CN115781136B - Intelligent recognition and optimization feedback method for welding heat input abnormality - Google Patents
Intelligent recognition and optimization feedback method for welding heat input abnormality Download PDFInfo
- Publication number
- CN115781136B CN115781136B CN202310053304.3A CN202310053304A CN115781136B CN 115781136 B CN115781136 B CN 115781136B CN 202310053304 A CN202310053304 A CN 202310053304A CN 115781136 B CN115781136 B CN 115781136B
- Authority
- CN
- China
- Prior art keywords
- heat input
- abnormal
- abnormality
- welding
- model
- 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
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an intelligent recognition and optimization feedback method for welding heat input abnormality, which comprises the steps of firstly collecting time sequence data in a normal welding process, obtaining a welding heat input data set through calculation, and training a welding heat input abnormality recognition model; the welding heat input abnormality detection model comprises an abnormality identification model and an abnormality classification model; training an abnormal recognition model by inputting heat input data during normal welding, and acquiring an abnormal recognition threshold; then, inputting a welding heat input set with abnormality into an abnormality recognition model, extracting an abnormality sample window and training an abnormality classification model; when the abnormal heat input detection device works, the abnormal recognition model receives heat input data in the actual welding process, and when the heat input is abnormal, a window sample where the abnormal heat input is located is output to the abnormal classification model; the abnormality classification model identifies specific categories of heat input abnormalities and gives out targeted adjustment and optimization strategies such as current, voltage, welding speed adjustment and the like according to different categories.
Description
Technical Field
The invention belongs to the technical field of intelligent welding, and particularly relates to an intelligent recognition and optimization feedback method for welding heat input abnormality.
Background
With the intelligent and large-scale development of welding, the number of welding automation equipment represented by a welding robot is greatly increased, and the application is increasingly wide. The automatic welding equipment is rapidly developed, is widely applied to industries such as high-speed locomotives, engineering machinery, steel structures, household appliances, automobiles and the like, and occupies a main position.
The automatic welding is required to follow the welding technological procedure WPS, and the WPS can provide reasonable range for various welding technological parameters including voltage and current according to different technological requirements. The core is to control the welding heat input within a reasonable range. However, in the prior art, a mature method is not formed for recognition of welding heat input abnormality and a control strategy to be adopted when abnormality occurs, and specific correspondence cannot be achieved for specific adjustment methods of different welding heat input abnormalities.
Disclosure of Invention
The invention aims to: aiming at the problems in the background technology, the invention provides an intelligent recognition and optimization feedback method for welding heat input abnormality. Through designing a welding heat input abnormality detection model comprising a heat input abnormality recognition model and an abnormality classification model, firstly, recognizing a window sample corresponding to an abnormality through the abnormality recognition model, then, specifically recognizing which abnormality belongs to through the abnormality classification model, and adopting a corresponding feedback regulation strategy.
The technical scheme is as follows: a welding heat input abnormality intelligent identification and optimization feedback method comprises the following steps:
step S1, firstly, collecting high-frequency time sequence data in the actual welding process, including current, voltage and welding speed; calculating and acquiring a welding heat input set; cutting the welding heat input set to obtain a plurality of window samples;
s2, constructing a welding heat input abnormality detection model; the welding heat input abnormality detection model comprises an indRNN-GAN-based abnormality identification model and an abnormality classification model; inputting the window sample into an anomaly identification model based on indRNN-GAN, extracting an anomaly heat input window sample when the anomaly identification model identifies that the window sample has heat input anomalies, and inputting the anomaly heat input window sample into an anomaly classification model; the abnormal classification model adopts a machine learning classification algorithm model; judging the abnormal state of the abnormal heat input window sample according to the difference of the actual abnormal conditions by the abnormal classification model; according to different abnormal states, different feedback regulation strategies are given;
and step S3, deploying the welding heat input abnormality detection model trained in the step S2 to the edge side, receiving high-frequency time sequence data in the actual welding process, judging what abnormal conditions belong to when the welding heat input abnormality exists, and giving out a corresponding feedback regulation strategy.
Further, the welding heat input set calculation method in the step S1 is as follows:
Heatinput = IU/V
wherein I is welding current, U is welding voltage, and V is welding speed; and according to the actual welding condition, sliding and cutting the welding heat input set obtained through calculation according to the time window size t and the time step size s, and finally obtaining a plurality of window samples under the conditions of normal heat input and abnormal heat input.
Further, in the step S2, the anomaly identification model based on indRNN-GAN is used to generate an countermeasure network structure GAN as a model body, which includes two parts of generating a network and discriminating the network; the generating network part adopts an Autoencoder architecture and comprises a decoder and an encoder which are sequentially connected, and an independent circulating network indRNN module is adopted; the discrimination network part adopts an MLP classifier, and the main structure comprises a linear layer, an activation function layer, a linear layer and a sigmoid layer which are connected in sequence; wherein the activation function layer employs a LeakyRelu function.
Further, the independent loop network indRNN module comprises a plurality of branches, each branch comprising a weight layer, a BN layer, a Relu activation function layer and a BN layer stacked in sequence.
Further, the training of the anomaly identification model based on indRNN-GAN in the step S2 specifically comprises the following steps:
s2.1, dividing a model entering data set, selecting a certain proportion from the normal welding heat input window samples obtained in the step S1 as a training set sample, and mixing the residual window samples with the abnormal heat input window samples to obtain a test set sample;
step S2.2, continuously iterating normal window samples based on training set samples, and calculating a mean square error MSE between each generation network generation data set and a real data set; after iterating the complete normal window samples, obtaining a set { MSE }; taking the 3/4+1.5IQR bit number of { MSE } as an anomaly threshold; after the window sample is input, generating a situation that the input window sample has abnormal heat input when the mean square error between the data set and the real window sample data set is more than the 3/4+1.5IQR bit number of { MSE }, otherwise, the situation is the normal heat input situation;
s2.3, after model training is completed, further performing model test; inputting test set data, and reconstructing the data set through a generation network; calculating a mean square error MSE between the reconstructed data set and the input data set; and judging based on the mean square error and the abnormal threshold value, determining an abnormal heat input window sample corresponding to the test set, and inputting the abnormal heat input window sample into an abnormal classification model for training a machine learning classification algorithm model.
Further, the welding abnormal heat input conditions in the step S2 comprise three conditions of point abnormality, abrupt abnormality and gradual abnormality; the corresponding feedback adjustment strategy includes: when the generated heat input abnormality belongs to the point abnormality, the system does not perform feedback regulation; when the generated heat input abnormality belongs to the abrupt change abnormality, the welding heat input is returned to the heat input level before the abrupt change by controlling the current, the voltage and the welding speed; when the abnormal heat input occurs belongs to the gradual change abnormality, the welding heat input is restored to the level before gradual change in a shorter time by controlling the current, the voltage and the welding speed as well.
Further, in the step S3, the preprocessed welding heat input data set is input into an anomaly identification model based on indRNN-GAN, and a window sample with an anomaly heat input is output; then the abnormal classification model receives a window sample with abnormal heat input, outputs a classification result, and confirms what kind of abnormality the window sample belongs to; and finally, carrying out corresponding adjustment based on the feedback adjustment strategy provided in the step S2.
Compared with the prior art, the technical scheme adopted by the invention has the following beneficial effects:
(1) According to the invention, a welding heat input abnormality detection model comprising an indRNN-GAN-based abnormality recognition model and an abnormality classification model is designed, a welding heat input abnormality recognition mechanism is firstly established through the abnormality recognition model, divided window samples are judged, samples with welding heat input abnormality are recognized, the samples are further input into the abnormality classification model, which abnormal conditions belong to are determined, and regulation is performed according to a corresponding feedback regulation strategy.
(2) The invention adopts the generation countermeasure network structure in the anomaly identification model part to carry out unsupervised anomaly identification. The generation network part adopts an indRNN module, so that the depth of the network structure can be effectively improved, and the problems of gradient disappearance and gradient explosion can be avoided by using a relu and other unsaturated functions as an activation function.
(3) The invention classifies the abnormal conditions possibly existing in the heat input in the actual welding process, gives corresponding feedback adjustment strategies according to different types of abnormalities, judges the abnormal conditions by an abnormality classification model, and can realize real-time feedback and accurate adjustment.
Drawings
FIG. 1 is a flow chart of a method for recognizing and adjusting abnormal welding heat input according to the present invention;
FIG. 2 is a schematic diagram of an anomaly identification model based on indRNN-GAN according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an indRNN module according to an embodiment of the present invention;
FIG. 4a is a schematic diagram of a thermal input waveform when a point anomaly (sudden rise) occurs in an embodiment of the present invention;
FIG. 4b is a schematic diagram of a thermal input waveform when a point anomaly (dip) occurs in an embodiment of the present invention;
FIG. 5a is a schematic diagram of a thermal input waveform when a sudden abnormality (sudden drop) occurs in an embodiment of the present invention;
FIG. 5b is a schematic diagram of a thermal input waveform when a sudden abnormality (sudden rise) occurs in an embodiment of the present invention;
FIG. 6a is a schematic diagram of a thermal input waveform when a gradual abnormality (dip) occurs in an embodiment of the present invention;
FIG. 6b is a schematic diagram of a thermal input waveform when a gradual abnormality (sudden rise) occurs in an embodiment of the present invention;
FIG. 7 is a flowchart of an exemplary machine learning classification algorithm model training method provided in an embodiment of the present invention.
Description of the embodiments
The invention is further explained below with reference to the drawings.
The invention discloses an intelligent recognition and optimization feedback method for welding heat input abnormality, which comprises the steps of firstly collecting time sequence data in a normal welding process, and obtaining a welding heat input data set through calculation; sliding window sampling is performed based on the set for training a welding heat input anomaly identification model. The welding heat input abnormality detection model designed by the invention comprises an indRNN-GAN-based abnormality identification model and an abnormality classification model. And training an anomaly identification model based on indRNN-GAN by inputting heat input data during normal welding, and acquiring an anomaly identification threshold value. The welding heat input dataset with anomalies is then input to the anomaly identification model, through which an anomaly sample window is extracted. Marking the actual abnormal situation according to the extracted abnormal sample window; extracting time domain features of the abnormal sample window, differentiating the abnormal sample window, extracting statistical features, and training an abnormal classification model based on the extracted features. The abnormal classification model adopts a machine learning classification model. After training is completed, the abnormal recognition model based on indRNN-GAN receives heat input data in the actual welding process, and when the heat input is abnormal, a window sample where the abnormal heat input is located is output to an abnormal classification model; the anomaly classification model identifies specific categories of heat input anomalies and adopts different feedback adjustment modes according to different categories. As shown in fig. 1, the method comprises the following steps:
and S1, collecting high-frequency time sequence data in the actual welding process, including current, voltage and welding speed. Based on the high frequency time series data, welding heat input data Heatinput can be obtained by:
Heatinput = IU/V
wherein I is welding current, U is welding voltage, and V is welding speed. The high-frequency time sequence data is divided into two parts of normal welding and abnormal welding, and the acquired heat input data correspondingly comprises two parts of normal heat input and abnormal heat input.
And performing segmentation processing on the obtained welding heat input data set. Specifically, the size of a time window is selected as t, sliding segmentation data is carried out according to the size s of a time step, and finally a plurality of window samples under the conditions of normal heat input and abnormal heat input are respectively obtained.
S2, constructing a welding heat input abnormality detection model and training;
in the embodiment, a welding heat input abnormality recognition model comprising an indRNN-GAN-based abnormality recognition model and an abnormality classification model is designed; the specific structure of the two-part model is described in detail below;
(1) Anomaly identification model based on indRNN-GAN
The model takes the generated countermeasure network structure GAN as a model main body, and comprises two parts of a generation network and a discrimination network, wherein the two parts are specifically shown in figure 2; the generating network part adopts an Autoencoder architecture and comprises a decoder and an encoder which are sequentially connected, and an independent cyclic network indRNN module is adopted. The indRNN module has a specific structure shown in FIG. 3 and comprises a plurality of branches, wherein each branch comprises a weight layer, a BN layer, a Relu activation function layer and a BN layer which are stacked in sequence.
Compared with the traditional RNN neural network, the invention adopts the indRNN network as the main body structure of the generator, and has the following effects:
(1) Because each neuron in the indRNN network is relatively independent, the behavior of each layer of indRNN neurons can be easily explained, and meanwhile, the indRNN network can keep long-term memory, so that the network structure depth can be improved, and the model effect can be effectively improved.
(2) The indRNN can well utilize the unsaturated functions of relu, etc. as activation functions and is very robust after training. The traditional RNN network has the problems of gradient disappearance and gradient explosion along with the deepening of the training degree, while the indRNN network does not have the problems.
Based on the characteristics, the GAN constructed by adopting the indRNN module can reconstruct training data more effectively, and the effect is better than that of the GAN constructed by using the RNN module.
The discrimination network part adopts an MLP classifier, and the main structure comprises a linear layer, an activation function layer, a linear layer and a sigmoid layer which are sequentially connected. Wherein the activation function layer employs a LeakyRelu function. And inputting the input real window sample data and the generated data set output by the generating network into the judging network for judgment, wherein the finally realized effect is that the judging network cannot respectively realize the real window sample and the generated data set, and the reconstruction effect of the representing generating network reaches the requirement.
In the model training phase, the input model dataset is first partitioned. For the normal heat input window samples obtained under the normal welding condition obtained in the step S1, a certain proportion is selected as a training set sample, the remaining samples are mixed with the abnormal heat input samples to be used as test set samples, and the proportion of the training set samples in the embodiment to all the normal heat input window samples is 80%. The normal window samples are iterated continuously based on the training set samples, and the mean square error MSE between the network generated data set and the real data set is calculated each time. After iterating the full normal window samples, the set { MSE } is obtained. The anomaly threshold is 3/4+1.5IQR bits of { MSE }. The meaning is as follows: when the window samples are input, a mean square error between the data set and the real window sample data set is generated to be greater than the 3/4+1.5IQR bit number of { MSE }, which represents the condition that abnormal heat input exists in the input window samples.
After model training is completed, further performing model test. Inputting test set data, and reconstructing the data set through a generation network. A mean square error MSE between the reconstructed data set and the input data set is calculated. And judging based on the mean square error and the abnormal threshold value, and determining an abnormal heat input window sample corresponding to the test set. The purpose of the anomaly identification model is to identify window samples when heat input anomalies occur.
(2) An anomaly classification model.
After the abnormality recognition model recognizes the window sample when the heat input is abnormal, it is necessary to further classify the abnormality for different abnormality situations. The purpose of the classification is to take different feedback adjustment measures according to the abnormal conditions of the actual welding heat input.
During actual welding, there are typically 3 situations where the heat input is abnormal, as shown in fig. 4 a-6 b:
1) Point anomalies
As shown in fig. 4 a-4 b, when a spot abnormality occurs, the welding heat input suddenly rises or falls within a very short period of time. This situation does not require feedback adjustment because of the time period in which it occurs, and the return is quick.
2) Mutant abnormalities
As shown in fig. 5 a-5 b, when a sudden abnormality occurs, the heat input suddenly increases or decreases abnormally at a certain time point and then continuously remains at the same level, and feedback adjustment is required at this time.
3) Gradual change abnormality
As shown in fig. 6 a-6 b, when a gradual abnormality occurs, the heat input suddenly increases or decreases abnormally at a certain time point, and then gradually returns to a normal level at a slower rate. At this time, corresponding feedback adjustment is also required.
In the embodiment, an abnormal classification model is constructed by adopting a machine learning classification algorithm, and the abnormal classification is realized by specifically selecting an XGboost classification model.
Acquiring a plurality of window samples of abnormal heat input through an abnormal recognition model, carrying out abnormal classification based on the classification standard, and extracting time domain characteristics of an abnormal heat input window; after the window samples are subjected to differential processing, statistical characteristics of the window samples are further extracted, so that a characteristic data set of each window is obtained, and the characteristic data set corresponds to a specific abnormal type. A typical machine learning classification algorithm model training flow is shown in figure 7,
according to the method shown in fig. 7, based on the feature data set training XGboost classification model, when a window sample of heat input abnormality is input, what kind of abnormality the heat input abnormality corresponding to the window belongs to can be accurately identified finally. According to different anomaly categories, the present embodiment provides the following feedback adjustment strategies:
when the generated heat input abnormality belongs to the point abnormality, the system does not perform feedback regulation; when the generated heat input abnormality belongs to the abrupt change abnormality, the welding heat input is returned to the heat input level before the abrupt change by controlling the current, the voltage and the welding speed; when the abnormal heat input occurs belongs to the gradual change abnormality, the welding heat input is restored to the level before gradual change in a shorter time by controlling the current, the voltage and the welding speed as well.
And S3, deploying the welding heat input abnormality detection model trained in the step S2 to the edge side, receiving high-frequency time sequence data in the actual welding process, and preprocessing. Inputting the preprocessed welding heat input data set into an anomaly identification model based on indRNN-GAN, and outputting a window sample with anomaly heat input; and then the abnormal classification model receives the window sample with abnormal heat input, outputs a classification result and confirms what kind of abnormality the window sample belongs to. And (3) carrying out corresponding adjustment based on the feedback adjustment strategy provided in the step (S2), and realizing the purposes of abnormal recognition of welding heat input and feedback adjustment.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (5)
1. The intelligent recognition and optimization feedback method for the abnormal welding heat input is characterized by comprising the following steps of:
step S1, firstly, collecting high-frequency time sequence data in the actual welding process, including current, voltage and welding speed; calculating and acquiring a welding heat input set; cutting the welding heat input set to obtain a plurality of window samples;
s2, constructing a welding heat input abnormality detection model; the welding heat input abnormality detection model comprises an indRNN-GAN-based abnormality identification model and an abnormality classification model; an anomaly identification model based on indRNN-GAN is used for generating an countermeasure network structure GAN as a model main body, and comprises two parts of a generation network and a discrimination network; the generating network part adopts an Autoencoder architecture and comprises a decoder and an encoder which are sequentially connected, and an independent circulating network indRNN module is adopted; the discrimination network part adopts an MLP classifier, and the main structure comprises a linear layer, an activation function layer, a linear layer and a sigmoid layer which are connected in sequence; wherein the activation function layer adopts a LeakyRelu function; the specific steps of training the indRNN-GAN-based anomaly identification model are as follows:
s2.1, dividing a model entering data set, selecting a certain proportion from the normal welding heat input window samples obtained in the step S1 as a training set sample, and mixing the residual window samples with the abnormal heat input window samples to obtain a test set sample;
step S2.2, continuously iterating normal window samples based on training set samples, and calculating a mean square error MSE between each generation network generation data set and a real data set; after iterating the complete normal window samples, obtaining a set { MSE }; taking the 3/4+1.5IQR bit number of { MSE } as an anomaly threshold; after the window sample is input, generating a situation that the input window sample has abnormal heat input when the mean square error between the data set and the real window sample data set is more than the 3/4+1.5IQR bit number of { MSE }, otherwise, the situation is the normal heat input situation;
s2.3, after model training is completed, further performing model test; inputting test set data, and reconstructing the data set through a generation network; calculating a mean square error MSE between the reconstructed data set and the input data set; judging based on the mean square error and an abnormal threshold value, determining an abnormal heat input window sample corresponding to the test set, and inputting the abnormal heat input window sample into an abnormal classification model for training a machine learning classification algorithm model;
inputting the window sample into an anomaly identification model based on indRNN-GAN, extracting an anomaly heat input window sample when the anomaly identification model identifies that the window sample has heat input anomalies, and inputting the anomaly heat input window sample into an anomaly classification model; the abnormal classification model adopts a machine learning classification algorithm model; judging the abnormal state of the abnormal heat input window sample according to the difference of the actual abnormal conditions by the abnormal classification model; according to different abnormal states, different feedback regulation strategies are given;
and step S3, deploying the welding heat input abnormality detection model trained in the step S2 to the edge side, receiving high-frequency time sequence data in the actual welding process, judging what abnormal conditions belong to when the welding heat input abnormality exists, and giving out a corresponding feedback regulation strategy.
2. The intelligent recognition and optimization feedback method for abnormal welding heat input according to claim 1, wherein the welding heat input set calculation method in step S1 is as follows:
Heatinput = IU/V
wherein I is welding current, U is welding voltage, and V is welding speed; and according to the actual welding condition, sliding and cutting the welding heat input set obtained through calculation according to the time window size t and the time step size s, and finally obtaining a plurality of window samples under the conditions of normal heat input and abnormal heat input.
3. The intelligent recognition and optimization feedback method for welding heat input anomalies according to claim 1, wherein the independent circulation network indRNN module comprises a plurality of branches, and each branch comprises a weight layer, a BN layer, a Relu activation function layer and a BN layer which are stacked in sequence.
4. The intelligent recognition and optimization feedback method for abnormal welding heat input according to claim 1, wherein the abnormal welding heat input conditions in the step S2 include three conditions of point abnormality, abrupt abnormality and gradual abnormality; the corresponding feedback adjustment strategy includes: when the generated heat input abnormality belongs to the point abnormality, the system does not perform feedback regulation; when the generated heat input abnormality belongs to the abrupt change abnormality, the welding heat input is returned to the heat input level before the abrupt change by controlling the current, the voltage and the welding speed; when the abnormal heat input occurs belongs to the gradual change abnormality, the welding heat input is restored to the level before gradual change in a shorter time by controlling the current, the voltage and the welding speed as well.
5. The intelligent recognition and optimization feedback method for welding heat input abnormality according to claim 1, wherein in the step S3, the preprocessed welding heat input data set is input into an indRNN-GAN-based abnormality recognition model, and a window sample with abnormal heat input is output; then the abnormal classification model receives a window sample with abnormal heat input, outputs a classification result, and confirms what kind of abnormality the window sample belongs to; and finally, carrying out corresponding adjustment based on the feedback adjustment strategy provided in the step S2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310053304.3A CN115781136B (en) | 2023-02-03 | 2023-02-03 | Intelligent recognition and optimization feedback method for welding heat input abnormality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310053304.3A CN115781136B (en) | 2023-02-03 | 2023-02-03 | Intelligent recognition and optimization feedback method for welding heat input abnormality |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115781136A CN115781136A (en) | 2023-03-14 |
| CN115781136B true CN115781136B (en) | 2023-04-25 |
Family
ID=85429614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310053304.3A Active CN115781136B (en) | 2023-02-03 | 2023-02-03 | Intelligent recognition and optimization feedback method for welding heat input abnormality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115781136B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116038076B (en) * | 2023-03-28 | 2023-10-13 | 苏芯物联技术(南京)有限公司 | Welding shielding gas control method based on scene matching |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5096087B2 (en) * | 2007-09-11 | 2012-12-12 | 株式会社神戸製鋼所 | High tensile strength steel plate for high heat input welding with excellent base metal low temperature toughness |
| FR2976804B1 (en) * | 2011-06-23 | 2013-07-12 | Oreal | PROCESS FOR TREATING KERATIN FIBERS WITH HEAT DELIVERY |
| CN103490734B (en) * | 2013-07-26 | 2016-01-20 | 江苏科技大学 | A kind of device of offsetting resistive feedback power current amplifier input stage thermal noise |
| US11033978B2 (en) * | 2018-02-28 | 2021-06-15 | Esab Ab | Arc voltage sensing and control for a welding apparatus |
| CN111141325A (en) * | 2018-11-05 | 2020-05-12 | 苏芯物联技术(南京)有限公司 | Safety monitoring device and monitoring method for fuel generator |
| CN109711465B (en) * | 2018-12-26 | 2022-12-06 | 西安电子科技大学 | Image subtitle generating method based on MLL and ASCA-FR |
| CN112114986B (en) * | 2019-06-20 | 2023-10-13 | 腾讯科技（深圳）有限公司 | Data anomaly identification method, device, server and storage medium |
| CN110480144B (en) * | 2019-08-26 | 2021-07-13 | 攀钢集团攀枝花钢铁研究院有限公司 | Process for controlling flash welding joint structure of medium-limit Mn content R350HT steel rail |
| CN111898514B (en) * | 2020-07-24 | 2022-10-18 | 燕山大学 | Multi-target visual supervision method based on target detection and action recognition |
| KR102387967B1 (en) * | 2020-11-26 | 2022-04-18 | 현대제철 주식회사 | electro gas welding system capable of measuring real time welding heat input |
-
2023
- 2023-02-03 CN CN202310053304.3A patent/CN115781136B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115781136A (en) | 2023-03-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107271925B (en) | Five level converter Fault Locating Method of modularization based on depth convolutional network | |
| CN109800875A (en) | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine | |
| CN106843195B (en) | The Fault Classification differentiated based on adaptive set at semi-supervised Fei Sheer | |
| CN117421684B (en) | Abnormal data monitoring and analyzing method based on data mining and neural network | |
| CN112330165B (en) | Power grid transient stability evaluation method and system based on feature separation type neural network | |
| CN115781136B (en) | Intelligent recognition and optimization feedback method for welding heat input abnormality | |
| CN108958217A (en) | A kind of CAN bus message method for detecting abnormality based on deep learning | |
| CN111862065B (en) | Power transmission line diagnosis method and system based on multitask deep convolutional neural network | |
| CN112000084A (en) | Intelligent BIT design method of controller module based on 1D-CNN and GRU-SVM | |
| CN112926620B (en) | Defect type identification method and system for partial discharge of transformer bushing | |
| Lou et al. | Fault diagnosis of blast furnace iron-making process with a novel deep stationary kernel learning support vector machine approach | |
| CN111985528B (en) | PDGAN-based cable partial discharge data enhancement method | |
| CN118014373B (en) | Risk identification model based on data quality monitoring and construction method thereof | |
| CN112214006A (en) | Intermittent process fault detection method and system considering two-dimensional dynamic characteristics | |
| CN111414943B (en) | Anomaly detection method based on mixed hidden naive Bayes model | |
| CN116431966A (en) | Reactor core temperature anomaly detection method of incremental characteristic decoupling self-encoder | |
| CN115481726A (en) | Industrial robot complete machine health assessment method and system | |
| CN110084301B (en) | Hidden Markov model-based multi-working-condition process working condition identification method | |
| Wang et al. | A SCADA-data-driven condition monitoring method of wind turbine generators | |
| CN112016597B (en) | Depth sampling method based on Bayesian unbalance measurement in machine learning | |
| Wang et al. | Thread quality classification of a tapping machine based on machine learning | |
| CN112163474A (en) | Intelligent gearbox diagnosis method based on model fusion | |
| CN111858343A (en) | Countermeasure sample generation method based on attack capability | |
| CN114565051A (en) | Test method of product classification model based on neuron influence degree | |
| CN114595750A (en) | Aluminum electrolysis fault identification method based on deep reinforcement learning parameter automatic adjustment |
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 |