Method for identifying welding operation risk in steel structure welding operation and related equipment

Abstract

The invention discloses a method for identifying welding operation risks in steel structure welding operation and related equipment, and relates to the technical field of steel structure welding construction. On the basis of a pre-trained welding risk identification model, risk identification is conducted on welding data characteristics of welding data in the welding operation process at the current moment, a risk identification result at the current moment is obtained, and the welding data at least comprise operation data of welding operation and image data of a welding area. The operation data is data, collected by a sensor installed in the welding area, of the welding equipment during welding. Whether risks exist in the welding process or not is recognized from the two aspects of the welding operation process and the welding operation result, the operation data of the welding operation is the data collected by the sensor and is more accurate and comprehensive, and therefore the welding risk recognition model is based on the more accurate and comprehensive welding data, and the welding risk recognition efficiency is improved. A more accurate risk identification result can be obtained through identification.

Description

Technical Field

[0001] The present application relates to the technical field of steel structure welding construction, and in particular to a method for identifying welding operation risks in steel structure welding operations and related equipment. Background Art

[0002] During the welding construction process of steel structures, there are many risk factors, such as welding defects (pores, cracks, incomplete penetration, etc.), welding deformation, construction environment factors (temperature, humidity, ventilation, etc.) and human operational errors. These risk factors will seriously affect the welding quality and construction safety of steel structures.

[0003] Based on this, existing technologies have proposed a method for identifying welding operation risks during steel structure welding operations, which primarily relies on manual judgment based on experience. Specifically, technicians observe workers' welding operations on-site or through real-time communication video recordings. Based on their own welding experience, they determine whether there are welding operation risks that may affect the quality of steel structure welding and construction safety during the welding operation. However, existing technologies suffer from strong subjectivity and high missed detection rates, resulting in low accuracy in welding operation risk detection results, which cannot meet the high-precision and high-efficiency requirements of modern steel structure welding construction.

[0004] Therefore, there is an urgent need for a method for identifying welding operation risks in steel structure welding operations to improve the accuracy of identifying welding operation risks in steel structure welding operations. Summary of the Invention

[0005] In view of the above problems, this application provides a method and related equipment for identifying welding operation risks in steel structure welding operations, so as to achieve the purpose of improving the accuracy of identifying welding operation risks during steel structure welding operations. The specific solution is as follows:

[0006] A first aspect of the present application provides a method for identifying welding operation risks in a steel structure welding operation, comprising:

[0007] Acquire welding data during a current welding operation, the welding data including at least: operation data of the welding operation and image data of the welding area, the operation data being data collected by a sensor installed in the welding area during welding by the welding equipment;

[0008] Extracting data features of the welding data to obtain welding data features;

[0009] A pre-trained welding risk identification model is called to perform risk identification on the welding data features to obtain a risk identification result at the current moment output by the welding risk identification model. The risk identification result at least includes information characterizing whether the welding operation has a welding operation risk. The welding risk identification model is trained by using welding data feature samples as feature values ​​and risk identification result samples corresponding to the welding data feature samples as target values. The convolutional layer of the trained welding risk identification model extracts at least one data feature characterizing welding risk corresponding to at least one data type from the welding data feature samples. The gated recurrent unit layer analyzes the temporal change relationship between the at least one data feature characterizing welding risk corresponding to each data type, and the fully connected layer determines the risk identification result of the welding data at the current moment based on the temporal change relationship corresponding to each data type. The temporal change relationship is used to characterize the trend of welding risk change over time caused by the welding data feature corresponding to the data type to which it corresponds.

[0010] In a possible implementation, the training process of the welding risk identification model includes:

[0011] Obtaining multiple welding data samples for training;

[0012] Extracting data features of each welding data sample respectively to obtain welding data feature samples corresponding to each welding data sample;

[0013] According to each welding data sample, respectively mark the risk identification result sample corresponding to each welding data feature sample;

[0014] Constructing an initial welding risk identification model, wherein the initial welding risk identification model comprises at least: a convolutional layer, a gated recurrent unit layer, and a fully connected layer;

[0015] The initial welding risk identification model is trained based on the plurality of welding data feature samples and the risk identification result samples marked corresponding to each welding data feature sample to obtain a welding risk identification model.

[0016] In a possible implementation, the initial welding risk identification model is trained based on the plurality of welding data feature samples and the risk identification result sample labeled corresponding to each welding data feature sample to obtain the welding risk identification model, including:

[0017] Dividing all the welding data feature samples into a training data sample set and a verification data sample set;

[0018] Training the initial welding risk identification model based on the welding data feature samples in the training data sample set to obtain a trained initial welding risk identification model;

[0019] Inputting each of the welding data feature samples in the verification data sample set into the trained initial welding risk identification model to obtain an initial risk identification result corresponding to each of the welding data feature samples output by the trained initial welding risk identification model;

[0020] Determining a loss value of the trained initial welding risk identification model based on the initial risk identification result corresponding to each welding data feature sample and the risk identification result sample marked corresponding to each welding data feature sample;

[0021] Determining whether the loss value is greater than a preset threshold;

[0022] When the loss value is greater than the preset threshold, the model parameters of the initial welding risk identification model are adjusted to obtain an adjusted initial welding risk identification model, and for the adjusted initial welding risk identification model, the initial welding risk identification model is trained according to the welding data feature samples in the training data sample set to obtain the trained initial welding risk identification model and subsequent steps, until the number of training times reaches a preset number of times, or the loss value is not greater than the preset threshold, to obtain a welding risk identification model.

[0023] In a possible implementation, adjusting the model parameters of the initial welding risk identification model to obtain an adjusted initial welding risk identification model includes:

[0024] The Adam optimizer is called to adjust the model parameters of the initial welding risk identification model to obtain an adjusted initial welding risk identification model, wherein the model parameters include at least a weight parameter and a bias parameter.

[0025] In a possible implementation, the method further includes:

[0026] If the risk identification result indicates that the current welding process has a welding risk, generating an operation adjustment parameter corresponding to the operation data of the welding operation according to a preset risk response rule, and adjusting the operation data of the welding operation according to the operation adjustment parameter to obtain target operation data;

[0027] The target operation data is sent to a device for performing a welding operation.

[0028] In a possible implementation, the method further includes:

[0029] When the risk identification result indicates that there is a welding risk in the current welding process, risk warning information corresponding to the risk identification result is output.

[0030] In a possible implementation, the operation data of the welding operation includes at least: a welding current value, a welding voltage value, a welding temperature, and a sound signal generated by welding.

[0031] The third aspect of the present application provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the method for identifying welding operation risks in steel structure welding operations according to the first aspect or any implementation of the first aspect.

[0032] A fourth aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0033] The memory is used to store computer programs;

[0034] The processor is used to execute the computer program so that the electronic device can implement the method for identifying welding operation risks in steel structure welding operations according to the first aspect or any implementation of the first aspect.

[0035] A fifth aspect of the present application provides a computer storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the method of identifying welding operation risks in steel structure welding operations according to the first aspect or any implementation of the first aspect.

[0036] By utilizing the above-described technical solution, the present application provides a method for identifying welding operation risks in steel structure welding operations. This method uses a pre-trained welding risk identification model to perform risk identification on the welding data acquired during the welding process. Since the welding data includes welding operation data and image data of the welding results in the welding area, the welding data features extracted from the model can include both the operation features during the welding operation and the features of the welding results in the welding area. Furthermore, the welding data features are input into the welding risk identification model, allowing the convolutional layer to extract the risk features of the welding operation and the risk features of the welding results. Furthermore, the fully connected layer identifies whether the welding process is risky based on both the welding operation process and the welding operation results, thereby improving the accuracy and comprehensiveness of the risk identification results. Furthermore, compared to existing manual identification methods, the welding operation data in the present application is sensor-collected data, which is more accurate and contains data that cannot be perceived by humans, making it more comprehensive. Therefore, the welding risk identification model, based on more accurate and comprehensive operation data, can identify and obtain more accurate risk identification results.

[0037] In addition, this application obtains welding data in real time during the welding process and identifies risks in the welding process in real time. Compared with the existing manual identification method, it improves the efficiency of risk identification and solves the problem of delayed risk identification results in the existing manual identification method. It enables technicians or welding machines to standardize welding operations in a timely manner according to the risk identification results, thereby improving welding quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the originals and elements are not necessarily drawn to scale.

[0039] Figure 1 A system architecture diagram for implementing a method for identifying welding operation risks in steel structure welding operations provided in an embodiment of the present application;

[0040] Figure 2 A flow chart of a method for identifying welding operation risks in steel structure welding operations provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of the structure of a one-dimensional convolutional neural network provided in an embodiment of the present application;

[0042] Figure 4 A schematic diagram of the structure of a gated recurrent unit layer provided in an embodiment of the present application;

[0043] Figure 5 A schematic diagram of the model training process provided in the embodiment of the present application;

[0044] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The following describes the embodiments of the present application in conjunction with the accompanying drawings. The terms used in the implementation methods of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0046] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0047] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0048] This application can be applied to technical fields such as engineering construction and welding construction. The following takes the identification of welding operation risks during welding operations on steel structures as an example to introduce multiple application scenarios of the product.

[0049] First, an optional application scenario of the present application is introduced. A technician holds a welding gun and performs welding operations on a steel structure. While welding, the technician observes the welding results of the steel structure, as well as the posture of holding the welding gun, and independently determines whether the current welding involves welding operation risks that affect welding quality and construction safety. However, the identification of the above-mentioned welding operation risks depends on the technician's welding experience, and the identification results are highly subjective. In addition, due to the limited welding operation data that the technician can observe, the risk identification results obtained are not accurate enough, and there is a possibility of missed detection, which cannot meet the high-precision and high-efficiency requirements of modern steel structure welding construction.

[0050] To solve the above problems, the present invention provides a method for identifying welding operation risks in steel structure welding operations. The method for identifying welding operation risks in steel structure welding operations according to the present invention is described in detail below with reference to the accompanying drawings.

[0051] Reference Figure 1 , an embodiment of the present application provides a system architecture diagram of a method for realizing the identification of welding operation risks in steel structure welding operations, the system comprising: a data acquisition module and a data processing module, wherein the data acquisition module comprises: a current sensor, a voltage sensor, a temperature sensor, a sound sensor, a camera, etc., the sensors and cameras contained in the data acquisition module can be installed at locations such as welding guns and welding areas that can collect welding data related to welding operations during the welding process, and transmit the collected welding data to the data processing module to identify welding operation risks in the welding operation.

[0052] The data processing module includes a welding risk identification model, which is used to implement a method for identifying welding operation risks in steel structure welding operations. The data processing module can be but is not limited to being deployed in an application with data processing capabilities or a cloud service provided by a cloud-side server.

[0053] Reference Figure 2 , Figure 2 A flow chart of a method for identifying welding operation risks in steel structure welding operations provided in an embodiment of the present application is shown as follows: Figure 2 As shown, a method for identifying welding operation risks in steel structure welding operations provided by an embodiment of the present application may include steps S110 to S130, and these steps are described in detail below.

[0054] Step S110, obtaining welding data during the current welding operation, the welding data at least including: operation data of the welding operation and image data of the welding area, the operation data being data collected by the sensor installed in the welding area during welding by the welding equipment.

[0055] In an embodiment of the present application, various sensors and cameras in the data acquisition module are used to comprehensively collect welding data during the steel structure welding operation, wherein the welding data may include: operation data of the welding operation and image data of the welding area.

[0056] The operation data of the welding operation include at least: welding current value, welding voltage value, welding temperature, and sound signal generated by welding, and may also include: welding gun deviation angle, welding speed, welding wire extension length, welding direction, etc. Figure 1On the basis of the system architecture, optionally, a current sensor and a voltage sensor are installed at the welding power source to collect the welding current value and the welding voltage value in real time to characterize the welding energy input state; considering that the temperature is closely related to the performance of the welding heat affected zone, a temperature sensor is installed in the welding area to monitor the temperature data during the welding process; and because welding defects are often accompanied by unique sound characteristics, a sound sensor is installed in the welding area to collect the sound signal generated during welding. An image acquisition device such as an industrial camera or a visual sensor is used to capture images of the welding area to obtain image information such as weld formation and molten pool morphology, so as to facilitate risk identification from aspects such as welding results and steel structure welding status. Based on this, the embodiment of the present application captures various types of information during the steel structure welding operation as comprehensively and accurately as possible to provide a sufficiently accurate and comprehensive risk identification basis for subsequent welding operation risk identification.

[0057] Step S120: extracting data features of welding data to obtain welding data features.

[0058] Step S130 , calling a pre-trained welding risk identification model, performing risk identification on the welding data features, and obtaining a risk identification result at the current moment output by the welding risk identification model.

[0059] Among them, the risk identification result at least includes information characterizing whether there is a welding operation risk in the welding operation. The welding risk identification model is obtained by training the initial welding risk identification model with welding data feature samples as feature values ​​and risk identification result samples corresponding to the welding data feature samples as target values, so that the convolution layer of the trained welding risk identification model extracts at least one data feature characterizing welding risk corresponding to at least one data type from the welding data feature samples, the gated recurrent unit layer analyzes the temporal change relationship between at least one data feature characterizing welding risk corresponding to each data type, and the fully connected layer determines the risk identification result of the welding data at the current moment based on the temporal change relationship corresponding to each data type. The temporal change relationship is used to characterize the welding risk caused by the welding data feature corresponding to its corresponding data type, and the change trend over time.

[0060] In the embodiments of the present application, the welding risk identification model first extracts data features from the collected welding data before performing risk identification. The extracted welding data features are an abstract and condensed representation of key welding information in the welding data. It is understood that welding data collected by sensors or cameras often contains a large amount of noise and redundant information, which may affect the accuracy of risk identification. By extracting key information from the welding data through features, redundant information is filtered out, and the data dimension is reduced, resulting in a simplified representation of the welding data. Optionally, data features of the welding data can be extracted using methods such as convolutional neural networks (CNNs), autoencoders, recurrent neural networks (RNNs), and their variants (LSTMs, GRUs), etc., to obtain welding data features in the form of vectors or matrices, etc., without limitation herein.

[0061] In addition, the embodiment of the present application also uses welding data features to train the welding risk identification model. It can be understood that the low-dimensional welding data features remove redundant information, allowing the model to focus more on learning the key features and rules in the data, thereby improving the generalization ability and stability of the model.

[0062] Optionally, the welding risk identification model can call on existing large models, such as GPT, BERT, or visual base models, among others. Existing large models typically include: convolutional layers, gated recurrent unit layers, and fully connected layers. The target scene data is screened and labeled. In the embodiment of this application, welding defect images, welding operation data, etc. are selected. Subsequently, the underlying network of part of the large model is frozen to retain general feature extraction capabilities, and only the top-level structure or adapter module is trained. Ultimately, the large model can quickly adapt to the task of identifying risks in steel structure welding operations while retaining broad cognition, achieving high-precision, low-cost domain function migration.

[0063] In another possible implementation, the training process of the welding risk identification model includes: obtaining multiple welding data samples for training; extracting data features of each welding data sample respectively to obtain welding data feature samples corresponding to each welding data sample; labeling the risk identification result samples corresponding to each welding data feature sample according to each welding data sample; constructing an initial welding risk identification model, the initial welding risk identification model including at least: a convolutional layer, a gated recurrent unit layer and a fully connected layer; training the initial welding risk identification model based on multiple welding data feature samples and the risk identification result samples labeled corresponding to each welding data feature sample to obtain a welding risk identification model.

[0064] The above-mentioned data acquisition module is used to collect multiple welding data samples during the steel structure welding process, wherein each welding data sample contains operation data and image data at the same time or time period. The feature extraction method of the above-mentioned example is used to extract the data features of each welding data sample, and obtain the welding data feature samples corresponding to each welding data sample. Furthermore, each welding data feature sample is annotated with a corresponding risk identification result. Optionally, manual annotation or open source tools can be used for annotation. Specifically, when annotating, based on the operation data and image data of the welding data sample, it is determined whether there is a welding operation risk in the welding scene corresponding to the welding data sample. Based on this, the corresponding risk identification result sample annotated with the welding data feature sample corresponding to the welding data sample can also be annotated with the type of welding operation risk for welding scenes with welding operation risks, such as welding angle deviation, welding temperature not meeting the standard, etc.

[0065] At the same time, an initial welding risk identification model is constructed. In an embodiment of the present application, the initial welding risk identification model includes at least: a convolution layer, a gated recurrent unit layer, and a fully connected layer. Optionally, the initial welding risk identification model is a deep learning model, including: an input layer, two one-dimensional convolution layers, two GRU (gated recurrent unit) layers, and two fully connected layers, wherein the input layer is used to receive data input to the initial welding risk identification model, and the two one-dimensional convolution layers are each composed of 16 convolution kernels of size 4, with a step size of 1, and use a linear activation function to extract features from the input data. The extracted features pass through two GRU layers and finally pass through two fully connected layers. The first fully connected layer is used to integrate all features, and the second fully connected layer with N=1 outputs the decomposition result of the integrated features, wherein the decomposition result can be understood as the risk identification result for the integrated features.

[0066] Among them, one-dimensional convolutional neural network (1DCNN) is a variant of CNN, which is used to process sequence data. Figure 3 , a schematic diagram of the structure of a one-dimensional convolutional neural network provided in an embodiment of the present application, wherein the one-dimensional convolutional neural network is composed of an input layer, a convolutional layer, a pooling layer, a fully connected layer, an activation function, and an output layer. The alternating use of convolutional layers and pooling layers progressively increases the learning depth of the model and can also reduce the data dimension while ensuring the extraction of key information. The above structural design enables 1DCNN to excel in extracting high-dimensional features in time or space from one-dimensional sequence data, thereby performing effective data recognition and classification. Among them, each convolutional layer applies multiple convolution kernels to filter the input data, captures local features in the input data, reduces the dimension of the local features by the pooling layer, and subsamples the local features after dimensionality reduction to obtain the final data features, so as to reduce the amount of calculation and prevent overfitting.

[0067] The GRU (Gated Recurrent Unit) layer contains: Reset Gate and Update Gate. Figure 4 , a schematic diagram of the structure of the gated recurrent unit layer provided in an embodiment of the present application, wherein the reset gate is used to determine the degree to which the state information at the previous moment has been forgotten, and is adjusted by a sigmoid activation function, with the output value ranging from 0 to 1. The update gate is used to determine the degree to which the state information at the previous moment affects the current moment, that is, to control how much memory of the previous moment should be retained in the hidden state at the current moment. Referring to the following formula (1), the update gate, reset gate, current memory content, and final memory at the current time are determined.

[0068] (1)

[0069] Among them, Z t is the update gate formula, r t To reset the gate formula, is the current memory content, h t is the final memory of the current time, h i is the hidden state at time i, and W and U are both weight matrices.

[0070] On this basis, refer to Figure 5 , a flow chart of model training provided in an embodiment of the present application, uses a plurality of welding data feature samples and the risk identification results corresponding to the welding data feature samples to train the initial welding risk identification model. Specifically, the welding data feature samples are used as input data sets and input into the initial welding risk identification model constructed by the CNN and GRU networks to train the model. The convolutional layer of the welding risk identification model finally trained is able to extract at least one data feature representing welding risk corresponding to at least one data type from the welding data features, wherein the data types may include: temperature, current value, voltage value, etc. The gated recurrent unit layer analyzes the temporal change relationship between the at least one data feature representing welding risk corresponding to each data type, and the fully connected layer determines the risk identification result of the welding data at the current moment based on the temporal change relationship corresponding to each data type. The temporal change relationship is used to characterize the trend of welding risk caused by the welding data corresponding to its corresponding data type over time.

[0071] Based on this, the welding risk identification model can predict whether there will be a welding operation risk at the next moment or the current moment based on the changing trend of each data type over a period of time. For example, based on the temperature data of the current welding process processed by the welding risk identification model before the current moment, and the temperature data obtained at the current moment, the temperature data at the current moment and before the current moment in the current welding process are processed to determine the changing trend of the temperature data. It is found that the welding temperature shows a downward trend. When the temperature is low, the weld strength of the steel structure is reduced, making the weld prone to cracking under stress, that is, the welding quality is reduced. Based on this, the fully connected layer outputs the risk identification result of the welding operation risk at the current moment. Optionally, the type of welding operation risk can also be output as a low welding temperature.

[0072] In a possible implementation, an initial welding risk identification model is trained based on multiple welding data feature samples and risk identification result samples corresponding to each welding data feature sample, to obtain a welding risk identification model, including: dividing all welding data feature samples into a training data sample set and a verification data sample set; training the initial welding risk identification model based on the welding data feature samples in the training data sample set to obtain a trained initial welding risk identification model; inputting each welding data feature sample in the verification data sample set into the trained initial welding risk identification model to obtain an initial risk identification result corresponding to each welding data feature sample output by the trained initial welding risk identification model; and The initial risk identification results corresponding to the data feature samples and the risk identification result samples corresponding to the annotated welding data feature samples are used to determine the loss value of the trained initial welding risk identification model; whether the loss value is greater than a preset threshold is judged; when the loss value is greater than the preset threshold, the model parameters of the initial welding risk identification model are adjusted to obtain an adjusted initial welding risk identification model, and for the adjusted initial welding risk identification model, the initial welding risk identification model is trained based on the welding data feature samples in the training data sample set to obtain the trained initial welding risk identification model and subsequent steps, until the number of training times reaches the preset number of times, or the loss value is not greater than the preset threshold, and the welding risk identification model is obtained.

[0073] The loss function used to train the model and the conditions for stopping training are determined in advance. In the embodiment of the present application, the loss function can refer to the mean square error loss function of the following formula (2), and the training stopping condition can be: the number of training times reaches a preset number, or the loss function converges or the loss value of the loss function is less than a preset threshold.

[0074] (2)

[0075] Among them, loss is the loss value, N is the length of the input data sequence, is the initial risk identification result corresponding to the domain input data sequence output by the initial welding risk identification model, y i The risk identification result samples are labeled corresponding to the input data series.

[0076] After each training session, at least one welding data feature sample is extracted from multiple welding data feature samples as validation data. Alternatively, all welding data feature samples are pre-divided into a training dataset and a validation dataset, with the welding data feature samples in the validation dataset serving as validation data. The validation data is input into the trained initial welding risk identification model, and the model outputs an initial risk identification result corresponding to the validation data. Based on the initial risk identification result corresponding to the validation data and the labeled risk identification result sample corresponding to the validation data, a loss value corresponding to this training session is calculated. If the loss value is greater than a preset threshold, the initial risk identification result output by the model differs significantly from its corresponding risk identification result sample, indicating that the output of the initial welding risk identification model is inaccurate. Therefore, the model parameters of the initial welding risk identification model need to be optimized and adjusted, and training should be repeated until the loss value is no greater than the preset threshold. If the loss value is no greater than the preset threshold, it indicates that the initial risk identification result output by the model differs little or even nothing from its corresponding labeled risk identification result sample, indicating that the output of the initial welding risk identification model is accurate. Training can then be stopped, and the initial welding risk identification model obtained from this training session is used as the final welding risk identification model.

[0077] In a possible implementation, an Adam optimizer is called to adjust the model parameters of the initial welding risk identification model to obtain an adjusted initial welding risk identification model, where the model parameters include at least a weight parameter and a bias parameter.

[0078] In the embodiments of the present application, the Adam optimizer is applied to adjust the model parameters to optimize the weight and bias parameters in the model. Adam is a gradient-based optimization algorithm that updates parameter values ​​by calculating parameter gradients. Adam introduces a momentum term to address the oscillation problem of gradient descent in non-convex functions.

[0079] By adjusting the model parameters and loss function, an effective model training process can be achieved. According to the pre-set number of training times, the model parameters are continuously adjusted until the loss value reaches the preset threshold. Even if the loss value still does not reach the preset threshold after multiple training times, the model will stop running when the pre-set number of training times is reached. For example, after three training times, if the loss value calculated by the loss function still does not improve significantly, then the training of the initial welding risk identification model will be stopped.

[0080] Based on this, the welding risk identification model trained as above is used to process the welding data at the current moment, identify whether there is a welding operation risk, and output it as a risk identification result. Optionally, if there is a welding operation risk, the type of welding operation risk can also be output synchronously.

[0081] In summary, the method for identifying welding operation risks in steel structure welding operations provided by this application calls a pre-trained welding risk identification model to perform risk identification on the welding data obtained during the welding process. Since the welding data includes welding operation data and image data of the welding results of the welding area, the welding data features extracted from it can include operation features during the welding operation and features of the welding results of the welding area. Furthermore, the welding data features are input into the welding risk identification model, allowing the convolutional layer to extract risk features of the welding operation and risk features of the welding results. Furthermore, the fully connected layer identifies whether the welding process is risky based on both the welding operation process and the welding operation results, thereby improving the accuracy and comprehensiveness of the risk identification results. Moreover, compared to existing manual identification methods, the welding operation operation data of this application is sensor-collected data, which is more accurate and contains data that cannot be perceived by humans, making it more comprehensive. Therefore, the welding risk identification model can identify more accurate risk identification results based on more accurate and comprehensive operation data.

[0082] In addition, this application obtains welding data in real time during the welding process and identifies risks in the welding process in real time. Compared with the existing manual identification method, it improves the efficiency of risk identification and solves the problem of delayed risk identification results in the existing manual identification method. It enables technicians or welding machines to standardize welding operations in a timely manner according to the risk identification results, thereby improving welding quality.

[0083] Next, other possible implementations of the method for identifying welding operation risks in steel structure welding operations provided in an embodiment of the present application are described.

[0084] In one possible implementation, the method for identifying welding operation risks in steel structure welding operations may also include: when the risk identification result is that there is a welding risk in the current welding process, generating operation adjustment parameters corresponding to the operation data of the welding operation according to preset risk response rules, and adjusting the operation data of the welding operation according to the operation adjustment parameters to obtain target operation data; and sending the target operation data to the equipment used to perform the welding operation.

[0085] If the risk identification result indicates that the current welding process presents a welding risk, an adjustment strategy for the current welding operation is determined to mitigate the detected welding risk. Optionally, response rules for various types of risks can be pre-set, such as increasing the welding torch current when the welding temperature is low. Based on these rules, when the welding risk identification model outputs a risk identification result indicating that the current welding process presents a welding risk, a welding operation adjustment strategy can be generated to address the welding risk. The welding operation adjustment strategy includes adjustment parameters for the welding operation. Furthermore, the adjustment parameters in the welding operation adjustment strategy are transmitted to each welding device to enable timely adjustment. Optionally, the adjustment parameters in the welding operation adjustment strategy can be displayed on a display device connected to the data processing module or sent to a technician via remote communication, such as text message or email. Optionally, the above-described method can also be used to output risk warning information corresponding to the risk identification result when the risk identification result indicates that the current welding process presents a welding risk.

[0086] In an embodiment of the present application, a welding operation adjustment strategy generation model trained using reinforcement learning can be defined to generate a welding operation adjustment strategy corresponding to the risk identification results. First, the reinforcement learning agent, environment, action, and reward are defined. The agent is the welding risk identification model, the environment is the relevant data during the entire steel structure welding operation, such as welding data, and the action includes the welding operation adjustment strategy. The strategy can include adjusting welding operation parameters such as current, voltage, welding speed, different levels of risk warnings, and pausing welding for inspection.

[0087] The reward function of the welding operation adjustment strategy generation model is designed based on factors such as the accuracy of risk identification results, the degree of risk reduction after implementing the strategy, and whether welding quality has improved. For example, accurate risk identification and timely implementation of effective measures to reduce the risk are rewarded positively, while misjudgment of risk or incorrect measures leading to increased risk are rewarded negatively. This ensures that the welding operation adjustment strategies output by the trained welding operation adjustment strategy generation model are more accurate and better adapted to the current risk identification results, effectively addressing welding operation risks.

[0088] An electronic device is also provided in an embodiment of the present application. Figure 6 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 6 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0089] like Figure 6 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 602 or programs loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing device 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0090] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Figure 6 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0091] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the methods for identifying welding operation risks in steel structure welding operations provided in the embodiments of the present application.

[0092] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any one of the methods for identifying welding operation risks in steel structure welding operations provided in the embodiment of the present application.

[0093] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0094] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

[0095] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0096] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a training device or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website, a computer, a training device or a data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. A method for identifying welding operation risks in steel structure welding operations, characterized in that: include: Acquire welding data during a current welding operation, the welding data including at least: operation data of the welding operation and image data of the welding area, the operation data being data collected by a sensor installed in the welding area during welding by the welding equipment; Extracting data features of the welding data to obtain welding data features; A pre-trained welding risk identification model is called to perform risk identification on the welding data features to obtain a risk identification result at the current moment output by the welding risk identification model. The risk identification result at least includes information characterizing whether the welding operation has a welding operation risk. The welding risk identification model is trained by using welding data feature samples as feature values ​​and risk identification result samples corresponding to the welding data feature samples as target values. The convolutional layer of the trained welding risk identification model extracts at least one data feature characterizing welding risk corresponding to at least one data type from the welding data feature samples. The gated recurrent unit layer analyzes the temporal change relationship between the at least one data feature characterizing welding risk corresponding to each data type, and the fully connected layer determines the risk identification result of the welding data at the current moment based on the temporal change relationship corresponding to each data type. The temporal change relationship is used to characterize the trend of welding risk change over time caused by the welding data feature corresponding to the data type to which it corresponds.

2. The method for identifying welding operation risks in steel structure welding operations according to claim 1, characterized in that: The training process of the welding risk identification model includes: Obtaining multiple welding data samples for training; Extracting data features of each welding data sample respectively to obtain welding data feature samples corresponding to each welding data sample; According to each welding data sample, respectively mark the risk identification result sample corresponding to each welding data feature sample; Constructing an initial welding risk identification model, wherein the initial welding risk identification model comprises at least: a convolutional layer, a gated recurrent unit layer, and a fully connected layer; The initial welding risk identification model is trained based on the plurality of welding data feature samples and the risk identification result samples marked corresponding to each welding data feature sample to obtain a welding risk identification model.

3. The method for identifying welding operation risks in steel structure welding operations according to claim 2, characterized in that: The initial welding risk identification model is trained based on the plurality of welding data feature samples and the risk identification result sample marked corresponding to each welding data feature sample to obtain a welding risk identification model, including: Dividing all the welding data feature samples into a training data sample set and a verification data sample set; Training the initial welding risk identification model based on the welding data feature samples in the training data sample set to obtain a trained initial welding risk identification model; Inputting each of the welding data feature samples in the verification data sample set into the trained initial welding risk identification model to obtain an initial risk identification result corresponding to each of the welding data feature samples output by the trained initial welding risk identification model; Determining a loss value of the trained initial welding risk identification model based on the initial risk identification result corresponding to each welding data feature sample and the risk identification result sample marked corresponding to each welding data feature sample; Determining whether the loss value is greater than a preset threshold; When the loss value is greater than the preset threshold, the model parameters of the initial welding risk identification model are adjusted to obtain an adjusted initial welding risk identification model, and for the adjusted initial welding risk identification model, the initial welding risk identification model is trained according to the welding data feature samples in the training data sample set to obtain the trained initial welding risk identification model and subsequent steps, until the number of training times reaches a preset number of times, or the loss value is not greater than the preset threshold, to obtain a welding risk identification model.

4. The method for identifying welding operation risks in steel structure welding operations according to claim 3, characterized in that: The adjusting the model parameters of the initial welding risk identification model to obtain an adjusted initial welding risk identification model includes: The Adam optimizer is called to adjust the model parameters of the initial welding risk identification model to obtain an adjusted initial welding risk identification model, wherein the model parameters include at least a weight parameter and a bias parameter.

5. The method for identifying welding operation risks in steel structure welding operations according to any one of claims 1 to 4, characterized in that: Also includes: If the risk identification result indicates that the current welding process has a welding risk, generating an operation adjustment parameter corresponding to the operation data of the welding operation according to a preset risk response rule, and adjusting the operation data of the welding operation according to the operation adjustment parameter to obtain target operation data; The target operation data is sent to a device for performing a welding operation.

6. The method for identifying welding operation risks in steel structure welding operations according to any one of claims 1 to 4, characterized in that: Also includes: When the risk identification result indicates that there is a welding risk in the current welding process, risk warning information corresponding to the risk identification result is output.

7. The method for identifying welding operation risks in steel structure welding operations according to any one of claims 1 to 4, characterized in that: The operation data of the welding operation includes at least a welding current value, a welding voltage value, a welding temperature, and a sound signal generated by welding.

8. A computer program product, characterized in that The method comprises computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the method for identifying welding operation risks in steel structure welding operations according to any one of claims 1 to 7.

9. An electronic device, characterized in that: comprising at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is configured to execute the computer program so as to enable the electronic device to implement the method for identifying welding operation risks in steel structure welding operations according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the method for identifying welding operation risks in steel structure welding operations as described in any one of claims 1 to 7.

Images