CN108788560A - Welding system based on XGBoost machine learning models - Google Patents
Welding system based on XGBoost machine learning models Download PDFInfo
- Publication number
- CN108788560A CN108788560A CN201810897373.1A CN201810897373A CN108788560A CN 108788560 A CN108788560 A CN 108788560A CN 201810897373 A CN201810897373 A CN 201810897373A CN 108788560 A CN108788560 A CN 108788560A
- Authority
- CN
- China
- Prior art keywords
- welding
- machine learning
- model
- unit
- parameter
- 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
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K37/00—Auxiliary devices or processes, not specially adapted to a procedure covered by only one of the preceding main groups
Abstract
The present invention provides a kind of welding systems based on XGBoost machine learning models.The welding system based on XGBoost machine learning models includes weld bond quality sample data acquisition unit, data analysis unit, parameter adjustment unit, welding unit, control unit.Using the present invention, it can automatically determine welding system according to weld bond welding quality sample data and need the relevant parameter adjusted, and carry out parameter adjustment automatically, to adjust unfavorable welding parameter in time, improve the tubing welding quality and welding efficiency of oil pipeline.
Description
Technical field
The present invention relates to oil pipeline tubing welding technology fields, more specifically, are related to a kind of based on XGBoost machines
The welding system of learning model.
Background technology
With the development of petroleum gas and petrochemical industry, China's long distance pipeline construction also obtains high speed development.With
Be referred to as controlling methodology of oil pipeline in the pipe-line system for transporting oil and oil product, mainly by petroleum pipeline, oil transportation station and its
He forms related ancillary equipment.The pipeline transportation mode of crude oil and oil product and the railway and public affairs for belonging to land transportation mode
Road oil transportation is compared, and has the advantages that big freight volume, good airproof performance, at low cost and safety coefficient are high.
The tubing of oil pipeline is generally steel pipe, and long-distance pipe is connected into using attachment devices such as welding and flanges, and
Control and flow-rate adjustment is opened and closed using valve.Wherein, the welding quality of tubing is outstanding in controlling methodology of oil pipeline process of construction
It is important.
In existing oil pipeline tubing welding process, pipeline full automatic welding connection technology is generally used, using microprocessor
Computer realizes the control of butt welding machine structure, carrys out analog manual operation.It is generally divided into three basic modules：Power plant module, control
Module and auxiliary equipment module.Power plant module is made of generating set and the source of welding current；Control module includes system dynamic supply
Distributor, welding parameter controller and welding manipulator three parts, auxiliary equipment module are by protective gas mixer, electronic instrument
The compositions such as cooling and dehumidifying equipment, welding gun cooling equipment, programmable device and print out equipment.
Existing pipeline full automatic welding connection technology carries out certainly petroleum pipeline tubing according to prior defined welding process flow
Dynamic welding, technological standards are easy to control, welding efficiency is high.But it is various uncertain in face of construction environment, tubing laying condition etc.
The presence of factor can not automatically adjust the relevant parameter in welding process, for the equipment management and weld bond quality of tubing welding
Guarantee still has some problems.
Invention content
In view of the above problems, the object of the present invention is to provide a kind of welding systems based on XGBoost machine learning models.It can be with
According to every factor such as the environment of welding scene, installations and facilities, the parameters of welding equipment are adjusted in real time, to ensure that tubing welds
The weld bond quality connect.
Technical solution is as follows：
A kind of welding system based on XGBoost machine learning models, including：
Weld bond quality sample data acquisition unit, for acquiring weld bond quality sample data, the weld bond quality sample data are extremely
Include geographic orientation parameter, environmental parameter and welding equipment parameter less；
Data analysis unit：For carrying out data analysis to the weld bond quality sample data acquired；
Parameter adjustment unit：For adjusting welding parameter according to the aspect of model list；
Welding unit：For being carried out surrounding automatic welding to oil transportation tubing according to welding parameter；
Control unit：For to the weld bond quality sample data acquisition unit, the data analysis unit, the parameter adjustment
Unit and the welding unit carry out unified regulation and control.
Further, the data analysis unit includes：
Sample data classification subelement：For the weld bond quality sample data to be divided into training set and test set；
Machine learning unit：For the machine learning model using integrated study model XGBoost；
Aspect of model list forms unit：Model training and model for reaching desired effects according to machine learning model are surveyed
The parameters of test result determine the aspect of model list for influencing model prediction target.
Further, the machine learning unit, can according to the training set model training AUC/ROC values and
Accuracy rate, rate of precision and the recall rate that model measurement is carried out to the test set, comment the effect of machine learning model
Estimate.
Further, the aspect of model list forms unit, can be according to each aspect of model to the shadow of simulated target
The degree of sound is ranked up the aspect of model is descending, and choosing sequence and being determined in five features of first five influences model prediction mesh
Target aspect of model list.
Further, the parameter adjustment mode of described control unit control is divided into two operating modes：
1）Automatically or manually intervened by control unit so that the parameter between each work package can be passed normally
It passs, control whole system is according to instruction works；
2）By connecting data analysis unit, the parameter adjustment for obtaining 5 most criticals instructs, and automatically controls associated component work,
Improve welding quality.
Further, the welding unit includes welded rails, welding gun, around automatic soldering device, and the welding gun is fixed
On the circular automatic soldering device, motion track upper edge that the circular automatic soldering device is limited in the welded rails
The tube wall of oil transportation tubing moves in a circle.
Further, the circular automatic soldering device further comprises：
Walking mechanism：The walking mechanism includes clicking and gear drive unit；
Wire feeder：For conveying welding wire to the welding gun.
Further, the weld bond quality sample data acquisition unit includes：
GPS geo-location instrument：Geographical position information for confirming tubing welding；
Environment monitor：For implementing monitoring and acquire the temperature of tubing pad, humidity, oxygen concentration, gas concentration lwevel；
System data monitoring system：For monitoring the parameters around automatic soldering device.
Further, the system data monitoring system includes：
Motor monitoring device：Electric current and voltage for monitoring electronics in each structure of the welding system；
Wire feed monitoring device：Wire feed rate for monitoring the wire feeder, rotary inertia, dynamic property, driving torque；
Welding gun monitoring device：For monitor welding gun with respect to weld seam swing frequency, left and right end stop frequency, turn up and down
Dynamic frequency, welding gun deflection angle；
Orbit monitoring device：For monitoring smoothness and position degree around automatic soldering device walking.
Further, the data analysis unit is according to the collected items of weld bond quality sample data acquisition unit
Data are collected for information table, using GPS positioning parameter as coordinate, using descruotuve statu statistical method, analysis it is every with welding quality it
Between relationship, find and the relevant Multiple factors of welding quality between relationship.
Compared with the prior art, a kind of welding system based on XGBoost machine learning models of the present invention, it can basis
The items factor such as environment, installations and facilities of welding scene, carries out Data acquisition and issuance, adjusts every ginseng of welding equipment in real time
Number, carries out oil transportation tubing according to the welding parameter after adjustment to surround automatic welding, to ensure the weld bond quality of tubing welding.
Description of the drawings
Fig. 1 is a kind of logical schematic of welding system based on XGBoost machine learning models of invention；
Fig. 2 is a kind of flow chart of welding method based on XGBoost machine learning models of the present invention.
Specific implementation mode
Hereinafter, specific embodiments of the present invention will be described in detail with reference to the accompanying drawings.
Specifically, as an example, the present invention is used around automatic soldering device as the main of Oil-pipeline welding scene
Welding equipment drives welding gun to surround vessel wall motion along track in the case where pipeline is relatively fixed around automatic soldering device.
Fig. 1 shows a kind of logical schematic of the welding system based on XGBoost machine learning models of the present invention.
As shown in Figure 1, a kind of welding system based on XGBoost machine learning models includes mainly weld bond quality sample number
The control of unified regulation and control is carried out according to collecting unit 110, data analysis unit 120, parameter adjustment unit 130 and welding unit 140
Unit 150.
Weld bond quality sample data acquisition unit 110 includes at least geographic orientation parameter, environmental parameter and welding equipment ginseng
Number.
Welding unit includes welded rails, welding gun, surround automatic soldering device around automatic soldering device.
Wherein, around driving mechanism of the automatic soldering device using walking mechanism as automatic Arc Welding, around automatic welding
Connection device is mounted on welded rails, and the motion track upper edge tube wall limited in the welded rails with welding gun makees circumference fortune
It is dynamic, it is one of the important equipment for realizing nozzle automatic welding；Walking mechanism is made of motor and gear drive, according to calculating
The position and speed order-driven that machine control unit is sent out is moved around automatic soldering device.
The mechanism that welding wire is conveyed for welding gun is wire feeder, and the consistent level of wire feeder wire feed directly affects welding matter
Amount.The wire feeding mode of wire feeder can simply be divided into wire drawing and push away a two ways.
Wire drawing mode is that wire feeder is installed within welding tractor, and wire-feed motor is closer from the installation position of welding gun, welding
It is smaller to leave the resistance being subject to after wire-feed motor for welding wire in the process, therefore can ensure that wire feed process is steady.But wire feed set welding wire
Disk must be mounted on welding tractor, and the heavy base for increasing welding tractor increases difficulty to artificial assemble and disassemble, and weight
Increase is also easy to cause welding tractor walking unstable.The shallow bid welding wire of a diameter of 0.8mm or 1.0mm is used in the welding process
（Weight is about 5kg）, weight and the load of welding tractor can be alleviated, and can welding process be easy to control, but by
Smaller in the amount of shallow bid welding wire, the used time is short, has a certain impact to welding efficiency.
It is that wire feeder is installed on except welding tractor to push away a mode, and wire-feed motor farther out from welding gun, can reduce in this way
The volume and weight of welding tractor, can use power wire-feed motor and the deep bid welding wire of a diameter of 1.2ram (weight is about
20kg), to improve welding efficiency.However, due to when pushing away thread wire-feed motor farther out from welding gun, must there is wire feed hose phase between the two
Even, when welding wire is continuously pushed at torch head, the frictional resistance that welding wire is subject to is larger, moreover, wire feed is soft in welding process
The curvature of pipe has a certain impact to the consistent level of wire feed, wire feed can be caused unsmooth when serious, therefore when use pushes away thread needs
Fully consider above-mentioned factor.
Welded rails are fixedly mounted on oil pipeline tubing for walking and positioning special around automatic soldering device
Mechanism, the structure of welded rails directly influence the smoothness and position degree of movement of the welding ring around automatic soldering device, in turn
Influence welding quality.
Between the specific mechanical structure and welding tractor, welding gun, walking mechanism, welded rails of automatic soldering device
Position, operation relationship be that this field notes technology, of the invention focuses on based on XGBoost machine learning models to scene
Weld bond quality sample data carry out data analysis, and according to adjusted site welding equipment and remote control equipment welding join
Number, welding procedure is improved according to welding scene situation in real time.Therefore, for around automatic soldering device specific mechanical structure and
Machinery, the component relationship of its ancillary equipment, unless should not, do not do additional description.
GPS geographic orientations position indicator, environmental monitoring detector and system data monitoring system are as weld bond quality sample number
According to the capital equipment of collecting unit 110, for acquiring the weld bond quality sample used during construction site and remote control and regulation
Data, weld bond quality sample data include at least geographic orientation parameter, environmental parameter and welding equipment parameter.
Specifically, confirming the geographic orientation of pipeline welding according to GPS geo-location instrument, pass through data transmission channel and data
Analytic unit 120 connects.Environment monitor mainly monitors the temperature of pipeline pad, humidity, oxygen concentration, carbon dioxide in real time
Concentration etc., and connect with data analysis unit by data transmission channel.
System data monitoring system is used to monitor the parameters around automatic soldering device, including：
Motor monitoring device monitors the electric current of electronics in each structure, and voltage is to ensure motor in each position of pipeline circumferential weld
Accurate contraposition, and have the function of preferable speed tracing；
Wire feed monitoring device monitors the parameters of automatic wire bonder, such as wire feed rate, rotary inertia, dynamic property, driving turn
Square etc., to ensure wire feed accurate stable；
Welding gun monitoring device, monitoring torch swinging adjust frequency, monitoring welding gun with respect to weld seam swing frequency, left and right end stop
Frequency, rotational frequency up and down, welding gun deflection angle etc.；
Orbit monitoring device monitors smoothness and position degree around automatic soldering device walking；
System monitoring data system is connect by data transmission channel with data analysis unit.
In the specific embodiment of the present invention, the technical parameter of weld bond quality sample data includes following content：
Technical parameter -1：
Adapt to 610~1420mm of caliber Φ；Input voltage 220V；Frequency 50Hz；Power 240W；Gage of wire Φ 0.9~
1.2mm；0~36m/h of the speed of travel；0~30mm of amplitude of fluctuation.
Around the technical parameter -2 of automatic soldering device（The speed of travel therein, amplitude of fluctuation, swing speed, when stopping
Between and peak excursion be to preset, is adjustable）：
0.08~1.63m/min of the speed of travel；0~3.0cm of amplitude of fluctuation；Swing hastens 0~42mm/s；0~2m of residence time；
Weld waveform：Trapezoidal, triangle, rectangle, straight line；Maximum offset 70mm；Trolley weight 13kg；Load welding wire 5kg；Welding wire is straight
1.0~1.2mm of diameter Φ.
Technical parameter -3：
| Tubing | Welding wire | Protective gas | Gas flow |
| X52 | H08Mn2siA | Ar75%+CO225% | 20L/min |
| X65 | JM-58 | Ar80%+CO220% | 20L/min |
| X70 | JM-68 | Ar80%+CO220% | 20L/min |
Technical parameter -4:
| Welding current（A） | Weldingvoltage（v） | Weldering speed（cm/min） | Weld interval（min） | |
| Hot weld | 180~190 | 17~18 | 30 | 4.19 |
| Filling weldering | 210~220 | 18~19 | 26 | 4.52 |
| Cosmetic welding | 160~180 | 16~17 | 20 | 8.1 |
GPS geo-locations instrument, environmental monitoring monitor, the collected each item data of system data monitoring system are input to data
Analytic unit is collected and, using descruotuve statu statistical method, is analyzed each in information table using GPS positioning parameter as coordinate for information table
Relationship between independent data item and welding quality finds the relationship between the relevant Multiple factors of welding quality.
In the present invention, information table is just stored in a tables of data inside RAM, and content therein includes that data are fixed
Justice, data type, data description, data value, incidence relation etc..A information table not instead of individual table, according to
Each collection instrument or system it is different, can all preserve individual table.It is exactly what each information table linked together
GPS geo-locations reference axis and time shaft.
Specifically, as an example, the data for being input to data analysis unit for data analysis may include following content：
1）Weld bond information
Weld bond information can specifically include：Weld seam number, unit in charge of construction, construction unit, welder, supervising engineer, supervisor,
Weld date, Guan Chang, caliber, wall thickness, welding rod lot number, welding wire lot number, bevel angle, reinforcement；
2）Detection data
Detection data may include：Weld seam number, detection date, detection mode, detection type, evaluation result；
3）Welding machine and welding gun data
Welding machine and welding gun data may include：Weld seam number, welding equipment, speed of welding, soldering angle, welding current, welding
Voltage, wire feed rate.
Box traction substation is drawn for the parameter indices for influencing welding quality, with reference to corresponding historical data, comparison is corresponding
The range intervals of welding technology standards welding quality qualification and the running parameter of unqualified solder joint find possible influence welding
The abnormal parameters interval value of quality, and closed respectively with welding quality around the running parameter of each mechanism in automatic soldering device
Relationship between lattice.
The testing result of pipe welding port quality be divided into it is qualified with it is unqualified, by descriptive statistical analysis and establish machine
Learning model, exploration and weld bond correlative factor off quality, including single factor and Multifactor Combination.Machine learning mould
The prediction target of type is whether weld bond quality is qualified.
After the completion of data analysis, it is data analysis report, data analysis report that can arrange the result of data analysis
It include mainly the following contents：
1）Data analysis conclusion is described, and enumerates two class factors：
1.1）Influence weld bond factor off quality；
1.2）With the whether qualified incoherent factor of weld bond quality；
2）The mode that factor and factor combination are described with correlation between unqualified weld bond, including：
2.1）About the relevant description of single factor test：The accounting of unqualified weld bond is counted according to single factor；
2.2）About multifactor relevant description：Listing influences weld bond blocking factor off quality, and according to each factor
The size of butt welding mouth influence degree off quality does descending sort；
3）Enumerate the data analysing method used in this project, title containing machine learning model.
Analyzing the relationship between the running parameter and welding quality qualification of each mechanism in automatic soldering device
Later, it is ranked up according to running parameter welding quality influence degree is descending, eventually finds the influence weldering of first five of sorting
Main 5 factors for connecing quality adjust this 5 running parameters by control unit 150, reduce deviation value range, improve in real time
Nozzle welding quality.
Control unit 150 carries out parameter adjustment according to above-mentioned 5 running parameters to welding unit.The ginseng of control unit 150
Number adjustment mode is divided into two operating modes, first, the automatically or manually intervention for passing through control system so that each working group
Parameter between part can be transmitted normally, and control whole system is according to instruction works；Second is that by connecting data analysis unit, obtain
The parameter adjustment instruction for obtaining 5 most criticals, automatically controls associated component work, improves welding quality.
Data analysis unit 120 is adopted using XGBoost machine learning model butt welding mouth quality sample data acquisition units
The weld bond quality sample data of collection carry out data analysis, weld bond quality sample data are divided into training set and test set, selection can
Realize two classification XGBoost machine learning models, according to training set model training AUC/ROC values and to the test
Collection carries out accuracy rate, rate of precision and the recall rate of model measurement, assesses the effect of machine learning model；If model
The index value of the parameters of training and model test results shows that machine learning model reaches desired effects, then exports influence
The aspect of model list of model prediction target.Parameter adjustment unit adjusts welding parameter according to the aspect of model list.
Wherein, during weld bond sample quality data being divided into training set and test set, the standard of division is divided into two
Part.A part is technical documentation, such as industry specifications, such as《It field device, industy pipe welding engineering construction and tests
Receive specification（GBJ236—82）》；Product design technology standard, this is also the standard for examining product whether to work normally.Other one
Part is experienced technical specialist, introduces that several numerical value in really work, and sample needs are paid close attention to.Based on the two
Part, the data that could will be collected define data type, data value range, attribute, incidence relation etc., in next step
Data analysis is prepared.Last result is exactly the data comparison of on-site collection by historical data, find it is crucial it is several because
Then element exerts one's influence to it.
Correspondingly, as shown in Figure 1, data analysis unit may include：Sample data classification subelement 121, is used for institute
It states weld bond quality sample data and is divided into training set and test set；Machine learning unit 122, for using integrated study model
The machine learning model of XGBoost, according to the training set model training AUC/ROC values and to the test set carry out
Accuracy rate, rate of precision and the recall rate of model measurement, assess the effect of machine learning model；Aspect of model list shape
At unit 123, every ginseng of model training and model test results for reaching desired effects according to machine learning model
Number determines the aspect of model list for influencing model prediction target.
Wherein, XGBoost machine learning model butt welding mouth quality sample data acquisition units are used in data analysis unit
During the weld bond quality sample data acquired carry out data analysis, it is necessary first to carry out Data Mining, explore sample number
According to the quality of data, it is ensured that the quality of data of each data item meets data analysis and machine learning modeling demand.Pass through description
Property statistical analysis, explores the data distribution rule of each data item.
During the factor of analyzing influence welding quality, using descruotuve statu statistical method, each independent data item is analyzed
The relationship for the welding quality that picks with full-automatic pipe outside weld finds the Relationship of Related Factors for influencing welding quality, finds the 5 of most critical
A key factor.
During carrying out aspect of model screening, in addition to the letter directly collected using system data monitoring system number
Table is ceased, the historical data information table in data center, the welding quality qualification in welding technology standards has related pass to unqualified
The factor of system, it is also necessary to consider to derive other aspect of model using the parameter of these data, such as count each solder joint and welded
It is the temperature welded in journey, humidity, oxygen concentration, gas concentration lwevel, the Current Voltage of motor, wire feed rate, rotary inertia, dynamic
State property energy, driving torque, welding gun with respect to weld seam swing frequency, left and right end stop frequency, up and down rotational frequency, weldering
The variance of each parameter items such as rifle deflection angle, orbital data monitoring device, the smoothness of monitoring walking and position degree, mean value, crowd
The statistics such as number.If these derivative features also have correlativity with quality of welding spot qualification, machine learning mould can be added together
Type carries out model training and model measurement.
During data analysis unit carries out XGBoost machine learning modelings, first have to complete modeling target and mould
Then type carries out model training and model measurement, finally influence quality of welding spot according to model training and model test results output
Correlative factor.
The target that quality of welding spot analyzes machine learning model building is whether prediction quality of welding spot is qualified, this is that one two classification is asked
The modeling requirement of topic.The present invention, as machine learning model, has exactly taken a fancy to XGBoost using integrated study model XGBoost
The following advantages of model：
Regular terms：Regular terms is added in the cost function of model, model is avoided the case where over-fitting occur；
Parallel computation：Parallel processing and calculating are carried out in aspect of model level, each attribute gain is calculated with multithreading simultaneously
Row；
Beta pruning process：All subtrees that can be established first are established from the top to bottom, then are reversed beta pruning from bottom to top, avoid being absorbed in
Local optimum；
Row sampling：It supports row sampling, over-fitting can not only be reduced, moreover it is possible to reduce calculation amount, lift scheme calculating speed；
Missing values processing：There is the sample of missing for the value of feature, can learn its cleavage direction automatically.
During carrying out model training with model measurement, sample data can be divided into training set（70%）With test
Collection（30%）, select XGBoost models to carry out model training and model measurement, according to the AUC/ROC values and mould of model training
Accuracy rate, precision ratio and the recall ratio of type test, assess the effect of machine learning model.In addition, in sample data
The ratio of training set test set can also be adjusted according to specific model training and model measurement demand.
Output influence quality of welding spot correlative factor during, assessment models training and model test results it is each
The index value of assessment parameter can export and model if assessment result shows that machine learning model has certain effect
Predict target（Whether solder joint is qualified）Relevant aspect of model list.It is meant that representated by the aspect of model list of output, this
A little factors combine collective effect, and whether qualification has correlativity with quality of welding spot.In output model target correlated characteristic
When, it can be according to each aspect of model（Correlative factor）To simulated target（Whether solder joint is qualified）Influence significance level by greatly to
Small is ranked up.
Fig. 1 describes a kind of welding system based on XGBoost machine learning models of the present invention.It is opposite with the system
It answers, the present invention also provides a kind of welding method based on XGBoost machine learning models, this method can be combined with software and hardware
Mode realize.
As shown in Fig. 2, a kind of welding method based on XGBoost machine learning models provided by the invention, including it is as follows
Step：
Step S210：Acquire weld bond quality sample data；
Step S220：Data analysis is carried out to the weld bond quality sample data acquired, determines the mould for influencing model prediction target
Type feature list；
Step S230：Welding parameter is adjusted according to the aspect of model list；
Step S240：Oil transportation tubing is carried out according to the welding parameter after adjustment to surround automatic welding.
Wherein, in step S220, weld bond quality sample data is divided into training set and test set, selects and can be achieved two points
The machine learning model of class, according to training set model training AUC/ROC values and to the test set carry out model measurement
Accuracy rate, rate of precision and recall rate, the effect of machine learning model is assessed；If model training and model are surveyed
The index value of the parameters of test result shows that machine learning model reaches desired effects, then exporting influences model prediction target
Aspect of model list.
When output influences the aspect of model list of model prediction target, shadow of each aspect of model to simulated target is determined
The degree of sound, and be ranked up according to each aspect of model is descending to the influence degree of simulated target, selected and sorted is forward
The aspect of model as adjustment welding parameter foundation.
Furthermore, it is necessary to explanation, it is above-mentioned based on the welding method of XGBoost machine learning models in use
The process that XGBoost machine learning model butt welding mouth quality sample data are trained, learn, testing is all aforementioned to being based on
Detailed statement has been done in the description of the welding system of XGBoost machine learning models, and XGBoost engineerings are based on using above-mentioned
The welding method for practising model carries out welding used associated monitoring, data analysis, parameter adjustment, welding equipment also all aforementioned
To having done detailed statement in the description of the welding system based on XGBoost machine learning models, then this is repeated no more.
Compared with prior art, the present invention can be according to weld bond welding quality sample data, and automatically determining welding system needs
The relevant parameter to be adjusted, and parameter adjustment is carried out automatically, to adjust unfavorable welding parameter in time, improve oil pipeline
Tubing welding quality and welding efficiency.
Above-described is only some embodiments of the present invention.For those of ordinary skill in the art, not
Under the premise of being detached from the invention design, various modifications and improvements can be made, these belong to the protection model of the present invention
It encloses.
Claims (10)
1. a kind of welding system based on XGBoost machine learning models, including：
Weld bond quality sample data acquisition unit, for acquiring weld bond quality sample data, the weld bond quality sample data are extremely
Include geographic orientation parameter, environmental parameter and welding equipment parameter less；
Data analysis unit：For carrying out data analysis to the weld bond quality sample data acquired；
Parameter adjustment unit：For adjusting welding parameter according to the aspect of model list；
Welding unit：For being carried out surrounding automatic welding to oil transportation tubing according to welding parameter；
Control unit：For to the weld bond quality sample data acquisition unit, the data analysis unit, the parameter adjustment
Unit and the welding unit carry out unified regulation and control.
2. a kind of welding system based on XGBoost machine learning models as described in claim 1, it is characterised in that：It is described
Data analysis unit includes：
Sample data classification subelement：For the weld bond quality sample data to be divided into training set and test set；
Machine learning unit：For the machine learning model using integrated study model XGBoost；
Aspect of model list forms unit：Model training and model for reaching desired effects according to machine learning model are surveyed
The parameters of test result determine the aspect of model list for influencing model prediction target.
3. a kind of welding system based on XGBoost machine learning models as claimed in claim 2, it is characterised in that：It is described
Machine learning unit is additionally operable to basis and carries out mould to the AUC/ROC values of the training set model training and to the test set
Accuracy rate, rate of precision and the recall rate of type test, assess the effect of machine learning model.
4. a kind of welding system based on XGBoost machine learning models as claimed in claim 3, it is characterised in that：It is described
Aspect of model list forms unit, is additionally operable to the aspect of model according to each aspect of model to the influence degree of simulated target by big
It is ranked up to small, chooses sequence and determine the aspect of model list for influencing model prediction target in five features of first five.
5. a kind of welding system based on XGBoost machine learning models as claimed in claim 4, it is characterised in that：It is described
The parameter adjustment mode of control unit control is divided into two operating modes：
1）Automatically or manually intervened by control unit so that the parameter between each work package is normally transmitted, control
Whole system processed is according to instruction works；
2）By connecting data analysis unit, the parameter adjustment for obtaining 5 most criticals instructs, and automatically controls associated component work,
Improve welding quality.
6. a kind of welding system based on XGBoost machine learning models as claimed in claim 5, it is characterised in that：It is described
Welding unit includes welded rails, welding gun, around automatic soldering device, and the welding gun is fixed on the circular automatic soldering device
On, the circular automatic soldering device does circumference fortune in the tube wall for the motion track upper edge oil transportation tubing that the welded rails limit
It is dynamic.
7. a kind of welding system based on XGBoost machine learning models as claimed in claim 6, it is characterised in that：It is described
Further comprise around automatic soldering device：
Walking mechanism：The walking mechanism includes clicking and gear drive unit；
Wire feeder：For conveying welding wire to the welding gun.
8. a kind of welding system based on XGBoost machine learning models as claimed in claim 7, it is characterised in that：It is described
Weld bond quality sample data acquisition unit includes：
GPS geo-location instrument：Geographical position information for confirming tubing welding；
Environment monitor：For implementing monitoring and acquire the temperature of tubing pad, humidity, oxygen concentration, gas concentration lwevel；
System data monitoring system：For monitoring the parameters around automatic soldering device.
9. a kind of welding system based on XGBoost machine learning models as claimed in claim 8, it is characterised in that：It is described
System data monitoring system includes：
Motor monitoring device：Electric current and voltage for monitoring electronics in each structure of the welding system；
Wire feed monitoring device：Wire feed rate for monitoring the wire feeder, rotary inertia, dynamic property, driving torque；
Welding gun monitoring device：For monitor welding gun with respect to weld seam swing frequency, left and right end stop frequency, turn up and down
Dynamic frequency, welding gun deflection angle；
Orbit monitoring device：For monitoring smoothness and position degree around automatic soldering device walking.
10. a kind of welding system based on XGBoost machine learning models as claimed in claim 9, it is characterised in that：It is described
Data analysis unit is collected according to the collected each item data of the weld bond quality sample data acquisition unit for information table, with
GPS positioning parameter is coordinate, and using descruotuve statu statistical method, the every relationship between welding quality of analysis is found and welding
Relationship between the relevant Multiple factors of quality.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810897373.1A CN108788560A (en) | 2018-08-08 | 2018-08-08 | Welding system based on XGBoost machine learning models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810897373.1A CN108788560A (en) | 2018-08-08 | 2018-08-08 | Welding system based on XGBoost machine learning models |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108788560A true CN108788560A (en) | 2018-11-13 |
Family
ID=64079625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810897373.1A Pending CN108788560A (en) | 2018-08-08 | 2018-08-08 | Welding system based on XGBoost machine learning models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108788560A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109961106A (en) * | 2019-04-18 | 2019-07-02 | 北京百度网讯科技有限公司 | The training method and device of track disaggregated model, electronic equipment |
| CN113909636A (en) * | 2021-09-30 | 2022-01-11 | 北京博清科技有限公司 | Control method and control device of welding equipment, processor and welding system |
| CN117340495A (en) * | 2023-12-05 | 2024-01-05 | 山东力乐包装股份有限公司 | Equipment quality control system based on artificial intelligence |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102371414A (en) * | 2010-08-20 | 2012-03-14 | 中国海洋石油总公司 | Network-based pipeline all-position welding control system |
| CN102941395A (en) * | 2012-11-27 | 2013-02-27 | 上海工业自动化仪表研究院 | Welding process parameter recording and welding quality analysis system |
| CN104200319A (en) * | 2014-08-28 | 2014-12-10 | 北京中盈安信技术服务有限公司 | Oil-gas pipeline construction data monitoring method based on pipeline integrity management technology |
| CN105478976A (en) * | 2016-01-26 | 2016-04-13 | 清华大学 | Edge micro-plasma arc welding forming control method based on identification of dynamical system |
| CN106020092A (en) * | 2016-07-29 | 2016-10-12 | 成都熊谷加世电器有限公司 | Remote monitoring system of welding machine |
| US20170028499A1 (en) * | 2015-07-31 | 2017-02-02 | Fanuc Corporation | Machine learning device, arc welding control device, arc welding robot system, and welding system |
| CN107427951A (en) * | 2015-02-06 | 2017-12-01 | 拉普兰塔理工大学 | Utilize the welding system of adaptive algorithm |
| CN108027911A (en) * | 2015-07-29 | 2018-05-11 | 伊利诺斯工具制品有限公司 | Promote the system and method that welding is service software |
-
2018
- 2018-08-08 CN CN201810897373.1A patent/CN108788560A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102371414A (en) * | 2010-08-20 | 2012-03-14 | 中国海洋石油总公司 | Network-based pipeline all-position welding control system |
| CN102941395A (en) * | 2012-11-27 | 2013-02-27 | 上海工业自动化仪表研究院 | Welding process parameter recording and welding quality analysis system |
| CN104200319A (en) * | 2014-08-28 | 2014-12-10 | 北京中盈安信技术服务有限公司 | Oil-gas pipeline construction data monitoring method based on pipeline integrity management technology |
| CN107427951A (en) * | 2015-02-06 | 2017-12-01 | 拉普兰塔理工大学 | Utilize the welding system of adaptive algorithm |
| CN108027911A (en) * | 2015-07-29 | 2018-05-11 | 伊利诺斯工具制品有限公司 | Promote the system and method that welding is service software |
| US20170028499A1 (en) * | 2015-07-31 | 2017-02-02 | Fanuc Corporation | Machine learning device, arc welding control device, arc welding robot system, and welding system |
| CN106392266A (en) * | 2015-07-31 | 2017-02-15 | 发那科株式会社 | Machine learning device, arc welding control device, and arc welding robot system |
| CN105478976A (en) * | 2016-01-26 | 2016-04-13 | 清华大学 | Edge micro-plasma arc welding forming control method based on identification of dynamical system |
| CN106020092A (en) * | 2016-07-29 | 2016-10-12 | 成都熊谷加世电器有限公司 | Remote monitoring system of welding machine |
Non-Patent Citations (1)
| Title |
|---|
| 钱学君等: "决策树算法在焊接工艺设计中的应用", 《2008年安徽省科协年会机械工程分年会论文集》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109961106A (en) * | 2019-04-18 | 2019-07-02 | 北京百度网讯科技有限公司 | The training method and device of track disaggregated model, electronic equipment |
| CN109961106B (en) * | 2019-04-18 | 2022-03-18 | 北京百度网讯科技有限公司 | Training method and device of trajectory classification model and electronic equipment |
| US11599825B2 (en) | 2019-04-18 | 2023-03-07 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for training trajectory classification model, and electronic device |
| CN113909636A (en) * | 2021-09-30 | 2022-01-11 | 北京博清科技有限公司 | Control method and control device of welding equipment, processor and welding system |
| CN117340495A (en) * | 2023-12-05 | 2024-01-05 | 山东力乐包装股份有限公司 | Equipment quality control system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108817724A (en) | Welding method based on XGBoost machine learning model | |
| CN108788560A (en) | Welding system based on XGBoost machine learning models | |
| CN100584503C (en) | Quality control module for tandem arc welding | |
| CA2355625C (en) | Apparatus and method for welding pipes together | |
| CN110091029A (en) | Double-torch Intelligent welding system | |
| CN106862813B (en) | A kind of robot automatic surfacing of caterpillar teeth wheel disc and vision detection system | |
| CN113001070A (en) | Pipeline cooling system | |
| US11267065B2 (en) | Systems and methods providing pattern recognition and data analysis in welding and cutting | |
| CN107107130B (en) | System and method for estimating smoke quantity and content and reporting results | |
| CN107020436A (en) | System and method for monitoring welding quality | |
| CN108581207A (en) | A kind of the chevilled silk spacing real-time adaptive regulating system and adjusting method of laser-arc hybrid welding in industry | |
| CN111790967A (en) | Gas control system for welding, control method thereof and welding system | |
| CN106891111A (en) | A kind of robot closed loop processing system for the welding of fin panel casing pin | |
| CN110449696A (en) | Welding system and welding method | |
| CN106077940B (en) | Agitating friction soldering set and tilt adjustment method based on laser ranging and gesture feedback | |
| CN113001066B (en) | Real-time monitoring system and monitoring method for pipeline welding working condition | |
| CN103978280A (en) | Method for assessing welding technique of stainless steel compound pipe | |
| CN113032941A (en) | Monitoring system and method for natural gas pipeline pigging process | |
| CN207087127U (en) | Full automatic convex welding machine | |
| DE102021214320A1 (en) | System for machining workpieces and method for controlling a machine tool | |
| CN104842046B (en) | Heavy wall large scale tube shaped weld work piece seam of intersection line submerged arc welding platform courses method | |
| CN210604287U (en) | Powder making system and device for detecting concentration and particle size of pulverized coal | |
| CN208688438U (en) | A kind of device for on-line measurement burr in welded tube height | |
| CN103831509A (en) | Spiral submerged arc steel tube welding procedure qualification method | |
| CN116748729B (en) | Coal cutter spiral roller tooth holder welding method based on multiple tests |
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: 20181113 |