CN114722883B - Welding quality real-time detection method and system based on high-frequency time sequence data - Google Patents
Welding quality real-time detection method and system based on high-frequency time sequence data Download PDFInfo
- Publication number
- CN114722883B CN114722883B CN202210633173.1A CN202210633173A CN114722883B CN 114722883 B CN114722883 B CN 114722883B CN 202210633173 A CN202210633173 A CN 202210633173A CN 114722883 B CN114722883 B CN 114722883B
- Authority
- CN
- China
- Prior art keywords
- data
- welding
- real
- time sequence
- sequence data
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
- B23K31/00—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups
- B23K31/12—Processes relevant to this subclass, specially adapted for particular articles or purposes, but not covered by only one of the preceding main groups relating to investigating the properties, e.g. the weldability, of materials
- B23K31/125—Weld quality monitoring
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C3/00—Registering or indicating the condition or the working of machines or other apparatus, other than vehicles
- G07C3/14—Quality control systems
- G07C3/146—Quality control systems during manufacturing process
-
- 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
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Manufacturing & Machinery (AREA)
- Automation & Control Theory (AREA)
- Mechanical Engineering (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a welding quality real-time detection method and a system based on high-frequency time sequence data, wherein the detection method comprises the following steps: firstly, collecting high-frequency time sequence data of normal welding, and constructing a data sample set by sliding a window according to a set window length; then building a fusion network model of Unet and TCN, and performing model training by using a data sample set; and finally, collecting new high-frequency welding time sequence data, inputting the new high-frequency welding time sequence data into the trained model for prediction according to the set window length, further calculating the reconstruction error of the real data and the model prediction data, and comparing the reconstruction error with a set threshold value to realize the real-time detection of welding abnormity. The method uses a new Unet and TCN fusion network model to carry out feature mapping and recovery reconstruction on the acquired high-frequency time sequence data, then calculates the reconstruction error of the real data and the reconstructed data, identifies the welding quality defect according to the error threshold, and has high accuracy and strong real-time performance and more practical value.
Description
Technical Field
The invention relates to a welding quality real-time detection method and system based on high-frequency time sequence data, and belongs to the technical field of automatic welding.
Background
In recent years, with the rapid development of industries such as automobiles, aerospace, construction, transportation and the like, the process and quality requirements on industrial equipment are higher and higher, and particularly, welding quality detection technology is widely applied to a plurality of fields. The welding quality can be classified into a direct welding quality and an indirect welding quality. The general welding joint has the main use performances of mechanical performance, internal and external defects, the geometric size of a welded product and the like, which is the direct welding quality. Indirect weld quality is a factor that can be detected by a sensor of the sense or nature of the welder during the welding process and that indirectly determines the quality of the direct weld. Although the indirect welding quality cannot directly indicate the service performance of the welding joint, the indirect welding quality can reflect whether the welding quality problem occurs in the welding process to a great extent.
At present, the deep learning is combined with data such as visual images, arc spectrums, arc sounds and the like to carry out welding quality diagnosis, which is a main technical means for carrying out welding quality diagnosis, but in an actual use scene, the data such as the welding images, the spectrums, the sounds and the like are difficult to collect, are greatly influenced by the environment, the defect types are difficult to define, a large amount of data marking is needed, the time and the economic cost are high, and the effect is poor. And the time series data such as the electric signal in the welding process not only contain power performance information, but also contain a large amount of welding quality information, and the acquisition cost of the time series data is low, and the time series data is not easily influenced by external environment factors, so that the corresponding relation between the time series data and the welding quality defect is analyzed and excavated through the time series data, and the welding quality detection cost and the effectiveness can be greatly reduced.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problems in the prior art, the invention provides a welding quality real-time detection method and a welding quality real-time detection system based on high-frequency time sequence data.
The technical scheme is as follows: in order to achieve the aim, the invention provides a welding quality real-time detection method based on high-frequency time sequence data, which comprises the following steps:
step 1: collecting high-frequency time sequence data (including but not limited to current, voltage, air flow speed and wire feeding speed) of normal welding, and carrying out sliding window construction samples according to a set window length to generate a data sample set;
step 2: building a fusion network model of Unet and TCN, and performing model training by using a data sample set;
and 3, step 3: and collecting new high-frequency welding time sequence data, inputting the new high-frequency welding time sequence data into the trained model for prediction according to the set window length, further calculating the reconstruction error of the real data and the model prediction data, and comparing the reconstruction error with a set threshold value to realize the real-time detection of welding abnormity.
Furthermore, the fusion network model comprises an encoder part and a decoder part, wherein the encoder part comprises a convolution network block and at least three TCN network blocks which are sequentially cascaded through downsampling, the decoder part comprises a plurality of convolution network blocks which are sequentially cascaded through upsampling, and the number of the network blocks of the encoder part is consistent with that of the network blocks of the decoder part;
the last TCN network block of the encoder portion is cascaded with the first convolutional network block of the decoder portion through upsampling, the last second TCN network block of the encoder portion is cascaded with the first convolutional network block of the decoder portion through a channel, the last third TCN network block of the encoder portion is cascaded with the second convolutional network block of the decoder portion through a channel, and so on until the first convolutional network block of the encoder portion is cascaded with the last second convolutional network block of the decoder portion through a channel.
Based on the characteristics of high-frequency time sequence data, an encoder firstly reduces the dimension of an input time sequence through a convolution network block, shortens the sequence length and enhances the real-time property of model prediction, and then adopts a plurality of cascaded TCN network blocks to enhance the learning capability of the model to time sequence characteristics.
Further, the TCN network block includes two one-dimensional expansion convolutional layers, two BN (batch normalization) layers, two activation function layers, and a jump connection, and the convolutional network block includes one-dimensional convolutional network layer, one BN layer, and one activation function layer.
Further, the reconstruction error is the MAE (mean absolute error) of the real data and the model prediction data, and the reconstruction error has better robustness to the abnormal point.
Further, a small sample testing process is further included between step 2 and step 3: collecting high-frequency time sequence data of normal welding and high-frequency time sequence data of abnormal welding (including defects of air holes, burn-through, welding leakage and the like), inputting the data into a trained model according to a set window length for prediction, and calculating a reconstruction error between real data and model prediction data so as to determine an abnormal threshold of the reconstruction error.
In addition, the invention also provides a welding quality real-time detection system based on the high-frequency time sequence data, which comprises a data acquisition module and a data processing module, wherein the data acquisition module comprises but is not limited to a current sensor, a pressure sensor, a speed sensor, a wire feeding sensor and the like, and the data processing module carries out real-time detection on welding abnormity according to the high-frequency welding time sequence data acquired by the data acquisition module by using the welding quality real-time detection method.
Has the advantages that: the method converts the welding quality detection problem into abnormal detection of the time sequence signal in the welding process, and compared with the method that data such as images, spectrums, sounds and the like are difficult to collect, the time sequence data is easy to collect and deploy, and the feasibility is high. In addition, the method uses an unsupervised mode to carry out modeling, and training data only use normal welding time sequence data, so that data marking is not needed, and the problem that defect data are difficult to mark is solved.
The invention provides a novel Unet and TCN fusion network structure which is divided into an encoder part and a decoder part, wherein the encoder part adopts a TCN network structure. Compared with a common self-encoder, the network structure of the Unet is used, the deep layer features can be fused, the deep layer features can represent more abstract and essential global features, the shallow layer features pay more attention to local features, and the network can learn more feature information and has higher accuracy; secondly, based on the characteristics of high-frequency time sequence data, TCN is fused in Unet, and the learning of time sequence characteristics is enhanced; in addition, the whole network structure of the method uses a convolutional neural network, and when model reasoning is performed, a prediction result can be given in real time only through parallel matrix operation aiming at new real-time high-frequency welding time sequence data, so that the method has strong real-time performance and higher practical value.
Drawings
FIG. 1 is a flow chart of a method for real-time detection of weld quality in an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a network model for fusing a Unet and a TCN according to an embodiment of the present invention;
FIG. 3 is a block diagram of a TCN network block according to an embodiment of the present invention;
FIG. 4 is a comparison graph of the actual value and the predicted value of the normal welding current in the embodiment of the invention;
FIG. 5 is a comparison graph of the actual value and the predicted value of the void defect current in the embodiment of the present invention;
FIG. 6 is a confusion matrix of the predicted results of small sample tests in an embodiment of the present invention.
Detailed Description
The following description of the preferred embodiments of the present invention with reference to the accompanying drawings will more clearly and completely illustrate the technical solutions of the present invention.
Fig. 1 shows a real-time welding quality detection method based on high-frequency time series data, which comprises the following steps:
step 1: preprocessing data;
step 1.1: high frequency current data during normal welding was collected and samples were constructed for sliding windows according to window _ size, in the example window _ size = 20000.
Step 1.2: randomly shuffle all the generated samples, set the proportion train _ size of the training set samples, cut the training set train _ data according to the proportion, and leave the verification set dev _ data, wherein train _ size = 0.7.
And 2, step: building a network;
the network structure is divided into an encoder and a decoder, and the network structure comprises the following parts:
the encoder section includes:
i, a convolutional network Block1 (comprising a one-dimensional convolutional network layer, stride =2, the output dimension is channel1, a BN layer and an activation function layer), and the output is eat 1;
II, maximally pooling a down-sampling layer;
III, TCN network block TCNBlock1 (output dimension is channel2, expansion coefficient is d 1), and output is feat 2;
IV, maximally pooling the down-sampling layer;
v, TCN network block TCNBlock2 (output dimension is channel3, expansion coefficient is d 2), and output is feat 3;
VI, maximally pooling down-sampling layers;
VII, a TCN network block TCNBlock3 (the output dimension is channel4, the expansion coefficient is d 3), and the output is feat 4;
VIII, a maximum pooling downsampling layer;
IX, TCN network block TCNBlock4 (output dimension is channel4, expansion coefficient is d 4), output is eat 5;
wherein channels 1, 2, 3 and 4 are progressively larger and equal to 4, 8, 12 and 16 in the examples; d1, d2, d3 and d4 are increased by multiples of 2, and are equal to 1, 2, 4 and 8 in the embodiment respectively.
The decoder portion includes:
1. the flat 5 passes through an upsampling layer and then is spliced with the flat 4 on a channel (channel dimension);
2. a convolutional network Block2 (comprising a one-dimensional convolutional network layer, the output dimension is channel4, a BN layer, and an activation function layer), and the output is P4;
3. p4 passes through an up-sampling layer and then is spliced with flat 3 on the channel;
4. a convolutional network Block3 (comprising a one-dimensional convolutional network layer, the output dimensions of which are channel3, a BN layer and an activation function layer), and the output of which is P3;
5. p3 passes through an upsampling layer and then is spliced with feat2 in the channel dimension;
6. a convolutional network Block4 (comprising a one-dimensional convolutional network layer, the output dimensions of which are channel2, a BN layer and an activation function layer), and the output of which is P2;
7. p2 passes through an upsampling layer and then is spliced with feat1 in the channel dimension;
8. a convolutional network Block5 (comprising a one-dimensional convolutional network layer, the output dimension is channel1, a BN layer, and an activation function layer), and the output is P1;
9. p1 passes through an upsampling layer;
10. convolutional network Block6 (comprising one-dimensional convolutional network layer with output dimension 1, one BN layer, one activation function layer).
As shown in fig. 3, each TCN network block contains (input and output are connected via one-dimensional convolutional layer hopping):
a. a one-dimensional expansion convolution layer is formed,
b. batch Normalization (BN, Batch standardization) layer,
c. a leakage ReLU (activation function) layer,
d. a one-dimensional expansion convolution layer is formed,
e. a layer of Batch Normalization,
f. a leak ReLU layer.
And 3, step 3: network training;
setting learning _ rate = α, batch _ size (number of samples captured in one training) = β, and performing model training using training set train _ data and verification set dev _ data, where α =0.001 and β =16 in the embodiment.
And 4, step 4: testing a small sample;
sampling high-frequency time sequence data of normal welding and high-frequency time sequence data of abnormal welding (including air holes, burn-through, welding leakage and the like), carrying out sliding window construction test data sets according to window _ size, carrying out small sample model test, calculating reconstruction errors MAE of real data and model prediction data for each test set sample, and setting an abnormal threshold th (namely finding a critical value between the reconstruction errors of the normal welding data and the abnormal welding data) according to analysis of a calculation result, wherein th =0.02 in the embodiment.
Fig. 4 shows a comparison between the actual value and the predicted value of a part of normal welding current in a small sample test, and fig. 5 shows a comparison between the actual value and the predicted value of a part of air hole defect current in a small sample test. As shown in fig. 6, for the sample of the test set, a confusion matrix is constructed by using the predicted result (determined by the anomaly threshold) and the true result, and the precision ratio of the detection method is about 95.09% according to the confusion matrix.
And 5: detecting the welding quality in real time;
collecting new high-frequency welding time sequence data in real time, performing sliding window according to window _ size, inputting data with the length of each window into a model for prediction, calculating a reconstruction error MAE of real data and model prediction data, comparing the reconstruction error MAE with a set threshold th, and if the reconstruction error MAE is larger than the threshold th, indicating that the welding quality is abnormal.
In addition, the invention also provides a welding quality real-time detection system based on the high-frequency time sequence data, which comprises a data acquisition module and a data processing module, wherein the data acquisition module adopts a high-precision current sensor, and the data processing module utilizes the welding quality real-time detection method to perform real-time detection of welding abnormity according to the acquired high-frequency current data.
The method comprises the steps of mapping high-frequency time sequence data to hidden vectors through an encoder part based on the collected high-frequency welding time sequence data, recovering the high-frequency time sequence data through decoder reconstruction, calculating reconstruction errors of real data and the reconstructed data, and identifying welding quality defects according to a reconstruction error threshold value.
The invention uses the network architecture of the Unet, can fuse the depth layer characteristics, so that the network can learn more characteristic information, and has higher accuracy; the encoder part adopts a TCN network structure, so that the learning capability of the model on the time sequence characteristics is greatly enhanced; in addition, the whole network structure of the method uses a convolutional neural network, and when a model is used for reasoning, a prediction result can be given in real time only through parallel matrix operation aiming at new high-frequency time sequence data, so that the method has strong real-time performance and higher practical value.
The above detailed description merely describes preferred embodiments of the present invention and does not limit the scope of the invention. Without departing from the spirit and scope of the present invention, it should be understood that various changes, substitutions and alterations can be made herein by those skilled in the art from the following detailed description and drawings.
Claims (5)
1. A welding quality real-time detection method based on high-frequency time sequence data is characterized by comprising the following steps:
step 1: collecting high-frequency time sequence data of normal welding, and carrying out sliding window construction samples according to a set window length to generate a data sample set;
and 2, step: building a fusion network model of the Unet and the TCN, and performing model training by using a data sample set;
and step 3: collecting new high-frequency welding time sequence data, inputting the new high-frequency welding time sequence data into a trained model according to a set window length for prediction, further calculating a reconstruction error between real data and model prediction data, and comparing the reconstruction error with a set threshold value to realize real-time detection of welding abnormity;
the converged network model comprises an encoder part and a decoder part, wherein the encoder part comprises a convolutional network block and at least three TCN network blocks which are sequentially cascaded through downsampling, the decoder part comprises a plurality of convolutional network blocks which are sequentially cascaded through upsampling, and the number of the network blocks of the encoder part is consistent with that of the network blocks of the decoder part;
the last TCN network block of the encoder portion is cascaded with the first convolutional network block of the decoder portion through upsampling, the last second TCN network block of the encoder portion is cascaded with the first convolutional network block of the decoder portion through a channel, the last third TCN network block of the encoder portion is cascaded with the second convolutional network block of the decoder portion through a channel, and so on until the first convolutional network block of the encoder portion is cascaded with the last second convolutional network block of the decoder portion through a channel.
2. The method of claim 1, wherein the TCN network block comprises two one-dimensional expansion convolution layers, two BN layers, two activation function layers and a jump connection, and the convolution network block comprises one-dimensional convolution network layer, one BN layer and one activation function layer.
3. The method of claim 1, wherein the reconstruction error is MAE of real data and model predicted data.
4. The method for detecting the welding quality in real time according to the claim 1, characterized in that a small sample testing process is further included between the step 2 and the step 3: and collecting high-frequency time sequence data of normal welding and high-frequency time sequence data of abnormal welding, inputting the data into a trained model according to a set window length for prediction, and calculating a reconstruction error between real data and model prediction data so as to determine an abnormal threshold of the reconstruction error.
5. A real-time welding quality detection system based on high-frequency time sequence data is characterized by comprising a data acquisition module and a data processing module, wherein the data processing module is used for carrying out real-time detection on welding abnormity according to the high-frequency welding time sequence data acquired by the data acquisition module by using the detection method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210633173.1A CN114722883B (en) | 2022-06-07 | 2022-06-07 | Welding quality real-time detection method and system based on high-frequency time sequence data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210633173.1A CN114722883B (en) | 2022-06-07 | 2022-06-07 | Welding quality real-time detection method and system based on high-frequency time sequence data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114722883A CN114722883A (en) | 2022-07-08 |
| CN114722883B true CN114722883B (en) | 2022-08-19 |
Family
ID=82232416
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210633173.1A Active CN114722883B (en) | 2022-06-07 | 2022-06-07 | Welding quality real-time detection method and system based on high-frequency time sequence data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114722883B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115106615B (en) * | 2022-08-30 | 2022-12-02 | 苏芯物联技术(南京)有限公司 | Welding deviation real-time detection method and system based on intelligent working condition identification |
| CN115255567B (en) * | 2022-09-26 | 2022-12-13 | 苏芯物联技术(南京)有限公司 | Welding deviation real-time detection method and system based on frequency domain and time-frequency domain characteristics |
| CN115255566B (en) * | 2022-09-26 | 2022-12-16 | 苏芯物联技术(南京)有限公司 | Welding deviation real-time intelligent detection method based on high-quality time domain characteristics |
| CN115647687B (en) * | 2022-12-29 | 2023-03-17 | 苏芯物联技术(南京)有限公司 | Welding abnormity real-time detection method integrating residual error network and self-attention mechanism |
| CN116010794B (en) * | 2023-02-17 | 2023-05-30 | 苏芯物联技术(南京)有限公司 | Welding defect real-time intelligent detection method based on small sample tag data |
| CN116204830B (en) * | 2023-04-28 | 2023-07-11 | 苏芯物联技术(南京)有限公司 | Welding abnormality real-time detection method based on path aggregation network |
| CN116630319B (en) * | 2023-07-24 | 2023-10-10 | 深圳信迈科技有限公司 | Machine learning-based data acquisition board card welding defect detection method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902832A (en) * | 2018-11-28 | 2019-06-18 | 华为技术有限公司 | Training method, predicting abnormality method and the relevant apparatus of machine learning model |
| CN113743188A (en) * | 2021-06-28 | 2021-12-03 | 杭州电子科技大学 | Internet video low-custom behavior detection method based on feature fusion |
| CN113870260A (en) * | 2021-12-02 | 2021-12-31 | 苏芯物联技术(南京)有限公司 | Welding defect real-time detection method and system based on high-frequency time sequence data |
| CN114548701A (en) * | 2022-01-28 | 2022-05-27 | 浙江大学 | Process early warning method and system for analyzing and estimating coupling structure of full-scale measuring point |
-
2022
- 2022-06-07 CN CN202210633173.1A patent/CN114722883B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902832A (en) * | 2018-11-28 | 2019-06-18 | 华为技术有限公司 | Training method, predicting abnormality method and the relevant apparatus of machine learning model |
| CN113743188A (en) * | 2021-06-28 | 2021-12-03 | 杭州电子科技大学 | Internet video low-custom behavior detection method based on feature fusion |
| CN113870260A (en) * | 2021-12-02 | 2021-12-31 | 苏芯物联技术(南京)有限公司 | Welding defect real-time detection method and system based on high-frequency time sequence data |
| CN114548701A (en) * | 2022-01-28 | 2022-05-27 | 浙江大学 | Process early warning method and system for analyzing and estimating coupling structure of full-scale measuring point |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114722883A (en) | 2022-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114722883B (en) | Welding quality real-time detection method and system based on high-frequency time sequence data | |
| CN113255733B (en) | Unsupervised anomaly detection method under multi-modal data loss | |
| CN113870260B (en) | Welding defect real-time detection method and system based on high-frequency time sequence data | |
| CN112461537B (en) | Wind power gear box state monitoring method based on long-time and short-time neural network and automatic coding machine | |
| CN111914883A (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN115824519B (en) | Comprehensive diagnosis method for valve leakage faults based on multi-sensor information fusion | |
| JP2021190716A (en) | Detection of failure in semiconductor specimen using weak labeling | |
| CN109871002B (en) | Concurrent abnormal state identification and positioning system based on tensor label learning | |
| CN114799610B (en) | Welding quality real-time detection method and system based on inverse Fourier transform and self-encoder | |
| CN111881970A (en) | Intelligent outer broken image identification method based on deep learning | |
| CN117076935B (en) | Digital twin-assisted mechanical fault data lightweight generation method and system | |
| CN118011994A (en) | Cross-modal industrial large model-based fault diagnosis system and method | |
| CN112147221B (en) | Steel rail screw hole crack identification method and system based on ultrasonic flaw detector data | |
| CN117520924B (en) | Island photovoltaic operation and maintenance fault cause analysis method and system based on multi-mode data | |
| CN117541534A (en) | Power transmission line inspection method based on unmanned plane and CNN-BiLSTM model | |
| CN116861369A (en) | Air compressor health assessment method and system for industrial internet heterogeneous data fusion | |
| CN117168861A (en) | Abnormality monitoring method and abnormality monitoring system for sewage treatment equipment | |
| CN101576445B (en) | Data reappearing method for structure health monitoring failure sensor simulating memory of people | |
| CN111753876B (en) | Product quality detection method based on deep neural network | |
| CN115081741A (en) | Natural gas metrological verification intelligent prediction method based on neural network | |
| Yu et al. | Deep learning (dl) based on-board indirect useful life prediction | |
| CN117984024B (en) | Welding data management method and system based on automatic production of ship lock lambdoidal doors | |
| Jency et al. | Enhancing Structural Health Monitoring: AI-Driven Image Processing for Automated Crack Identification in Concrete Surfaces | |
| CN116718382A (en) | Bearing early fault online detection method based on contrast learning | |
| CN118821056A (en) | Semi-supervised anomaly detection method for tunnel cavity and uncompacted defect |
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 |