Self-adaptive welding control method and system, electronic equipment and medium

Abstract

The invention relates to the technical field of oil and gas pipeline welding, and discloses a self-adaptive welding control method and system, electronic equipment and a medium, and the method comprises the following steps: acquiring a first visual image and a first welding parameter; establishing a mapping relation between each first visual image and each first welding parameter; the first visual images and the corresponding first welding parameters serve as a data set to be input into a dense convolutional neural network for training, and a welding parameter prediction model is determined; acquiring a second visual image of the surface of the oil and gas pipeline in real time when the oil and gas pipeline is welded, inputting the second visual image into the welding parameter prediction model, and predicting each second welding parameter; and the welding robot is controlled to conduct pipeline welding according to the second welding parameters. According to the method, the corresponding welding parameters can be predicted in real time from the visual images through the welding parameter prediction model, so that the welding parameters of the welding robot are adjusted in a self-adaptive mode, and the stability of the welding process and the welding seam quality are guaranteed.

Description

Background Art

[0002] Existing oil and gas pipeline welding machines have some basic adjustment parameters, such as welding current, welding speed, arc voltage, etc., and the welding parameters can be directly set according to welding needs to achieve better welding effects. However, in actual use, due to the small number of adjustable motion parameters and process parameters, the real-time adjustment is poor, and therefore it is impossible to achieve synchronous joint adjustment and other problems, which will limit the flexibility and applicability of the equipment. For example, the prior art records an automatic welding control method, which generates the welding parameters required for the corresponding welding according to the weld morphology of the weld layer and the weld bead, reversely predicts the weld morphology according to the obtained welding parameters, and then compares the predicted weld morphology with the planned morphology, and adjusts the welding parameters according to the comparison results. The above-mentioned prior art adjusts the welding parameters of the welding robot by predicting the weld morphology, and cannot adaptively control the welding robot according to the real-time weld morphology.

[0003] For example, the prior art records a molten pool center recognition system and weld tracking method based on HDR images, which achieves weld tracking by obtaining the deviation between the molten pool center and the weld and correcting the position of the K-TIG welding gun in real time according to the deviation. This can be used for automatic tracking of K-TIG welding. However, the above technical solution is not suitable for welding of oil and gas pipelines, and it is also unable to adaptively control welding robots.

[0004] In summary, the existing pipeline welding technology cannot adaptively adjust the welding parameters of the welding robot according to the morphological state of the pipeline during welding. Summary of the invention

[0005] In order to overcome the problem that the existing pipeline welding technology cannot adaptively adjust the welding parameters of the welding robot according to the morphological state of the pipeline during welding, the present invention provides an adaptive welding control method, system, electronic equipment and medium.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides an adaptive welding control method, comprising:

[0007] Acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot;

[0008] Establishing a mapping relationship between each first visual image and each first welding parameter, and determining each first welding parameter corresponding to each first visual image;

[0009] The first visual images and the first welding parameters with established mapping relationships are used as data sets and input into a dense convolutional neural network for training to build a welding parameter prediction model;

[0010] Acquire a second visual image of the surface of the oil and gas pipeline in real time during welding, input the second visual image into a welding parameter prediction model, and predict various second welding parameters of the welding robot corresponding to the second visual image;

[0011] The welding robot is controlled to perform pipeline welding according to the second welding parameters.

[0012] In a second aspect, the present invention provides an adaptive welding control system, comprising:

[0013] A first data acquisition module, used to acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot;

[0014] A mapping relationship establishing module, used to establish a mapping relationship between each first visual image and each first welding parameter, and determine each first welding parameter corresponding to each first visual image;

[0015] A model training module is used to take each first visual image and each first welding parameter with a mapping relationship as a data set, and input them into a dense convolutional neural network for training, so as to train and construct a welding parameter prediction model;

[0016] The second data acquisition module is used to acquire in real time a second visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded, and input the second visual image into the welding parameter prediction model to predict various second welding parameters of the welding robot corresponding to the second visual image;

[0017] The welding robot control module is used to control the welding robot to perform pipeline welding according to various second welding parameters.

[0018] In a third aspect, the present invention provides a computing device, including a memory, a processor, and a program stored in the memory and running on the processor, and when the processor executes the program, the steps of an adaptive welding control method as described above are implemented.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes the steps of an adaptive welding control method as described above.

[0020] The beneficial effects of the present invention are: a dense convolutional neural network is trained by using the first visual image of the oil and gas pipeline during welding and the corresponding first welding parameters to obtain a welding parameter prediction model, and the second visual image of the oil and gas pipeline during welding obtained in real time is input into the welding parameter prediction model to obtain the corresponding second welding parameters, so as to control the welding robot to perform welding in real time through the second welding parameters. The present application can predict the corresponding welding parameters from the visual image in real time through the welding parameter prediction model, so as to adaptively adjust the various welding parameters of the welding robot to ensure the stability of the welding process and the quality of the weld. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0022] Figure 1 A schematic flow chart of an adaptive welding control method according to an embodiment of the present invention;

[0023] Figure 2 Schematic diagram of the structure of the channel attention mechanism module and the space attention mechanism module of the present invention;

[0024] Figure 3 The figure is a schematic diagram of the structure of an adaptive welding control system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following examples are provided to further explain and supplement the present invention and do not constitute any limitation to the present invention.

[0026] An adaptive welding control method, system, electronic device and medium according to an embodiment of the present invention will be described below in conjunction with the accompanying drawings.

[0027] like Figure 1 As shown, the present invention provides an adaptive welding control method, comprising:

[0028] S1. Acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot.

[0029] S2. Establish a mapping relationship between each first visual image and each first welding parameter, and determine each first welding parameter corresponding to each first visual image.

[0030] S3. Taking each first visual image and each first welding parameter with established mapping relationship as a data set, and inputting them into a dense convolutional neural network for training, so as to construct a welding parameter prediction model.

[0031] S4. Acquire in real time a second visual image of the surface of the oil and gas pipeline during welding, input the second visual image into a welding parameter prediction model, and predict various second welding parameters of the welding robot corresponding to the second visual image.

[0032] S5. Controlling the welding robot to perform pipeline welding according to each second welding parameter.

[0033] In this embodiment, a dense convolutional neural network is trained by using the first visual image of the oil and gas pipeline during welding and the corresponding first welding parameters to obtain a welding parameter prediction model, and the second visual image of the oil and gas pipeline during welding obtained in real time is input into the welding parameter prediction model to obtain the corresponding second welding parameters, so as to control the welding robot to perform welding in real time through the second welding parameters. The present application can predict the corresponding welding parameters in real time from the visual image through the welding parameter prediction model, so as to adaptively adjust the various welding parameters of the welding robot to ensure the stability of the welding process and the quality of the weld.

[0034] Optionally, establishing a mapping relationship between each first visual image and each first welding parameter, and determining each first welding parameter corresponding to each first visual image, includes:

[0035] When a frame of the first visual image is collected, each first welding parameter of the welding robot corresponding to the first visual image is obtained, and each first welding parameter is marked in the first visual image as marking information.

[0036] In this embodiment, different first visual images are associated with welding parameters of a specific welding robot, for example:

[0037] The first visual image of the first frame is a weld surface topography image;

[0038] The marking information corresponding to the first visual image of the first frame is: welding current: 180A, welding voltage: 24V, welding speed: 12cm / min, welding gun swing frequency: 3Hz and welding gun posture angle: 20°.

[0039] The first visual image of the Nth frame is a molten pool morphology image;

[0040] The marking information corresponding to the first visual image of the Nth frame is: welding current: 175A, welding voltage: 22V, welding speed: 11.5cm / min, welding gun swing frequency: 2Hz and welding gun posture angle: 10°.

[0041] Optionally, the welding parameter prediction model is a prediction model formed by sequentially adding a channel attention mechanism module and a spatial attention mechanism module between a convolutional layer and a fully connected layer of a dense convolutional neural network;

[0042] Inputting the second visual image into the welding parameter prediction model to predict each second welding parameter of the welding robot corresponding to the second visual image includes:

[0043] Performing multiple convolution operations on the second visual image through a convolution layer to extract local features of the second visual image and determine multiple feature maps;

[0044] According to the global information of each feature map, the target feature map is filtered out from each feature map through the channel attention mechanism module;

[0045] Through the spatial attention mechanism module, the target area is filtered out from the target feature map;

[0046] The features of the target area are extracted through the fully connected layers in the dense convolutional neural network to determine the high-level abstract global features;

[0047] Each second welding parameter is mapped as a high-level abstract global feature via a normalized exponential function in a dense convolutional neural network.

[0048] In this embodiment, a dense convolutional neural network (DenseNet) based on a mixed attention mechanism (mixed attention module) is built, and its original dense convolutional neural network structure consists of an input layer, a convolution layer, a fully connected layer and a softmax layer (normalization function) connected in sequence. In this embodiment, a mixed attention mechanism module is added between the convolution layer and the fully connected layer. The mixed attention mechanism module consists of a channel attention mechanism module (channel attention module, CA module) and a spatial attention mechanism module (spatial attention module, SA module), which is used to capture important channel and spatial features of weld and molten pool morphology. The Softmax layer is used to predict welding parameters.

[0049] The core feature of the dense convolutional neural network based on the hybrid attention mechanism used in this embodiment is that the output of each layer is connected to all subsequent layers, thereby achieving better feature reuse and gradient propagation. This structure enables the model to capture local and global information in visual images.

[0050] The role of the hybrid attention mechanism (including the spatial attention mechanism and the channel attention mechanism) is to enhance the network's ability to express features, so that the model can focus more accurately on important information in the welding process, thereby improving the accuracy of welding process control.

[0051] In this embodiment, the data needs to be preprocessed before the second visual image is input into the welding parameter prediction model, including steps such as denoising and normalizing the collected second visual image to ensure that the data format and range meet the input requirements of the dense convolutional neural network. At the same time, such preprocessing also helps to improve the generalization ability of the neural network.

[0052] Optionally, according to the global information of each feature map, a channel attention mechanism module is used to filter out a target feature map from each feature map, including:

[0053] Each feature map is input into different channels of the channel attention mechanism module;

[0054] Calculating a first weight value according to the global information of each feature map and the preset weight of the corresponding channel; wherein the first weight value represents the importance of the global information in the feature map;

[0055] The feature map corresponding to the maximum value among the first weight values ​​is used as the target feature map.

[0056] In this embodiment, the channel attention mechanism module assigns weights to each channel, highlights the key feature channels, suppresses unimportant or redundant feature channels, and then decides which feature map channels are more important based on the global information of each feature map (such as weld morphology, molten pool state, etc.). In this way, the feature maps corresponding to the most important global information can be screened out from each channel.

[0057] like Figure 2 As shown in the figure, each channel in the channel attention mechanism module (CA module) includes two branches, where:

[0058] The first branch includes Maxpool (maximum pooling layer), FC (Fully Connected Layer), ReLU (Rectified Linear Unit activation function) and FC (Fully Connected Layer), which are mainly used to reduce the spatial dimension of the feature map by selecting the maximum value in the pooling window to obtain global information, and then identify and classify the global features through two fully connected layers. After identification and classification, the first weight value is calculated through the activation function.

[0059] The second branch includes Avgpool (average pooling layer), FC (Fully Connected Layer), ReLU (Rectified Linear Unit activation function) and FC (Fully Connected Layer), which are mainly used to reduce the spatial dimension of the feature map by averaging all values ​​in the pooling window to obtain global information, and then identify and classify the global features through two fully connected layers. After identification and classification, the first weight value is calculated through the activation function.

[0060] After the first weight values ​​calculated by the first branch and the second branch are merged, the calculation result is mapped to a range between 0 and 1 through the Sigmoid function to obtain the final first weight value.

[0061] Optionally, the target area is filtered out from the target feature map through a spatial attention mechanism module, including:

[0062] Calculate the second weight value of each pixel in the target feature map through the spatial attention mechanism; wherein the second weight value represents the importance of the spatial position in the target feature map;

[0063] Taking pixels whose second weight value is greater than or equal to a preset value as target pixels;

[0064] The area corresponding to each target pixel is taken as the target area.

[0065] In this embodiment, the spatial attention mechanism module assigns weights to each spatial position (pixel) in the feature map, highlights the features of key positions, and suppresses the features of irrelevant or background areas. The spatial attention mechanism module focuses on extracting important areas in the image, such as weld edges, weld start and end points, molten pool boundaries, etc. It ensures that the model can focus on these key spatial positions during the welding process instead of being disturbed by irrelevant background information.

[0066] like Figure 2 As shown in the figure, the spatial attention mechanism module (SA module) includes Maxpool (maximum pooling layer), Avgpool (average pooling layer), Concatenate layer (neural network layer), Conv (convolutional layer) and Sigmoid function.

[0067] The spatial dimension of the feature map is reduced by taking the average value of all values ​​in the pooling window, and then by taking the maximum value of all values ​​in the pooling window. The feature maps reduced by Maxpool and Avgpool are fused through the Concatenate layer, and then the features of each spatial position are extracted through Conv and the second weight value is calculated. Finally, the second weight value is mapped to between 0 and 1 through the Sigmoid function.

[0068] Optionally, the first visual image and the second visual image include a weld surface morphology image and a molten pool morphology image, and the first welding parameter and the second welding parameter include welding current, welding voltage, welding speed, welding gun swing frequency and welding gun posture angle.

[0069] In this embodiment, the weld surface topography image can be obtained by a line profile scanner, the molten pool topography data can be obtained by a molten pool camera, and the first welding parameter can be collected by a multi-source sensor.

[0070] Optionally, controlling the welding robot to perform pipeline welding according to each second welding parameter includes:

[0071] The welding current, welding voltage and welding speed are input into the welding power supply control system of the welding robot, and the welding gun swing frequency and welding gun posture angle are input into the motion control system of the welding robot.

[0072] In this embodiment, the welding parameters of the welding robot can be adaptively and synchronously adjusted to achieve intelligent decision-making for welding control, thereby assisting in improving welding quality and production efficiency.

[0073] In addition, based on the adaptive welding control method of deep learning, it is extracted as the numerical process or motion parameters that the robot needs to execute, and a dense convolutional neural network for adaptive welding is constructed. The welding parameters are adaptively adjusted according to the closed-loop feedback parameters to ensure the stability and consistency of the welding quality. At the same time, through the adaptive adjustment of welding parameters, it can be used to automatically compensate for the surface quality defects of the upper layer, thereby effectively ensuring the welding qualification rate and weld quality.

[0074] like Figure 3 As shown, the present invention provides an adaptive welding control system, comprising:

[0075] A first data acquisition module, used to acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot;

[0076] A mapping relationship establishing module, used to establish a mapping relationship between each first visual image and each first welding parameter, and determine each first welding parameter corresponding to each first visual image;

[0077] A model training module is used to take each first visual image and each first welding parameter with a mapping relationship as a data set, and input them into a dense convolutional neural network for training, so as to train and construct a welding parameter prediction model;

[0078] The second data acquisition module is used to acquire in real time a second visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded, and input the second visual image into the welding parameter prediction model to predict various second welding parameters of the welding robot corresponding to the second visual image;

[0079] The welding robot control module is used to control the welding robot to perform pipeline welding according to various second welding parameters.

[0080] Optionally, the mapping relationship establishment module is specifically used to:

[0081] When a frame of the first visual image is collected, each first welding parameter of the welding robot corresponding to the first visual image is obtained, and each first welding parameter is marked in the first visual image as marking information.

[0082] Optionally, the dense convolutional neural network includes a convolutional layer, a channel attention mechanism module, a spatial attention mechanism module and a Softmax layer;

[0083] The second data acquisition module is specifically used for:

[0084] Performing multiple convolution operations on the second visual image through a convolution layer to extract local features of the second visual image and determine multiple feature maps;

[0085] According to the global information of each feature map, the target feature map is filtered out from each feature map through the channel attention mechanism module;

[0086] Through the spatial attention mechanism module, the target area is filtered out from the target feature map;

[0087] The features of the target area are extracted through the fully connected layers in the dense convolutional neural network to determine the high-level abstract global features;

[0088] Each second welding parameter is mapped as a high-level abstract global feature via a normalized exponential function in a dense convolutional neural network.

[0089] Optionally, the second data acquisition module is specifically used to:

[0090] Each feature map is input into different channels of the channel attention mechanism module;

[0091] Calculating a first weight value according to the global information of each feature map and the preset weight of the corresponding channel; wherein the first weight value represents the importance of the global information in the feature map;

[0092] The feature map corresponding to the maximum value among the first weight values ​​is used as the target feature map.

[0093] Optionally, the second data acquisition module is specifically used to:

[0094] Calculate the second weight value of each pixel in the target feature map through the spatial attention mechanism; wherein the second weight value represents the importance of the spatial position in the target feature map;

[0095] Taking pixels whose second weight value is greater than or equal to a preset value as target pixels;

[0096] The area corresponding to each target pixel is taken as the target area.

[0097] Optionally, the welding robot control module is specifically used for:

[0098] The welding current, welding voltage and welding speed are input into the welding power supply control system of the welding robot, and the welding gun swing frequency and welding gun posture angle are input into the motion control system of the welding robot.

[0099] The embodiment of the present invention also provides a computing device, including a memory, a manager, and a program stored in the memory and running on the manager. When the manager executes the program, some or all steps of the above-mentioned adaptive welding control method are implemented.

[0100] Among them, the computing device can be a computer, and correspondingly, its program is computer software. The above-mentioned parameters and steps in the computing device of the present invention can refer to the parameters and steps in the embodiment of an adaptive welding control method mentioned above, and will not be repeated here.

[0101] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code. Computer-readable storage media can be, for example, but not limited to - electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above.

[0102] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0103] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. An adaptive welding control method, characterized in that: include: Acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot; Establishing a mapping relationship between each of the first visual images and each of the first welding parameters, and determining each of the first welding parameters corresponding to each of the first visual images; Using the first visual images and the first welding parameters with established mapping relationships as data sets, training and constructing a welding parameter prediction model; Acquire a second visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded in real time, input the second visual image into the welding parameter prediction model, and predict each second welding parameter of the welding robot corresponding to the second visual image; The welding robot is controlled to perform pipeline welding according to each of the second welding parameters.

2. The method according to claim 1, characterized in that The establishing of a mapping relationship between each of the first visual images and each of the first welding parameters, and determining each of the first welding parameters corresponding to each of the first visual images, includes: When a frame of the first visual image is collected, each first welding parameter of the welding robot corresponding to the first visual image is obtained, and each first welding parameter is marked in the first visual image as marking information.

3. The method according to claim 1, characterized in that The welding parameter prediction model is a prediction model formed by sequentially adding a channel attention mechanism module and a space attention mechanism module between the convolution layer and the fully connected layer of the dense convolutional neural network; The step of inputting the second visual image into the welding parameter prediction model to predict each second welding parameter of the welding robot corresponding to the second visual image includes: Performing a plurality of convolution operations on the second visual image through the convolution layer to extract local features of the second visual image and determine a plurality of feature maps; According to the global information of each of the feature maps, a target feature map is filtered out from each feature map through the channel attention mechanism module; Filtering a target area from the target feature map through the spatial attention mechanism module; Extracting features of the target area through the fully connected layer to determine high-level abstract global features; Each second welding parameter is mapped to the high-level abstract global feature through a normalized exponential function in the dense convolutional neural network.

4. The method according to claim 3, characterized in that: The method of filtering out a target feature map from each feature map by using the channel attention mechanism module according to the global information of each feature map includes: Input each of the feature maps into a different channel of the channel attention mechanism module; Calculate a first weight value according to the global information of each feature map and the preset weight of the corresponding channel; wherein the first weight value represents the importance of the global information in the feature map; The feature map corresponding to the maximum value of each of the first weight values ​​is used as the target feature map.

5. The method according to claim 3, characterized in that: The method of filtering out a target area from the target feature map by using a spatial attention mechanism module includes: Calculating a second weight value of each pixel in the target feature map by a spatial attention mechanism; wherein the second weight value represents the importance of the spatial position in the target feature map; Taking each pixel whose second weight value is greater than or equal to a preset value as a target pixel; The area corresponding to each of the target pixels is taken as the target area.

6. The method according to any one of claims 1 to 5, characterized in that: The first visual image and the second visual image include a weld surface morphology image and a molten pool morphology image, and the first welding parameter and the second welding parameter include welding current, welding voltage, welding speed, welding gun swing frequency and welding gun posture angle.

7. The method according to claim 6, characterized in that The controlling the welding robot to perform pipeline welding according to each of the second welding parameters comprises: The welding current, the welding voltage and the welding speed are input into a welding power supply control system of a welding robot, and the welding gun swing frequency and the welding gun posture angle are input into a motion control system of the welding robot.

8. An adaptive welding control system, characterized in that: include: A first data acquisition module, used to acquire a first visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded and a first welding parameter of the welding robot; A mapping relationship establishing module, used to establish a mapping relationship between each of the first visual images and each of the first welding parameters, and determine each of the first welding parameters corresponding to each of the first visual images; A model training module, used to take each of the first visual images and each of the first welding parameters with established mapping relationships as a data set, and input them into a dense convolutional neural network for training to determine a welding parameter prediction model; A second data acquisition module is used to acquire in real time a second visual image of the surface of the oil and gas pipeline when the oil and gas pipeline is welded, and input the second visual image into the welding parameter prediction model to predict various second welding parameters of the welding robot corresponding to the second visual image; The welding robot control module is used to control the welding robot to perform pipeline welding according to each of the second welding parameters.

9. A computing device comprising a memory, a processor, and a program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the steps of an adaptive welding control method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on the terminal device, the terminal device executes the steps of an adaptive welding control method as described in any one of claims 1 to 7.

Images