CN114266286A - Online detection method and device for welding process information - Google Patents
Online detection method and device for welding process information Download PDFInfo
- Publication number
- CN114266286A CN114266286A CN202111365983.5A CN202111365983A CN114266286A CN 114266286 A CN114266286 A CN 114266286A CN 202111365983 A CN202111365983 A CN 202111365983A CN 114266286 A CN114266286 A CN 114266286A
- Authority
- CN
- China
- Prior art keywords
- welding
- prediction model
- characteristic information
- welding process
- correlation coefficient
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for detecting welding process information on line, which comprises the steps of firstly, acquiring dynamic image information data of a welding process molten pool shot by an industrial CCD camera by means of a software development platform; extracting characteristic information of a molten pool image through an image processing means; based on a deep learning algorithm, establishing the correlation between the characteristic information of the molten pool image and the welding state; and performing state recognition on the weld pool image by using an algorithm model obtained by a deep learning algorithm, and finally realizing weld quality evaluation. The invention aims at the problem that the randomness of image information change and the diversity of welding states in the welding process are difficult to describe accurately, can learn the correlation information between the images and the welding states by utilizing a neural network, and can adjust through a self-learning mechanism. The image information in the welding process is effectively decomposed, characterized and evaluated by using the characteristics of the neural network, so that a result which cannot be compared with the traditional method is obtained.
Description
Technical Field
The invention relates to an online detection method and device of welding process information, and belongs to the technical field of welding image analysis.
Background
The real-time information collection of the welding process has become a main means for identifying the welding state, and the collection of the image information of the welding process is gradually the mainstream as a detection means. However, the welding process is complex, the interference items are numerous, and in order to obtain useful molten pool characteristic information as a basis for measuring stability and defects of the welding process, a clear image with high acquisition frequency and a richer image processing means are required to be obtained, so that the data volume of image information is huge, the image information cannot be well corresponding to different welding states, and the inaccuracy of a process monitoring result is caused.
Most of the conventional detection methods match the weld pool image characteristic information with the welding state through empirical and regular cognition, and evaluate the weld quality through the change of the characteristic information. With the development of computer technology, more choices are provided as evaluation means, wherein the deep learning method is extremely suitable for being applied to the field of weld process quality evaluation as an emerging means for realizing artificial intelligence.
How to use artificial intelligence for the analysis of welding process information and more accurately and efficiently carry out the on-line detection of the welding state is a technical problem which needs to be solved urgently by technical personnel in the field.
Disclosure of Invention
The purpose is as follows: aiming at the problem that the randomness of image information change and the diversity of welding states in the welding process are difficult to describe accurately, the invention can learn the correlation information between the image and the welding state by utilizing a neural network and adjust the correlation information by a self-learning mechanism. The image information in the welding process is effectively decomposed, characterized and evaluated by using the characteristics of the neural network, so that a result which cannot be compared with the traditional method is obtained.
The technical scheme is as follows: in order to solve the technical problems, the technical scheme adopted by the invention is as follows:
in a first aspect, a method for online detection of welding process information includes the steps of:
and constructing a first prediction model, taking the characteristic information of the molten pool image and the welding state as input values of the first prediction model, and calculating a first correlation coefficient between the characteristic information and the welding state as an output value of the first prediction model.
And setting a preset value of a correlation coefficient, and selecting the welding state and the weld pool image characteristic information corresponding to the output value of the first prediction model larger than the preset value of the correlation coefficient as a first training set.
And constructing a second prediction model, wherein the input value of the second prediction model is molten pool image characteristic information, the output value of the second prediction model is a welding state, and the first training set is utilized to train the second prediction model to obtain the trained second prediction model.
Extracting actual weld pool image characteristic information and a welding state in the welding process, and inputting the actual weld pool image characteristic information into the trained second prediction model to obtain a predicted welding state; and calculating a second correlation coefficient of the actual welding state and the predicted welding state, taking the actual molten pool image characteristic information, the welding state and the second correlation coefficient as a second training set, training the first prediction model by using the second training set to obtain an updated first prediction model, and taking the second correlation coefficient as an updated preset value of the correlation coefficient.
And according to the updated first prediction model and the correlation coefficient preset value, welding state and molten pool image characteristic information corresponding to the output value of the updated first prediction model being larger than the updated correlation coefficient preset value are obtained again to serve as an updated first training set.
And training the second prediction model by using the updated first training set to obtain an updated second prediction model.
And inputting the real-time molten pool image characteristic information into the updated second prediction model to obtain the final predicted welding state.
Preferably, the weld pool image characteristic information is acquired by an industrial CCD camera, and includes but is not limited to the following geometric dimension characteristic information of the weld pool image on the surface of the welding process: molten bath width, molten bath length, molten bath area, keyhole length, keyhole area.
Preferably, the welding state includes, but is not limited to: good welding, unstable welding process, welding deviation, welding collapse, welding leakage and splashing.
Preferably, the welding states are represented by binary vectors, one column element of the vector corresponding to each state is 1, the other elements are 0, and the number of columns of the vector is equal to the number of the welding states.
Preferably, the preset value of the correlation coefficient is set to 0.5.
Preferably, when all the column elements in the vector values output by the second prediction model are not 1, the welding state corresponding to the numerical value closest to 1 in the column elements is taken as the output value.
Preferably, the weld pool image characteristic information can be replaced by welding process parameters.
Preferably, the welding process parameters include, but are not limited to: welding speed, welding power and welding angle.
In a second aspect, an apparatus for online detection of welding process information includes the following modules:
a first model building module: the method is used for constructing a first prediction model, the characteristic information of the molten pool image and the welding state are used as input values of the first prediction model, and a first correlation coefficient between the characteristic information of the molten pool image and the welding state is calculated as an output value of the first prediction model.
A training set generation module: and the method is used for setting a preset value of the correlation coefficient, and selecting the welding state and the weld pool image characteristic information corresponding to the output value of the first prediction model being larger than the preset value of the correlation coefficient as a first training set.
A second model building module: and the method is used for constructing a second prediction model, the input value of the second prediction model is the characteristic information of the molten pool image, the output value of the second prediction model is the welding state, and the first training set is utilized to train the second prediction model to obtain the trained second prediction model.
A model adjustment module: the weld condition prediction model is used for extracting actual weld pool image characteristic information and a weld condition in the welding process, and inputting the actual weld pool image characteristic information into the trained second prediction model to obtain a predicted weld condition; and calculating a second correlation coefficient of the actual welding state and the predicted welding state, taking the actual molten pool image characteristic information, the welding state and the second correlation coefficient as a second training set, training the first prediction model by using the second training set to obtain an updated first prediction model, and taking the second correlation coefficient as an updated preset value of the correlation coefficient.
A training set updating module: and the weld state and weld pool image characteristic information corresponding to the output value of the updated first prediction model being greater than the updated correlation coefficient preset value is obtained again as an updated first training set according to the updated first prediction model and the updated correlation coefficient preset value.
A second model update module: and the second prediction model is trained by utilizing the updated first training set to obtain an updated second prediction model.
A welding state evaluation module: and the real-time weld pool image characteristic information is input into the updated second prediction model to obtain the final predicted welding state.
Has the advantages that: according to the method and the device for detecting the welding process information on line, provided by the invention, an industrial CCD (charge coupled device) vision sensor is taken as a data acquisition source, surface molten pool image characteristic information is extracted through an image processing means, a correlation relation between a welding image and a welding state is established by utilizing a deep learning algorithm, a defect identification standard is formulated, and the accuracy of an on-line detection means is improved.
Drawings
Fig. 1 is a schematic diagram of an online detection process of welding process information.
FIG. 2 is a flow chart of a predictive model training algorithm.
Detailed Description
The present invention will be further described with reference to the following examples.
As shown in fig. 1-2, an online detection method for welding process information includes the following steps:
step 1: and acquiring dynamic image information of a surface molten pool in the welding process by an industrial CCD camera.
Step 2: extracting geometric dimension characteristic information of the molten pool image as molten pool image characteristic information by using image processing means, including but not limited to: molten bath width, molten bath length, molten bath area, keyhole length, keyhole area.
And step 3: the welding state types are preset, and include but are not limited to: good welding, welding deviation, welding collapse and unstable welding process, and different states are described through a four-dimensional vector mode. For example: the good welding corresponds to an ideal vector [ 1000 ], the welding deviation corresponds to an ideal vector [ 0100 ], the welding collapse corresponds to an ideal vector [ 0010 ], and the welding process is unstable and corresponds to [ 0001 ].
And 4, step 4: establishing a first prediction model, taking the characteristic information of the molten pool image and the welding state as input values of the first prediction model, calculating a first correlation coefficient between the characteristic information of the molten pool image and the welding state as an output value of the first prediction model, wherein the output value range is between 0 and 1, and the basis is used for judging the corresponding relation between the characteristic information of the molten pool image and the welding state.
And 5: according to daily production experience, setting the preset value of the correlation coefficient of the characteristic information of the molten pool image and the welding state to be 0.5, namely, when the output value obtained through the first prediction model is more than 0.5, the preset value is regarded as a 'true' criterion; and selecting the welding state information and the molten pool image characteristic information which are output by the first prediction model and are larger than the preset value as a first training set.
Step 6: establishing a second prediction model by using the obtained first training set, wherein the input value is molten pool image characteristic information, and the output value is a welding state; it should be noted that the output value passing through the first prediction model is not always the ideal vector value, and then the judgment criterion should be adjusted accordingly, and the vector with the largest component is regarded as the value closest to the corresponding welding state. For example, if the output obtained after the model is [ 0.20.80.30.6 ], wherein 0.8 is closest to 1, the welding state corresponding to the output is considered as the weld deviation.
And 7: after the welding process, extracting the corresponding characteristic information of the molten pool image, and simultaneously extracting the corresponding monitoring quantity, namely the actually obtained welding state.
And 8: and comparing the actually obtained monitored quantity with an output value obtained through a second prediction model to obtain a second correlation coefficient of the monitored quantity and the output value, taking the actually obtained monitored quantity, the molten pool image characteristic information and the correlation coefficient as a second training set, adopting the second training set to retrain the first prediction model to obtain an updated first prediction model, and taking the obtained second correlation coefficient as a preset value.
And step 9: and reselecting the first training data according to the updated first prediction model, and retraining the second prediction model by adopting the reselected first training data to obtain an updated second prediction model.
Step 10: and in the new welding process, the updated second prediction model is adopted to carry out welding state evaluation on the characteristic information of the molten pool obtained by image processing.
Optionally, according to the actual working condition, the welding state information includes but is not limited to: good welding, unstable welding process, welding deviation, welding collapse, welding leakage and splashing;
optionally, besides the weld pool characteristic parameters, the welding process parameters, such as welding speed, welding power, welding angle, etc., may also be used as input values of the prediction model.
Example (b):
firstly, acquiring dynamic image information data of a welding process molten pool shot by an industrial CCD camera by means of a software development platform; extracting characteristic information of a molten pool image through an image processing means; based on a deep learning algorithm, establishing the correlation between the characteristic information of the molten pool image and the welding state; and performing state recognition on the weld pool image by using an algorithm model obtained by a deep learning algorithm, and finally realizing weld quality evaluation.
The method effectively utilizes a deep learning algorithm to establish the relevance between the weld pool image and the welding state in the welding process, obtains a second prediction model according to training, and can be used for representing the mapping from the weld pool image information to the welding state. Therefore, the characteristic information of the molten pool in the welding process is obtained through image processing and is transmitted into the prediction model, and the corresponding state information can be obtained. Meanwhile, the accuracy of the prediction model is greatly improved through a feedback regulation mechanism, so that a high-efficiency and accurate online detection means for the welding process is obtained, the welding state can be evaluated in real time, and the welding risk is avoided.
The invention can also map the welding process information except the image information with the post-welding state, such as process parameters, sound signals, equipment state and the like. However, it should be noted that other welding process signals need to be processed correspondingly, so as to reduce the learning process of the prediction model and improve the prediction accuracy.
The invention can also simplify the characteristic signal model and the state information aiming at other prediction algorithms and aiming at the rules obtained by the actual welding condition and the daily engineering experience.
The welding process information online detection method provided by the invention is within the protection scope of the invention.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (16)
1. An online detection method for welding process information is characterized in that: the method comprises the following steps:
constructing a first prediction model, taking the characteristic information of the molten pool image and the welding state as input values of the first prediction model, and calculating a first correlation coefficient between the characteristic information and the welding state as an output value of the first prediction model;
setting a preset value of a correlation coefficient, and selecting welding state and molten pool image characteristic information corresponding to the output value of the first prediction model larger than the preset value of the correlation coefficient as a first training set;
constructing a second prediction model, wherein the input value of the second prediction model is molten pool image characteristic information, the output value of the second prediction model is a welding state, and the second prediction model is trained by utilizing the first training set to obtain a trained second prediction model;
extracting actual weld pool image characteristic information and a welding state in the welding process, and inputting the actual weld pool image characteristic information into the trained second prediction model to obtain a predicted welding state; calculating a second correlation coefficient of the actual welding state and the predicted welding state, taking the actual molten pool image characteristic information, the welding state and the second correlation coefficient as a second training set, training the first prediction model by using the second training set to obtain an updated first prediction model, and taking the second correlation coefficient as an updated correlation coefficient preset value;
according to the updated first prediction model and the correlation coefficient preset value, welding state and molten pool image characteristic information corresponding to the fact that the output value of the updated first prediction model is larger than the updated correlation coefficient preset value are obtained again and serve as an updated first training set;
training the second prediction model by using the updated first training set to obtain an updated second prediction model;
and inputting the real-time molten pool image characteristic information into the updated second prediction model to obtain the final predicted welding state.
2. The method for detecting the welding process information on line according to the claim 1, characterized in that: the weld pool image characteristic information is used for acquiring geometric dimension characteristic information of a weld pool image on the surface of a welding process through an industrial CCD camera, and comprises the following steps of: molten bath width, molten bath length, molten bath area, keyhole length, keyhole area.
3. The method for detecting the welding process information on line according to the claim 1, characterized in that: the welding conditions include, but are not limited to: good welding, unstable welding process, welding deviation, welding collapse, welding leakage and splashing.
4. The method for detecting the welding process information on line according to the claim 1, characterized in that: the welding states are represented by binary vectors, one column element in the vector corresponding to each state is 1, other elements are 0, and the number of columns of the vector is equal to the number of the welding states.
5. The method for detecting the welding process information on line according to the claim 1, characterized in that: the preset value of the correlation coefficient is set to 0.5.
6. The method for detecting the welding process information on line according to claim 4, wherein the method comprises the following steps: and when the column elements in the vector values output by the second prediction model are not 1, taking the welding state corresponding to the numerical value closest to 1 in the column elements as an output value.
7. The method for detecting the welding process information on line according to the claim 1, characterized in that: the weld pool image characteristic information can be replaced by welding process parameters.
8. The method for detecting the welding process information on line according to claim 7, wherein the method comprises the following steps: the welding process parameters include, but are not limited to: welding speed, welding power and welding angle.
9. The utility model provides an on-line measuring device of welding process information which characterized in that: the system comprises the following modules:
a first model building module: the device is used for constructing a first prediction model, taking the characteristic information of the molten pool image and the welding state as input values of the first prediction model, and calculating a first correlation coefficient between the characteristic information and the welding state as an output value of the first prediction model;
a training set generation module: the device is used for setting a preset value of a correlation coefficient, and selecting welding state and molten pool image characteristic information corresponding to the output value of the first prediction model larger than the preset value of the correlation coefficient as a first training set;
a second model building module: the method comprises the steps of constructing a second prediction model, training the second prediction model by utilizing a first training set to obtain a trained second prediction model, wherein the input value of the second prediction model is molten pool image characteristic information, the output value of the second prediction model is a welding state;
a model adjustment module: the weld condition prediction model is used for extracting actual weld pool image characteristic information and a weld condition in the welding process, and inputting the actual weld pool image characteristic information into the trained second prediction model to obtain a predicted weld condition; calculating a second correlation coefficient of the actual welding state and the predicted welding state, taking the actual molten pool image characteristic information, the welding state and the second correlation coefficient as a second training set, training the first prediction model by using the second training set to obtain an updated first prediction model, and taking the second correlation coefficient as an updated correlation coefficient preset value;
a training set updating module: the correlation coefficient preset value obtaining unit is used for obtaining a correlation coefficient preset value of the first prediction model, and obtaining weld state and weld pool image characteristic information corresponding to the fact that an output value of the first prediction model is larger than the correlation coefficient preset value as an updated first training set;
a second model update module: the second prediction model is trained by utilizing the updated first training set to obtain an updated second prediction model;
a welding state evaluation module: and the real-time weld pool image characteristic information is input into the updated second prediction model to obtain the final predicted welding state.
10. The on-line detection device for the welding process information according to claim 9, wherein: the weld pool image characteristic information is used for acquiring geometric dimension characteristic information of a weld pool image on the surface of a welding process through an industrial CCD camera, and comprises the following steps of: molten bath width, molten bath length, molten bath area, keyhole length, keyhole area.
11. The on-line detection device for the welding process information according to claim 9, wherein: the welding conditions include, but are not limited to: good welding, unstable welding process, welding deviation, welding collapse, welding leakage and splashing.
12. The on-line detection device for the welding process information according to claim 9, wherein: the welding states are represented by binary vectors, one column element in the vector corresponding to each state is 1, other elements are 0, and the number of columns of the vector is equal to the number of the welding states.
13. The on-line detection device for the welding process information according to claim 9, wherein: the preset value of the correlation coefficient is set to 0.5.
14. The on-line detection device for the welding process information according to claim 12, wherein: and when the column elements in the vector values output by the second prediction model are not 1, taking the welding state corresponding to the numerical value closest to 1 in the column elements as an output value.
15. The on-line detection device for the welding process information according to claim 9, wherein: the weld pool image characteristic information can be replaced by welding process parameters.
16. The on-line detection device for the welding process information according to claim 15, wherein: the welding process parameters include, but are not limited to: welding speed, welding power and welding angle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111365983.5A CN114266286A (en) | 2021-11-18 | 2021-11-18 | Online detection method and device for welding process information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111365983.5A CN114266286A (en) | 2021-11-18 | 2021-11-18 | Online detection method and device for welding process information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114266286A true CN114266286A (en) | 2022-04-01 |
Family
ID=80825153
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111365983.5A Pending CN114266286A (en) | 2021-11-18 | 2021-11-18 | Online detection method and device for welding process information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114266286A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117745718A (en) * | 2024-02-19 | 2024-03-22 | 工业云制造(四川)创新中心有限公司 | Information interaction method based on cloud manufacturing |
| CN117911416A (en) * | 2024-03-20 | 2024-04-19 | 广州市市政工程试验检测有限公司 | Welding quality online detection method, device, equipment and storage medium |
| CN117911416B (en) * | 2024-03-20 | 2024-05-31 | 广州市市政工程试验检测有限公司 | Welding quality online detection method, device, equipment and storage medium |
-
2021
- 2021-11-18 CN CN202111365983.5A patent/CN114266286A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117745718A (en) * | 2024-02-19 | 2024-03-22 | 工业云制造(四川)创新中心有限公司 | Information interaction method based on cloud manufacturing |
| CN117911416A (en) * | 2024-03-20 | 2024-04-19 | 广州市市政工程试验检测有限公司 | Welding quality online detection method, device, equipment and storage medium |
| CN117911416B (en) * | 2024-03-20 | 2024-05-31 | 广州市市政工程试验检测有限公司 | Welding quality online detection method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111468989B (en) | Five-axis linkage numerical control manipulator polishing control system and method | |
| US10518357B2 (en) | Machine learning device and robot system to learn processing order of laser processing robot and machine learning method thereof | |
| WO2019001220A1 (en) | Decision tree based turning process energy consumption prediction system and method of numerically controlled lathe | |
| CN114048600A (en) | Digital twin-driven multi-model fusion industrial system anomaly detection method | |
| US20180307203A1 (en) | Machining defect factor estimation device | |
| CN109472057B (en) | Product processing quality prediction device and method based on cross-process implicit parameter memory | |
| CN114329940A (en) | Continuous casting billet quality prediction method based on extreme learning machine | |
| JP2019162712A (en) | Control device, machine learning device and system | |
| JP2017130094A (en) | Cell control device, and production system for managing operation situation of multiple manufacturing machines in manufacturing cell | |
| CN116307289B (en) | Textile processing procedure parameter detection and prediction method, system and storage medium | |
| CN116822353A (en) | Digital twin model rapid construction method in manufacturing process | |
| CN116050678A (en) | Die-casting product processing test system and method based on cloud computing | |
| CN108931621B (en) | Zinc ore grade soft measurement method based on process texture characteristics | |
| CN116224947B (en) | Deep learning industrial control system capable of actively integrating domain rule knowledge | |
| CN114266286A (en) | Online detection method and device for welding process information | |
| CN112884717A (en) | System and method for real-time workpiece surface detection and tool life prediction | |
| CN115129003A (en) | Automatic production line intelligent monitoring system based on self-learning time-varying digital twin | |
| CN116703254B (en) | Production information management system for mechanical parts of die | |
| CN116213095B (en) | Intelligent clean coal product ash content adjusting method and system based on dense medium separation | |
| CN110956308A (en) | Chemical equipment health state prediction method based on time sequence | |
| CN115393288A (en) | Processing technology control system and method | |
| CN110174409B (en) | Medium plate periodic defect control method based on real-time detection result | |
| CN112612782A (en) | MES system data online filling method and system based on LSTM network | |
| CN117787480B (en) | Res-LSTM-based weld joint forming quality real-time prediction method | |
| CN117668670B (en) | Port lifting equipment fault diagnosis method and system |
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 |