Cover beam intelligent welding cloud control system

Abstract

The invention relates to the technical field of industrial intelligent control, in particular to a capping beam intelligent welding cloud control system, a pre-welding proofreading module comprises a capping beam steel bar frame unit and a suspicious steel bar judgment unit, and the pre-welding proofreading module verifies the number of steel bars based on comparison of a capping beam frame diagram and actual image data to ensure correct positioning of the steel bars before welding; and the reinforcing steel bar distance is verified, the reinforcing steel bars with unreasonable layout are judged, and warning information is generated. And determining the to-be-welded position of the reinforcing steel bar from the image data based on the convolutional neural network and generating a two-dimensional matrix to ensure that the welding sequence and position are reasonable and safe. And according to the verification result, the welding robot is adjusted in real time to conduct precise welding, and welding data are continuously updated. Welding head pressure, welding head depth and sound parameters are monitored in real time, a shutdown instruction is sent out when abnormity is detected, and dislocation or foreign matter invasion is avoided. According to the method, the welding efficiency and quality are remarkably improved, personal errors and risks are reduced, the production cost is reduced, the construction period is shortened, and intellectualization and automation of the bent cap welding process are achieved.

Description

Technical Field

[0001] The present invention relates to the field of industrial intelligent control technology, and in particular to a cap beam intelligent welding cloud control system. Background Art

[0002] The cap beam is an important component of bridges and other large structures that bear the load of the superstructure. It is usually located on the top of the pier or abutment. Its main function is to transfer the bridge deck or bridge truss to the lower structure. The quality of the cap beam directly affects the stability and safety of the entire bridge. The cap beam skeleton is the basic component of the cap beam, which refers to the steel bars connected together by welding to form a skeleton that can bear and transfer loads. This skeleton needs to be accurately positioned and connected during the construction process to ensure the integrity and uniformity of the structure. In the prior art, the welding of the cap beam skeleton mainly relies on manual operation. The specific steps include: workers manually measure and arrange the steel bars according to the design drawings, and then use welding equipment for welding. Although this method meets the construction needs to a certain extent, it has the following significant disadvantages: the manual operation welding process is complicated, time-consuming, and requires a lot of manpower, resulting in a longer construction period. Due to manual measurement and arrangement of steel bars, it is difficult to avoid human errors, the positioning and welding accuracy of steel bars is difficult to guarantee, and problems such as steel bar offset and uneven spacing are prone to occur, affecting the overall quality of the cap beam.

[0003] In the prior art, the steel bars are fixed by a welding robot and welded by a welding robot that can move along the XY axis. Most of the existing welding robots rely on preset welding paths and programs and lack adaptive capabilities, resulting in welding work still being performed when steel bar deviations occur or the steel bars are not ready. The robots are highly dependent on repeated manual inspections and supervision, and are unable to stop automatically according to real-time conditions, resulting in a high degree of dependence on personnel and reduced production efficiency. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a cap beam intelligent welding cloud control system.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A cap beam intelligent welding cloud control system, comprising: an image receiving module, a pre-welding proofreading module, a welding robot cloud control module and an abnormality monitoring module, wherein the pre-welding proofreading module comprises a cap beam reinforcement frame unit and a suspicious reinforcement judgment unit, the image receiving module, the pre-welding proofreading module and the abnormality monitoring module are respectively connected to the welding robot cloud control module, and the image receiving module is connected to the pre-welding proofreading module;

[0007] The image receiving module is used to receive image data of the steel bar fixing platform;

[0008] The cap beam reinforcement frame unit is used to receive the cap beam frame diagram, compare the cap beam frame diagram with the image data of the reinforcement fixing platform, and verify the number of reinforcements;

[0009] The suspicious steel bar determination unit is used to verify the steel bar spacing according to the image data of the steel bar fixing platform, and determine the suspicious steel bars based on the steel bar spacing;

[0010] The welding robot cloud control module determines the steel bar to be welded position from the image data of the steel bar fixing platform based on the convolutional neural network, generates a two-dimensional matrix according to the steel bar to be welded position, updates the parameters of the two-dimensional matrix based on the skip welding logic, controls the welding robot to move and weld based on the verification result of the pre-welding proofreading module and according to the updated two-dimensional matrix, receives the displacement and welding data of the welding robot in real time and updates the two-dimensional matrix;

[0011] The abnormality monitoring module is used to receive the working parameters of the welding robot during welding, judge the welding abnormality according to the working parameters and send a stop command to the welding robot cloud control module, wherein the working parameters include welding head pressure, welding head depth and sound parameters.

[0012] Furthermore, the comparison between the cap beam frame diagram and the image data of the steel bar fixing platform to verify the steel bar position and the steel bar quantity includes the following steps:

[0013] Preprocess the cap beam frame diagram and real-time image data, including denoising, grayscale conversion and image enhancement;

[0014] Use image recognition algorithm to extract features from preprocessed images, identify separated steel bars and count the number of steel bars;

[0015] The number of steel bars in the identified steel bar fixing platform is compared with the number in the cap beam frame drawing to generate a deviation value for the number of steel bars.

[0016] Further, the method of verifying the steel bar spacing according to the image data of the steel bar fixing platform, determining suspicious steel bars based on the steel bar spacing and generating an alarm includes the following steps:

[0017] Classify and mark the steel bars based on the identified separated steel bars and steel bar location data;

[0018] Calculate the distance between adjacent bars based on the marking results;

[0019] Compare the distance between adjacent steel bars with the preset standard spacing to identify suspicious steel bars whose spacing does not meet the design requirements.

[0020] Furthermore, the convolutional neural network is trained by the following steps:

[0021] The steel bars identified and separated in the image data of the steel bar fixing platform are edge-calibrated based on the same color value to obtain an image to be welded;

[0022] Annotate the image to be welded, including marking the welded position with another same color value to obtain the target image:

[0023] The convolutional neural network is trained with the image to be welded as input and the target image as output.

[0024] Furthermore, the formula of the convolutional neural network is as follows:

[0025]

[0026] Among them, Y is the output vector, which represents the predicted value of the steel bar to be welded; X k is the input vector, specifically the image to be welded; i and j are the row index and column index of the convolution kernel; k is the number of channels of the input image; W ij is the weight matrix of the convolutional layer; W ijk is the weight parameter of the convolution kernel; b ij is the bias vector of the convolutional layer; φ is the activation function of the convolutional layer; σ is the activation function of the fully connected layer; c is the bias vector of the fully connected layer; n is the height of the convolution kernel; m is the width of the convolution kernel.

[0027] Furthermore, the generating of a two-dimensional matrix according to the positions of the steel bars to be welded and updating the parameters of the two-dimensional matrix based on the skip welding logic comprises the following steps:

[0028] Generate a two-dimensional matrix through the target image output by the convolutional neural network, each element in the two-dimensional matrix represents the coordinates and state information of a point to be welded;

[0029] The jump welding logic is applied to update the parameters of the two-dimensional matrix, including updating the parameters of the continuous points to be welded from the parameters to be welded to the parameters to be welded through the preset step size.

[0030] Furthermore, the control of the welding robot to perform displacement and welding based on the verification result of the pre-welding verification module and according to the updated two-dimensional matrix includes the following steps:

[0031] Obtain the verification result of the pre-welding verification module;

[0032] If the deviation value of the number of steel bars is 0 and no suspicious steel bars are found, the welding robot is controlled to move and weld according to the updated two-dimensional matrix;

[0033] If the deviation value of the number of steel bars is not 0 or suspicious steel bars are found, the welding robot is controlled to stop displacement and welding.

[0034] Furthermore, controlling the welding robot to perform displacement and welding comprises the following steps:

[0035] According to the parameters to be welded in the two-dimensional matrix, the welding robot is controlled to move the X-axis and Y-axis to the position of the parameters to be welded and weld through the welding head;

[0036] After each point to be welded is completed, the location is obtained and the two-dimensional matrix is ​​updated, and the parameters to be welded are updated to the welded parameters.

[0037] Further, judging welding abnormality according to working parameters comprises the following steps:

[0038] Get the welding head depth, welding head pressure and sound parameters and the corresponding timestamp of a single welding when there is no abnormality in the history;

[0039] Based on the same timestamp, the value ranges of welding head pressure and sound parameters at different welding head depths are obtained;

[0040] If the welding head pressure or sound parameters in real-time welding work do not belong to their value range at the current welding head depth, it is judged as welding abnormality;

[0041] If the welding head pressure or sound parameters do not belong to the value range at the current welding head depth during real-time welding work, it is judged that there is no abnormality.

[0042] The beneficial effect of the present invention is that the system can obtain the layout and quantity information of the steel bars in real time through the image processing technology, reducing the need for manual intervention. The pre-welding proofreading module includes a cap beam steel bar frame unit and a suspicious steel bar judgment unit. The cap beam steel bar frame unit compares the cap beam frame diagram with the actual image data to check the number of steel bars and ensure the correct positioning of the steel bars before welding. The suspicious steel bar judgment unit verifies whether there are suspicious steel bars with unreasonable layout by verifying the steel bar spacing, and generates an alarm message to prompt the operator to check and adjust. The pre-welding proofreading module improves the accuracy of steel bar layout through image recognition algorithm and feature extraction technology. The welding robot cloud control module determines the position of the steel bar to be welded from the image data based on the convolutional neural network and generates a corresponding two-dimensional matrix. The convolutional neural network can accurately identify the steel bar to be welded and generate the image to be welded by training a large amount of welding image data. The parameters of the two-dimensional matrix are updated based on the jump welding logic to ensure the rationality and safety of the welding sequence and position. According to the verification result of the pre-welding proofreading module, the welding robot cloud control module adjusts the welding robot in real time for precise welding, and continuously updates the welding data to ensure that each welding point meets the design requirements. The abnormality monitoring module monitors the working parameters in real time during the welding process, including welding head pressure, welding head depth and sound parameters. Once an abnormality is detected, the system will immediately issue a stop command to prevent welding defects. The module accurately determines whether the welding is abnormal by comparing the real-time working parameters with the parameter range of historical normal operation, thereby avoiding dislocation or debris in the welding and ensuring the stability and reliability of quality. The present invention significantly improves the welding efficiency and quality through real-time multi-parameter monitoring, reduces the errors and risks of human operation, reduces the production cost and construction period, and realizes the intelligence and automation of the cap beam welding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a structural schematic diagram of a cap beam intelligent welding cloud control system in the present invention.

[0044] Figure 2 It is a flow chart of the training steps of the convolutional neural network used to determine the position of the steel bar to be welded in the present invention. DETAILED DESCRIPTION

[0045] See also Figure 1-2 As shown, the present invention relates to a cap beam intelligent welding cloud control system, comprising: an image receiving module, a pre-welding proofreading module, a welding robot cloud control module and an abnormality monitoring module, wherein the pre-welding proofreading module comprises a cap beam reinforcement frame unit and a suspicious reinforcement judgment unit, the image receiving module, the pre-welding proofreading module and the abnormality monitoring module are respectively connected to the welding robot cloud control module, and the image receiving module is connected to the pre-welding proofreading module;

[0046] The image receiving module is used to receive image data of the steel bar fixing platform;

[0047] The cap beam reinforcement frame unit is used to receive the cap beam frame diagram, compare the cap beam frame diagram with the image data of the reinforcement fixing platform, and verify the number of reinforcements;

[0048] The suspicious steel bar determination unit is used to verify the steel bar spacing according to the image data of the steel bar fixing platform, and determine the suspicious steel bars based on the steel bar spacing;

[0049] The welding robot cloud control module determines the steel bar to be welded position from the image data of the steel bar fixing platform based on the convolutional neural network, generates a two-dimensional matrix according to the steel bar to be welded position, updates the parameters of the two-dimensional matrix based on the skip welding logic, controls the welding robot to move and weld based on the verification result of the pre-welding proofreading module and according to the updated two-dimensional matrix, receives the displacement and welding data of the welding robot in real time and updates the two-dimensional matrix;

[0050] The abnormality monitoring module is used to receive the working parameters of the welding robot during welding, judge the welding abnormality according to the working parameters and send a stop command to the welding robot cloud control module, wherein the working parameters include welding head pressure, welding head depth and sound parameters.

[0051] It should be noted that the image receiving module is responsible for receiving the image data of the steel bar fixing platform. The steel bar fixing platform is a work surface for placing and fixing the cap beam steel bars to ensure the stability of the position and posture of the steel bars during the welding process. The image receiving module uses a high-resolution industrial camera and advanced image processing technology to capture and process the image data of the steel bar layout in real time. These image data include key information such as the spatial position, quantity and spacing of the steel bars. The image receiving module also has pre-processing functions such as denoising, grayscale conversion and image enhancement to ensure the clarity and accuracy of the image data and provide a reliable basis for subsequent image analysis. The cap beam steel bar frame unit compares the cap beam frame diagram with the real-time image data of the steel bar fixing platform, extracts the steel bar features in the image using an image recognition algorithm, calculates the actual number of steel bars, and compares it with the number of steel bars in the design drawing to generate a deviation report on the number of steel bars. This step ensures the correct arrangement of the steel bars before welding and avoids welding quality problems caused by incorrect steel bar numbers. The suspicious steel bar judgment unit verifies the steel bar spacing based on the image data of the steel bar fixing platform. By calculating the distance between adjacent steel bars and comparing it with the preset standard spacing, suspicious steel bars whose spacing does not meet the design requirements are identified. For the identified suspicious steel bars, the unit locates their specific positions in the image and generates an alarm message to prompt the operator to check and adjust. This step ensures that the spacing of the steel bars meets the design requirements and improves the overall quality of the cap beam through precise image analysis. The welding robot cloud control module is the core part of the system. It determines the steel bar to be welded from the image data of the steel bar fixing platform based on the convolutional neural network. Through training on a large amount of welding image data, the convolutional neural network can accurately identify the position of the steel bar to be welded and generate a two-dimensional matrix representing the points to be welded. Each element in the matrix represents the coordinates and status information of a point to be welded. The system applies the jump welding logic to update the parameters of the two-dimensional matrix, optimize the welding path and sequence, and avoid material deformation caused by heat concentration. According to the updated two-dimensional matrix, the welding robot accurately controls the displacement and welding operation of the welding robot, and receives and processes the displacement and welding data generated during the welding process in real time to ensure the accuracy and quality of each welding point. Specifically, the moving path design of the welding robot system is based on the XY axis plane coordinate system, and covers the entire top of the steel bar fixing platform by setting two parallel tracks for Y-axis movement and a horizontal axis for X-axis movement. The specific description is as follows: Y-axis movement, the welding robot is installed on two parallel tracks, which are arranged along the longitudinal direction of the steel bar fixing platform, so that the robot can move forward and backward along the Y-axis direction. The track adopts high-precision linear guide rails and servo motor drive to ensure that the robot moves stably and accurately in the Y-axis direction. X-axis movement: A horizontal axis is set between the two Y-axis tracks, and a welding gun and an up and down telescopic device are installed on the horizontal axis. The horizontal axis moves left and right along the X-axis direction through a driving mechanism. The X-axis movement is also driven by linear guide rails and servo motors to ensure the movement accuracy in the X-axis direction.The design of the horizontal axis enables the welding robot to cover the entire upper area of ​​the steel bar fixing platform, ensuring that each welding point can be accurately located and welded. The welding robot is operated by an integrated control system, which receives data from the image receiving module and the pre-welding proofreading module, and adjusts the welding operation according to the preset welding path and real-time data. The control system generates and updates a two-dimensional matrix through convolutional neural networks and jump welding logic to guide the mechanical telescopic device to perform precise welding operations. The control system can also receive and process the displacement and welding data generated during the welding process in real time, dynamically adjust the welding parameters, and improve the accuracy and consistency of welding. The abnormal monitoring module monitors the working parameters of the welding robot in real time during the welding process, including welding head pressure, welding head depth and sound parameters. The module determines whether the welding is normal by comparing with the preset standard parameters. Once an abnormal situation is detected, such as welding head pressure or welding temperature beyond the normal range, the system will immediately issue a stop command to prevent welding defects. The abnormal monitoring module also records abnormal data to provide a basis for subsequent analysis and improvement.

[0052] Furthermore, the comparison between the cap beam frame diagram and the image data of the steel bar fixing platform to verify the steel bar position and the steel bar quantity includes the following steps:

[0053] Preprocess the cap beam frame diagram and real-time image data, including denoising, grayscale conversion and image enhancement;

[0054] Use image recognition algorithm to extract features from preprocessed images, identify separated steel bars and count the number of steel bars;

[0055] The number of steel bars in the identified steel bar fixing platform is compared with the number in the cap beam frame drawing to generate a deviation value for the number of steel bars.

[0056] In some embodiments, first, the real-time image data of the cap beam frame diagram and the steel bar fixing platform are preprocessed. The preprocessing steps include denoising, grayscale transformation and image enhancement. Denoising is to remove noise in the image and improve the quality of the image. Common denoising methods include Gaussian filtering, median filtering, etc. Grayscale transformation is to convert a color image into a grayscale image to simplify the subsequent image processing steps. Image enhancement makes the edges of the steel bars clearer by adjusting the contrast and brightness of the image, thereby improving the recognizability of the image. Next, the image recognition algorithm is used to extract features from the preprocessed image. The purpose of feature extraction is to identify and separate the steel bars in the image. Commonly used image recognition algorithms include edge detection algorithms (such as Canny edge detection), Hough transforms, and template matching. These algorithms can accurately identify the edges and positions of steel bars and generate a binary image containing steel bar features. In the binary image, the steel bars are marked as white areas, and the background is black areas. Then, the extracted steel bar features are further processed to calculate the number of steel bars. This step identifies each independent steel bar by performing a connected domain analysis on the white areas in the binary image. The connected domain analysis algorithm can detect and mark the connected regions in the image and count the number of these regions. In this way, the number of steel bars on the steel bar fixing platform can be accurately calculated. Finally, the number of steel bars on the identified steel bar fixing platform is compared with the design number in the cap beam frame drawing to generate the deviation value of the number of steel bars.

[0057] Further, the method of verifying the steel bar spacing according to the image data of the steel bar fixing platform, determining suspicious steel bars based on the steel bar spacing and generating an alarm includes the following steps:

[0058] Classify and mark the steel bars based on the identified separated steel bars and steel bar location data;

[0059] Calculate the distance between adjacent bars based on the marking results;

[0060] Compare the distance between adjacent steel bars with the preset standard spacing to identify suspicious steel bars whose spacing does not meet the design requirements.

[0061] In some embodiments, the system obtains real-time image data of the steel bar fixing platform and preprocesses the data. The preprocessing steps include denoising, grayscale conversion and image enhancement to ensure the clarity and accuracy of the image. These preprocessing techniques make the steel bar features in the image more obvious, providing a reliable basis for subsequent identification and analysis. Then, the system uses an image recognition algorithm to extract features from the preprocessed image, identify and separate the specific positions of the steel bars. The image recognition algorithm can be trained and applied using a convolutional neural network (CNN), and by learning a large number of steel bar images, it can accurately identify the positions of the steel bars. The identified steel bar position data will serve as the basis for subsequent processing. Based on the identified steel bar position data, the system classifies and marks the steel bars. The classification and marking process includes numbering each steel bar and marking its specific position in the image. The purpose of the classification and marking is to clarify the spatial position of each steel bar, which is convenient for subsequent distance calculation and analysis. After completing the classification and marking of the steel bars, the system begins to calculate the distance between adjacent steel bars. The specific method is to traverse all the marked steel bar positions and calculate the straight-line distance between each pair of adjacent steel bars. The calculation result will generate a set of distance data, reflecting the spacing of each steel bar in actual construction. After completing the classification and marking of the steel bars, the system begins to calculate the distance between adjacent steel bars. The specific method is to traverse all the marked steel bar positions and calculate the straight-line distance between each pair of adjacent steel bars. The calculation results will generate a set of distance data, which reflects the spacing of each steel bar in actual construction. Once suspicious steel bars are identified, the system will highlight the location of these steel bars in the image and generate an alarm message. The alarm information includes the specific location of the suspicious steel bars, the deviation value between the actual spacing and the standard spacing, and the recommended treatment measures. The system will push the alarm information to the operator at the construction site, prompting them to check and adjust the suspicious steel bars.

[0062] Furthermore, the convolutional neural network is trained by the following steps:

[0063] The steel bars identified and separated in the image data of the steel bar fixing platform are edge-calibrated based on the same color value to obtain an image to be welded;

[0064] Annotate the image to be welded, including marking the position to be welded by another same color value to obtain a target image;

[0065] The convolutional neural network is trained with the image to be welded as input and the target image as output.

[0066] In some embodiments, first, the system acquires image data of the steel bar fixing platform, which includes the spatial position and arrangement of all steel bars on the platform. In order to ensure the quality of the data, the image data needs to be preprocessed, including denoising, grayscale conversion and image enhancement. These preprocessing steps can improve the clarity of the image and make the edges of the steel bars more obvious. Next, the system extracts features from the preprocessed image data to identify and separate the steel bars in the image. The feature extraction process usually uses an edge detection algorithm, such as Canny edge detection, to identify the outline and position of the steel bars. The steel bars after identification and separation will be marked as edges with the same color value in the image, and this step is called edge calibration. Through edge calibration, the system can clearly identify the specific position and shape of each steel bar and generate an image to be welded. After generating the image to be welded, the system annotates the image, and the annotation process refers to clearly marking the specific positions where welding is required in the image. To achieve this goal, the system uses another same color value to mark these positions to be welded. The marked positions to be welded will be displayed as specific color blocks in the image, called target images. This step ensures the clear identification of the welding position, which is convenient for subsequent neural network training. After completing the image annotation, the system uses the image to be welded as input and the target image as output to start training the convolutional neural network (CNN). The convolutional neural network is a deep learning algorithm that can automatically extract high-level features in the image and realize the recognition and classification of complex image patterns by learning from a large amount of annotated data. During the training process, the system inputs the image to be welded into the convolutional neural network, and the network gradually extracts the feature information in the image through multiple convolutional layers, pooling layers, and fully connected layers. Each convolutional layer uses several convolutional kernels to perform convolution operations on the input image to extract local features of the image. The pooling layer reduces the amount of data by downsampling, retains important features, and improves computational efficiency. The fully connected layer maps the extracted features to the output space to generate the final prediction result, that is, the position of the steel bar to be welded. During the training process, the system continuously adjusts the weight parameters and bias vectors of the convolution kernel through the back-propagation algorithm to minimize the error between the prediction result and the target image. The error is usually measured by the mean square error or cross entropy loss function. Through multiple iterations and training with a large amount of data, the convolutional neural network gradually learns how to accurately identify the position to be welded in the image.

[0067] Furthermore, the formula of the convolutional neural network is as follows:

[0068]

[0069] Among them, Y is the output vector, which represents the predicted value of the steel bar to be welded; X k is the input vector, specifically the image to be welded; i and j are the row index and column index of the convolution kernel; k is the number of channels of the input image; W ij is the weight matrix of the convolutional layer; Wijk is the weight parameter of the convolution kernel; b ij is the bias vector of the convolutional layer; φ is the activation function of the convolutional layer; σ is the activation function of the fully connected layer; c is the bias vector of the fully connected layer; n is the height of the convolution kernel; m is the width of the convolution kernel.

[0070] First, the system acquires image data of the steel bar fixing platform, which is preprocessed, including denoising, grayscale transformation, and image enhancement, to ensure the clarity and accuracy of the image. These preprocessing steps make the steel bar features in the image more obvious, providing a reliable basis for subsequent recognition and analysis. During the training process, the network optimizes parameters through the back propagation algorithm. The loss function usually selects the mean square error (MSE) or cross-entropy loss function to measure the error between the predicted value and the true label. Through the gradient descent method, the network continuously adjusts the weight matrix W and the bias vector b to minimize the loss function. The specific training steps include: Forward propagation: The input image is processed through multiple layers of convolution, pooling, and fully connected layers to generate a predicted output. Calculate the loss: Calculate the loss function based on the predicted output and the true label. Back propagation: Calculate the gradient of the loss function relative to each parameter, and update the parameters based on these gradients.

[0071] Furthermore, the generating of a two-dimensional matrix according to the positions of the steel bars to be welded and updating the parameters of the two-dimensional matrix based on the skip welding logic comprises the following steps:

[0072] Generate a two-dimensional matrix through the target image output by the convolutional neural network, each element in the two-dimensional matrix represents the coordinates and state information of a point to be welded;

[0073] The jump welding logic is applied to update the parameters of the two-dimensional matrix, including updating the parameters of the continuous points to be welded from the parameters to be welded to the parameters to be welded through the preset step size.

[0074] In some embodiments, first, a two-dimensional matrix is ​​generated by a target image output by a convolutional neural network. The convolutional neural network is trained to accurately identify and locate the points to be welded. During the recognition process, the network extracts features of the steel bars in the image and generates a target image containing all the points to be welded. Each point to be welded is represented as a bright spot in the image, and the position of these bright spots is the position of the steel bars to be welded. The generated target image is processed and converted into a two-dimensional matrix. Each element in the two-dimensional matrix represents the state information of a point to be welded, and a numerical value is used to represent different states: 0 represents to be welded, 1 represents welded, and 2 represents current welding. In the initial state, the state of all welding points is set to 0 (to be welded). Select the initial welding point: select a point to be welded from the two-dimensional matrix as the starting point, usually select the first point to be welded in the matrix, update its state to 2 (current welding), and then start the welding operation. Perform the welding operation: The welding robot moves to the selected welding point for welding operation. After welding is completed, the state of the welding point is updated from 2 (current welding) to 1 (welded). A jump welding step size (such as 2 welding points) is preset to determine the position of the next point to be welded. The step length can be adjusted according to specific welding requirements. Repeat the above steps until the status of all welding points is updated to 1 (welded).

[0075] Furthermore, the control of the welding robot to perform displacement and welding based on the verification result of the pre-welding verification module and according to the updated two-dimensional matrix includes the following steps:

[0076] Obtain the verification result of the pre-welding verification module;

[0077] If the deviation value of the number of steel bars is 0 and no suspicious steel bars are found, the welding robot is controlled to move and weld according to the updated two-dimensional matrix;

[0078] If the deviation value of the number of steel bars is not 0 or suspicious steel bars are found, the welding robot is controlled to stop displacement and welding.

[0079] In some embodiments, first, the system obtains the verification result of the pre-welding proofreading module. The pre-welding proofreading module includes a cap beam reinforcement frame unit and a suspicious reinforcement judgment unit, which are responsible for verifying the number and spacing of reinforcements. The cap beam reinforcement frame unit compares the image data of the cap beam frame diagram with the reinforcement fixing platform to determine whether the actual number of reinforcements is consistent with the design drawing. The suspicious reinforcement judgment unit verifies the reinforcement spacing according to the image data of the reinforcement fixing platform, identifies and marks suspicious reinforcements whose spacing does not meet the design requirements. After obtaining the verification result, the system first checks the deviation value of the number of reinforcements. A deviation value of 0 indicates that the actual number of reinforcements is completely consistent with the design drawing. At this time, the system also needs to check whether there are suspicious reinforcements. If the verification result shows that no suspicious reinforcements are found, the system can control the welding robot to perform displacement and welding operations according to the updated two-dimensional matrix. For example, the element state in the two-dimensional matrix is ​​0, indicating that it is to be welded, and 1 indicates that it has been welded. The welding robot first moves itself to the position of the first welding point to be welded in the two-dimensional matrix with a state of 0, and starts welding. After welding is completed, the system updates the state of the welding point to 1 (welded), and selects the next welding point according to the preset jump welding logic. The welding robot continues to move and weld according to the instructions of the two-dimensional matrix until all the points to be welded have completed the welding operation. However, if the verification result shows that the deviation value of the number of steel bars is not 0, or suspicious steel bars are found, the system immediately stops the welding robot from moving and welding. At this time, the system generates an alarm message and notifies the on-site operator of the specific location of the deviation and suspicious steel bars. The operator needs to check and adjust the actual construction situation to ensure that the number and spacing of steel bars meet the design requirements. For example, in actual construction, if the cap beam steel bar frame unit detects that the number of steel bars is less than the requirements of the design drawings, the system will immediately suspend the welding operation and prompt the operator to supplement the missing steel bars. Similarly, if the suspicious steel bar judgment unit identifies that the steel bar spacing is too large or too small, the system will prompt the operator to readjust the steel bar spacing. Through the above steps, the system can effectively ensure the accuracy and quality of the welding operation. Based on the verification results of the pre-welding proofreading module, the system can dynamically adjust the welding plan to avoid welding defects caused by steel bar quantity or spacing problems. This not only improves construction efficiency, but also significantly enhances the overall stability and safety of the structure.

[0080] Furthermore, controlling the welding robot to perform displacement and welding comprises the following steps:

[0081] According to the parameters to be welded in the two-dimensional matrix, the welding robot is controlled to move the X-axis and Y-axis to the position of the parameters to be welded and weld through the welding head;

[0082] After each point to be welded is completed, the location is obtained and the two-dimensional matrix is ​​updated, and the parameters to be welded are updated to the welded parameters.

[0083] Further, judging welding abnormality according to working parameters comprises the following steps:

[0084] Get the welding head depth, welding head pressure and sound parameters and the corresponding timestamp of a single welding when there is no abnormality in the history;

[0085] Based on the same timestamp, the value ranges of welding head pressure and sound parameters at different welding head depths are obtained;

[0086] If the welding head pressure or sound parameters in real-time welding work do not belong to their value range at the current welding head depth, it is judged as welding abnormality;

[0087] If the welding head pressure or sound parameters do not belong to the value range at the current welding head depth during real-time welding work, it is judged that there is no abnormality.

[0088] In some embodiments, first, the system obtains the welding head depth, welding head pressure and sound parameters when there is no abnormal operation in the history, as well as the timestamps of the single welding corresponding to these parameters. The collection of historical data is to ensure that these data can accurately reflect the parameter range under normal working conditions by long-term monitoring and recording the changes of various parameters during normal welding operation. After obtaining the historical data without abnormalities, the system obtains the value range of welding head pressure and sound parameters under different welding head depths based on the same timestamp. This step uses a data analysis method to perform statistical analysis on the historical data to determine the normal value range of welding head pressure and sound parameters at each welding head depth. For example, when the welding head depth is 5mm, the system may determine that the welding head pressure is between 50-60N and the sound parameter is between 70-80dB as the normal range. Next, the system monitors the welding head depth, welding head pressure and sound parameters in real-time welding work. When the welding robot starts the welding operation, the system records the current welding head depth in real time, and compares and judges the welding head pressure and sound parameters according to the normal value range corresponding to the current welding head depth. Acquiring real-time parameters: The system acquires the welding head depth, welding head pressure and sound parameters in the current welding operation in real time. For example, at a certain moment, the system records that the current welding head depth is 8mm, the welding head pressure is 65N, and the sound parameter is 85dB. The system compares the welding head pressure and sound parameters acquired in real time with the preset normal value range. If the current welding head depth is 5mm, the normal value range of the welding head pressure is 50-60N, and the normal value range of the sound parameter is 70-80dB. The system determines whether the current welding head pressure and sound parameters are within the normal value range. If the welding head pressure or sound parameter does not belong to its value range at the current welding head depth, it is judged as welding abnormality. For example, if the current welding head pressure is 65N, which is not within the normal range of 50-60N, the system determines that the welding is abnormal, generates an alarm message and records the abnormal time and parameters. The system continuously monitors the welding head depth, welding head pressure and sound parameters in each welding operation cycle, and updates the judgment results in real time. If the welding head pressure and sound parameters are always within the normal value range during the entire welding process, the system determines that there is no abnormality and the welding operation is carried out normally. In particular, during the welding process, by monitoring the welding head depth, welding head pressure and sound parameters, welding abnormalities caused by the dislocation of steel bars or the presence of foreign objects can be discovered in time. For example, when the steel bars are misplaced, the welding head pressure may deviate significantly from the normal range, and the sound parameters will also fluctuate abnormally. The system can immediately identify this abnormality, suspend the welding operation, and issue an alarm to notify the operator to check and adjust. This method not only improves the accuracy of the welding process, but also effectively avoids welding defects caused by dislocation of steel bars or foreign objects. Through the system's real-time monitoring and automatic exception handling functions, the cap beam intelligent welding cloud control system can significantly improve the welding quality and ensure that each weld point meets the design requirements, thereby ensuring the overall stability and safety of bridges and other large structures.

[0089] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A cap beam intelligent welding cloud control system, characterized in that: include: An image receiving module, a pre-welding proofreading module, a welding robot cloud control module and an abnormality monitoring module, wherein the pre-welding proofreading module comprises a cap beam reinforcement frame unit and a suspicious reinforcement determination unit, the image receiving module, the pre-welding proofreading module and the abnormality monitoring module are respectively connected to the welding robot cloud control module, and the image receiving module is connected to the pre-welding proofreading module; The image receiving module is used to receive image data of the steel bar fixing platform; The cap beam reinforcement frame unit is used to receive the cap beam frame diagram, compare the cap beam frame diagram with the image data of the reinforcement fixing platform, and verify the number of reinforcements; The suspicious steel bar determination unit is used to verify the steel bar spacing according to the image data of the steel bar fixing platform, and determine the suspicious steel bars based on the steel bar spacing; The welding robot cloud control module determines the steel bar to be welded position from the image data of the steel bar fixing platform based on the convolutional neural network, generates a two-dimensional matrix according to the steel bar to be welded position, updates the parameters of the two-dimensional matrix based on the skip welding logic, controls the welding robot to move and weld based on the verification result of the pre-welding proofreading module and according to the updated two-dimensional matrix, receives the displacement and welding data of the welding robot in real time and updates the two-dimensional matrix; The abnormality monitoring module is used to receive the working parameters of the welding robot during welding, judge the welding abnormality according to the working parameters and send a stop command to the welding robot cloud control module, wherein the working parameters include welding head pressure, welding head depth and sound parameters.

2. According to claim 1, a cap beam intelligent welding cloud control system is characterized in that: The comparison between the cap beam frame diagram and the image data of the steel bar fixing platform to check the steel bar position and the steel bar quantity includes the following steps: Preprocess the cap beam frame diagram and real-time image data, including denoising, grayscale conversion and image enhancement; Use image recognition algorithm to extract features from preprocessed images, identify separated steel bars and count the number of steel bars; The number of steel bars in the identified steel bar fixing platform is compared with the number in the cap beam frame drawing to generate a deviation value for the number of steel bars.

3. The cap beam intelligent welding cloud control system according to claim 2 is characterized in that: The method of verifying the steel bar spacing according to the image data of the steel bar fixing platform, determining suspicious steel bars based on the steel bar spacing and generating an alarm comprises the following steps: Classify and mark the steel bars based on the identified separated steel bars and steel bar location data; Calculate the distance between adjacent bars based on the marking results; Compare the distance between adjacent steel bars with the preset standard spacing to identify suspicious steel bars whose spacing does not meet the design requirements.

4. The cap beam intelligent welding cloud control system according to claim 1 is characterized in that: The convolutional neural network is trained by the following steps: The steel bars identified and separated in the image data of the steel bar fixing platform are edge-calibrated based on the same color value to obtain an image to be welded; Annotate the image to be welded, including marking the position to be welded by another same color value to obtain a target image; The convolutional neural network is trained with the image to be welded as input and the target image as output.

5. The cap beam intelligent welding cloud control system according to claim 4 is characterized in that: The formula of the convolutional neural network is as follows: Among them, Y is the output vector, which represents the predicted value of the steel bar to be welded; X k is the input vector, specifically the image to be welded; i and j are the row index and column index of the convolution kernel; k is the number of channels of the input image; W ij is the weight matrix of the convolutional layer; W ijk is the weight parameter of the convolution kernel; b ij is the bias vector of the convolutional layer; φ is the activation function of the convolutional layer; σ is the activation function of the fully connected layer; c is the bias vector of the fully connected layer; n is the height of the convolution kernel; m is the width of the convolution kernel.

6. The cap beam intelligent welding cloud control system according to claim 1 is characterized in that: The method of generating a two-dimensional matrix according to the steel bar to be welded position and updating the parameters of the two-dimensional matrix based on the skip welding logic comprises the following steps: Generate a two-dimensional matrix through the target image output by the convolutional neural network, each element in the two-dimensional matrix represents the coordinates and state information of a point to be welded; The jump welding logic is applied to update the parameters of the two-dimensional matrix, including updating the parameters of the continuous points to be welded from the parameters to be welded to the parameters to be welded through the preset step size.

7. The cap beam intelligent welding cloud control system according to claim 3 is characterized in that: The method of controlling the welding robot to perform displacement and welding based on the verification result of the pre-welding verification module and according to the updated two-dimensional matrix comprises the following steps: Obtain the verification result of the pre-welding verification module; If the deviation value of the number of steel bars is 0 and no suspicious steel bars are found, the welding robot is controlled to move and weld according to the updated two-dimensional matrix; If the deviation value of the number of steel bars is not 0 or suspicious steel bars are found, the welding robot is controlled to stop displacement and welding.

8. The cap beam intelligent welding cloud control system according to claim 1 is characterized in that: Controlling the welding robot to perform displacement and welding comprises the following steps: According to the parameters to be welded in the two-dimensional matrix, the welding robot is controlled to move the X-axis and Y-axis to the position of the parameters to be welded and weld through the welding head; After each point to be welded is completed, the location is obtained and the two-dimensional matrix is ​​updated, and the parameters to be welded are updated to the welded parameters.

9. The cap beam intelligent welding cloud control system according to claim 1 is characterized in that: The method of judging welding abnormality according to working parameters comprises the following steps: Get the welding head depth, welding head pressure and sound parameters and the corresponding timestamp of a single welding when there is no abnormality in the history; Based on the same timestamp, the value ranges of welding head pressure and sound parameters at different welding head depths are obtained; If the welding head pressure or sound parameters in real-time welding work do not belong to their value range at the current welding head depth, it is judged as welding abnormality; If the welding head pressure or sound parameters do not belong to the value range at the current welding head depth during real-time welding work, it is judged that there is no abnormality.

Images