagnostic procedure is shown in fig. 7.
After the type of the fault is set and the vector mapping between the fault and the output is good, the GA-RBF fault diagnosis is realized in a similar way to the prediction process of the end data. The method selects feedback data of each axis servo as a monitoring variable, screens a large amount of fault data from a fault database tested by a robot as sample data, sets five types of fault state data according to fault types to form five state models, and takes the deviation between a fault servo pulse and a normal operation pulse as the input of the RBF network. And 5 fault data of the robot are selected, wherein the fault data comprise waist faults, large arm faults, elbow faults, small arm faults and wrist faults, and 10 groups of data of each fault are selected, so that 50 groups of training samples are total.
When the failure neural network is designed, the diagnosis parameters are initialized to be consistent with the previous parameters, but the output dimension is changed, so that the initial value of the network structure is set to be m-5, a-2.68, the network structure of 6-6-5 is obtained, the number of central variables is 36, the width parameters are still 6, and the number of chromosomes is also unchanged. Further, the error of the target is set to 0.0001.
Comparing the tested and expected outputs, the result of the RBF diagnosis is very close to the expected output, and basically completely consistent, and the error is only 0.0023. By selecting a large amount of test sample data, the accuracy of fault classification is about 99.585%, and therefore the conclusion that the self-adaptive GA-RBF can meet the requirement for fault diagnosis of the robot body can be drawn.
Although the present invention has been described with reference to a preferred embodiment, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (9)
1. The construction method of the robot end performance prediction model based on the GA-RBF network is characterized by comprising the following steps: the method comprises the following steps:
(1) an axis joint acquisition hardware platform for acquiring data at the tail end of the robot is built, and an EtherCAT bus and a laser tracker are used as auxiliary tools for tail end test and are respectively used as input and output data acquisition modes for training a GA-RBF network;
(2) and acquiring the position, speed and torque feedback of each joint in real time through an EtherCAT bus to obtain data serving as the input of a GA-RBF network, and acquiring tail end data serving as the output of the GA-RBF network by using a laser tracking coordinate measuring system to train a robot tail end performance prediction model based on the GA-RBF network.
2. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 1, wherein the method comprises the following steps: the shaft joint acquisition hardware platform comprises a plurality of servo drivers, an RC controller, a demonstrator and an upper computer, wherein one end of each servo driver is connected with each joint and used for acquiring position, speed and torque data of each joint, the other end of each servo driver is gathered to the RC controller through an EtherCAT bus, and the RC controller is further connected with the demonstrator and the upper computer.
3. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 2, wherein: the servo driver comprises an EtherCAT communication module.
4. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 1, wherein the method comprises the following steps: the laser tracking coordinate measuring system comprises a laser tracker host, a tracking camera, an upper computer and a T-Mac sensor, wherein the laser tracker host is respectively connected with the tracking camera, the upper computer and the T-Mac sensor through communication cables, and the system is used for collecting the tail end position and speed parameter data of the robot in real time.
5. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 4, wherein: the T-Mac sensor is arranged in the center of an end plane of a robot end flange.
6. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 1, wherein the method comprises the following steps: the forming process of the robot end performance prediction model comprises the following steps:
(1) selecting sample data: selecting and analyzing according to sample data, taking at least 240 pairs of data as RBF network training samples, namely taking the positions of six axis joints acquired by an EtherCAT real-time data acquisition bus as input, correspondingly, acquiring the tail end position by a laser tracker as output, and forming tail end position parameter training sample data;
(2) normalization: all data are converted into a [0,1] interval for normalization, and the normalization method is realized by a linear conversion function, as shown in the formula:
in the formula: x' k is an input value of the network after normalization conversion, xk is original sample data, and xmax and xmin are respectively a maximum value and a minimum value in the original sample data;
(3) GA (genetic algorithm) optimizes RBF network parameters: the method comprises the steps of initializing parameter setting, initial population and chromosome coding, and GA optimization program;
(4) selecting a test sample: as a verification for the established RBF model, at least 50 groups of data are selected from the previously collected training data, or at least 50 groups of data are tested in addition to be used as test data samples;
(5) reverse normalization: and (3) substituting the test sample data into the terminal data prediction model based on the RBF network obtained by training, performing inverse normalization to obtain terminal data of the predicted robot, comparing the terminal data with the terminal data of the actual robot to confirm the feasibility of the model, and retraining the RBF network according to the steps (1) to (3) when the deviation between the predicted data and the measured data is large.
7. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 6, wherein: the genetic initialization parameter setting in the step (3) comprises the following steps: the population size N is 30, the evolution generation P is 100, the cross probability Pc is 0.8, and the variation probability Pm is 0.15.
8. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 6, wherein: the initial population and the chromosome code in the step (3) specifically comprise: first, according to the general formula of the number of hidden layer nodes:
wherein m is the number of nodes of the output layer, and n is the number of nodes of the input layer, and the structure of the network is determined; in the process of building a prediction model based on the terminal parameters of the RBF network, the pulse of the six-freedom-degree axial joint is used as an input, so that an input node n is set to be 6, the terminal position is a three-dimensional spatial position, so that an output node m is set to be 3, and a is set to be [1,10 & lt/EN & gt ]]Within the range, for rounding setting a to 3, then h to 6; the network is a 6-6-3 structure, and then the genetic RBF center vector is 6 multiplied by 6, the basis function width vector is 6 multiplied by 1, and the weight vector is 6 multiplied by 3.
9. The method for constructing the GA-RBF network-based robot end performance prediction model according to claim 6, wherein: and (3) the GA optimization program specifically adopts a genetic algorithm to optimize RBF network parameters to obtain a base function width vector Bp, a center vector Cp and a weight vector Wp which meet the precision requirement and are most suitable.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811501005.7A CN109656229B (en) | 2018-12-10 | 2018-12-10 | Construction method of robot end performance prediction model based on GA-RBF network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811501005.7A CN109656229B (en) | 2018-12-10 | 2018-12-10 | Construction method of robot end performance prediction model based on GA-RBF network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109656229A CN109656229A (en) | 2019-04-19 |
| CN109656229B true CN109656229B (en) | 2022-03-15 |
Family
ID=66113409
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811501005.7A Active CN109656229B (en) | 2018-12-10 | 2018-12-10 | Construction method of robot end performance prediction model based on GA-RBF network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109656229B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110501903B (en) * | 2019-09-12 | 2022-09-23 | 南京邮电大学 | Self-adjusting and optimizing method for parameters of robot inverse solution-free control system |
| CN114648148A (en) * | 2020-12-18 | 2022-06-21 | 广东博智林机器人有限公司 | Parameter prediction method and device for robot, storage medium and processor |
| CN113673153B (en) * | 2021-08-11 | 2024-08-13 | 北京小米移动软件有限公司 | Method and device for determining electromagnetic torque of robot, storage medium and electronic device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1382997A (en) * | 2002-06-13 | 2002-12-04 | 上海交通大学 | Precise tracking method based on nerve network for moving target |
| JP2010054429A (en) * | 2008-08-29 | 2010-03-11 | Mitsutoyo Corp | Laser tracker |
| CN101791801A (en) * | 2010-01-15 | 2010-08-04 | 广东工业大学 | Industrial robot motion planning and performance testing system and implementation method thereof |
| CN106705956A (en) * | 2017-02-28 | 2017-05-24 | 南京工程学院 | Rapid industrial robot tail end pose measuring device and measuring method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101680743B (en) * | 2007-03-05 | 2012-05-02 | 绝对机器人技术有限公司 | Determining positions |
-
2018
- 2018-12-10 CN CN201811501005.7A patent/CN109656229B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1382997A (en) * | 2002-06-13 | 2002-12-04 | 上海交通大学 | Precise tracking method based on nerve network for moving target |
| JP2010054429A (en) * | 2008-08-29 | 2010-03-11 | Mitsutoyo Corp | Laser tracker |
| CN101791801A (en) * | 2010-01-15 | 2010-08-04 | 广东工业大学 | Industrial robot motion planning and performance testing system and implementation method thereof |
| CN106705956A (en) * | 2017-02-28 | 2017-05-24 | 南京工程学院 | Rapid industrial robot tail end pose measuring device and measuring method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109656229A (en) | 2019-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109656229B (en) | Construction method of robot end performance prediction model based on GA-RBF network | |
| CN110000787B (en) | Control method of super-redundant mechanical arm | |
| CN113192112B (en) | Partial corresponding point cloud registration method based on learning sampling | |
| US20140343727A1 (en) | Robot positioning | |
| CN110815206A (en) | Stewart type parallel robot kinematics calibration method | |
| CN113168564A (en) | Method and system for generating artificial intelligence model | |
| CN113910218B (en) | Robot calibration method and device based on kinematic and deep neural network fusion | |
| CN106610584A (en) | Remanufacturing process quality control method based on neural network and expert system | |
| CN115514679B (en) | Abnormal source monitoring method and system based on communication module | |
| Mohebbi et al. | Trends in concurrent, multi-criteria and optimal design of mechatronic systems: A review | |
| CN115135463A (en) | Prediction model learning method, device and system for industrial system | |
| CN117901104A (en) | Feature active learning-feedback digital twin method for mechanical arm performance monitoring and error compensation | |
| CN113799137B (en) | Mobile processing robot rapid visual positioning method based on neural network | |
| Indri et al. | Integration of a production efficiency tool with a general robot task modeling approach | |
| CN106446502B (en) | The time-varying operation mode online recognition method of feature vector recursion with forgetting factor | |
| Lu et al. | Digital twin-based operation simulation system and application framework for electromechanical products | |
| CN114609969B (en) | Numerical control machine tool track error compensation method based on cloud computing | |
| Guerra-Zubiaga et al. | Computer vision and machine learning to create an advanced pick-and-place robotic operation using industry 4.0 Trends | |
| CN116337322A (en) | Array type force sensor calibration method and measuring method and device | |
| Huang et al. | Research on the kinematic calibration of the 3-PTT parallel mechanism | |
| CN115256397A (en) | Mechanical arm force control method and system based on deep learning | |
| Liu et al. | Research on intelligent control system of manipulator based on multi degree of freedom | |
| CN114415575A (en) | Real-time data-driven three-dimensional virtual monitoring and intelligent early warning system for welding workshop | |
| CN112380655A (en) | Robot inverse kinematics solving method based on RS-CMSA algorithm | |
| CN117621090B (en) | Industrial robot control method and system and industrial robot |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240808 Address after: 230000 B-1015, wo Yuan Garden, 81 Ganquan Road, Shushan District, Hefei, Anhui. Patentee after: HEFEI MINGLONG ELECTRONIC TECHNOLOGY Co.,Ltd. Country or region after: China Address before: 226000 No. 9 Siyuan Road, Chongchuan District, Nantong City, Jiangsu Province Patentee before: NANTONG University Country or region before: China |
|
| TR01 | Transfer of patent right |