Busbar welding device

Abstract

The invention provides a busbar welding device and relates to the field of welding, the busbar welding device comprises a workbench, a welding mechanical arm and a conveying device set, and each conveying device comprises two sliding tables, a protection box and a smoke exhaust device; an updatable smoke concentration-pumping and discharging parameter mapping relation is pre-stored in the controller, a pre-trained welding state recognition model is integrated, and the controller is configured to generate splicing efficiency parameters corresponding to operators and update the splicing efficiency parameters; the earliest weldable moment of each sliding table is predicted in combination with sliding table position data, and the operation sequence of the welding mechanical arm is scheduled; identifying a welding stage based on the image data, and outputting a smoke concentration predicted value; dynamically adjusting operation parameters of the smoke exhaust device; according to the method, intelligent sensing and self-adaptive control are integrated, so that the welding efficiency of the busbar is improved, and the production energy consumption is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding, in particular to a busbar welding device. BACKGROUND

[0002] As a key connecting component in power transmission and distribution system, busbar is often composed of welding mechanical arm, conveying track and centralized smoke purification system to form an automatic welding workstation, which replaces manual operation to a certain extent and realizes the automation of basic functions. However, in the low-batch and multi-specification busbar welding scene, it is often necessary to alternate welding of multiple specifications on the same device. This process is still difficult to separate from manual operation in terms of busbar tool assembly, and the assembly efficiency and quality of manual assembly vary greatly, directly affecting the automation production efficiency of the welding device, which is prone to idle welding mechanical arms due to waiting for materials, and mismatching of conveying rhythm and welding rhythm, resulting in the need to improve the utilization rate of the device. At the same time, the smoke purification system is often continuously operated at a fixed power or simply linked to the arc ignition signal, and cannot sense the actual smoke concentration change, which easily leads to the inability to further reduce the energy consumption of the device. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a busbar welding device that integrates intelligent sensing and adaptive control to improve the welding efficiency of busbars and reduce production energy consumption.

[0004] To solve the above technical problems, the technical solution of the present application is as follows: A busbar welding device, comprising a workbench, comprising: a welding mechanical arm installed on the surface of the workbench for welding busbars; a conveying device group composed of a plurality of conveying device groups symmetrically distributed on both sides of the welding mechanical arm, each conveying device comprising two sliding tables for conveying between welding stations and assembly stations and carrying busbars fixed by tools; a protective box installed on the surface of the workbench for isolating high-temperature splashes, strong arc light and smoke; an exhaust smoke device installed on the surface of the protective box for exhausting and filtering the smoke generated in the protective box; a sensor group for collecting image data of the welding station, position data of the sliding table and state data of the busbar; a controller electrically connected to the welding mechanical arm, the drive of the conveying device, the exhaust smoke device and the sensor group, respectively; The controller pre-stores an updateable smoke concentration-exhaust parameter mapping relationship and integrates a pre-trained welding state recognition model, and is configured to: Generate and update the assembly efficiency parameter of the corresponding operator based on the operator's identity information and the parent row state data; predict the earliest weldable time of each slide based on the slide position data to schedule the moving order of the slide; Identify the welding stage based on the image data, and output the smoke concentration prediction value based on the morphological features, brightness features and smoke shielding features of the weld fusion zone; generate the target extraction parameter according to the smoke concentration-extraction parameter mapping relationship, and dynamically adjust the operation parameter of the smoke extraction device, so that the operation parameter real-time tracks the target extraction parameter; When the welding stage is identified as the end of welding and the slide meets the carry condition, the smoke extraction device is adjusted to the operation parameter for maintaining extraction; and based on the model information of the next to-be-welded parent row and the pre-set welding path, the smoke concentration-extraction parameter mapping relationship is dynamically corrected.

[0005] Further, the controller is configured to perform welding stage identification and smoke concentration prediction based on image data when: Analyze the image sequence of the welding station to extract the appearance features and smoke plume shielding features of the weld fusion zone, input the extracted feature vector into the pre-trained welding state recognition model, and output the current welding stage, smoke concentration prediction value and corresponding confidence; When the smoke concentration prediction value is higher than the preset safety concentration threshold, increase the extraction parameter to the upper limit of the target interval of the smoke concentration-extraction parameter mapping relationship; When the smoke concentration prediction value is lower than the preset safety concentration threshold, gradually reduce the extraction parameter to the corresponding target extraction parameter in the smoke concentration-extraction parameter mapping relationship; When the confidence is lower than the confidence threshold, use the extraction parameter higher than the corresponding set interval of the smoke concentration prediction value for control, and mark the corresponding data for subsequent online correction.

[0006] Further, the controller is configured to perform slide scheduling when: Based on the identity of the operator, generate a reference efficiency parameter by associating the historical data of the corresponding operator at the assembly station; calculate the efficiency decay factor according to the number of times the assembly position deviation exceeds the tolerance threshold in the current task cycle, combine the reference efficiency parameter and the efficiency decay factor to generate a dynamically updated assembly efficiency parameter; Based on the real-time position of the slide and combined with the pre-set motion performance parameters of the conveying device, calculate the shortest transmission time of the slide to reach the welding station, combine the assembly efficiency parameter to predict the remaining assembly time, take the larger value of the shortest transmission time and the remaining assembly time as the earliest weldable time, and use the earliest weldable time for the scheduling control of the slide.

[0007] Further, the controller is configured to perform slide position scheduling when: determining the welding advance window based on the remaining time of the current operation of the welding robot and the time required for the target skid located in the welding standby area to advance to the welding station; determining the assembly advance window based on the remaining assembly time determined based on the operator assembly efficiency parameter and the busbar state data and the estimated time required for the skid located in the assembly standby area to advance to the assembly station; When any of the advance windows meets the preset threshold and the corresponding target station is available, the controller controls the corresponding skid to advance to the target station within the corresponding advance window; When both types of advance windows meet and the corresponding target stations are available, the controller drives the two skids to advance to the corresponding target stations in parallel within the same control cycle; When only one side is allowed to advance in one control cycle, the controller selects the advancing side based on the criterion of shortening the idle time of the welding robot, and the other side remains standby until the next control cycle reevaluates the advance window.

[0008] Further, the conveying device comprises: A mounting plate is fixedly connected to the surface of the workbench, and the surface of the mounting plate is fixedly connected with a linear conveying table, the surface of the linear conveying table is fixedly connected with a C-shaped sliding arm, the inside of the C-shaped sliding arm is provided with a first lead screw, the first lead screw is rotatably connected with the C-shaped sliding arm through two first bearings, the surface of the first lead screw is threadedly connected with a connecting block, the surface of the connecting block is provided with a plurality of constraint rods, the constraint rods are slidably connected with the connecting block through first circular grooves formed on the surface of the connecting block, the constraint rods are fixedly connected with the C-shaped sliding arm, the constraint rods are arranged in an annular array around the central axis of the first lead screw, the skid is fixedly connected to the upper surface of the connecting block, and a first motor is installed on the surface of the C-shaped sliding arm.

[0009] Further, the controller is configured to: When it is determined based on the welding state recognition model that the welding stage is over and the target skid meets the advance condition, the smoke extraction device is switched from the target extraction parameter in the welding stage to the maintenance extraction operation parameter, and the maintenance extraction operation parameter is set to have a trapping effect in the protection box that is not lower than a trapping effect threshold to suppress smoke deposition; and based on the next to be welded busbar model information and its associated preset welding path, the smoke demand of the next welding cycle is predicted in combination with the material of the corresponding busbar model and the weld set characteristics, and the smoke concentration-extraction parameter mapping relationship is corrected online accordingly to generate the target extraction parameter for the next welding cycle.

[0010] Further, the controller is further configured to, when performing the slide table scheduling, continuously update the earliest weldable time of each slide table by using a rolling time window, and set a freezing time and a minimum holding time to avoid frequent rearrangement; when both sides simultaneously meet the carry condition and the target station is acceptable, introduce a stability weight for priority judgment on the basis of sorting by the earliest weldable time, and the stability weight is determined based on the assembly efficiency parameter fluctuation degree and the historical accuracy of the corresponding operator; When the deviation between the predicted earliest weldable time and the actual completion time exceeds a threshold value in a plurality of consecutive control periods, the assembly efficiency parameter of the corresponding operator is automatically converged and corrected, and a review prompt is triggered, so as to ensure the stability of the rhythm, reduce the loss of waiting, and improve the robustness of the scheduling.

[0011] Further, the controller is further configured to perform global energy efficiency optimization control: Based on the historical correction record of the smoke density-exhaust parameter mapping relationship, the update record of the assembly efficiency parameter, and the actual operation rhythm of the welding robot, an energy efficiency model with the optimization target of the lowest energy consumption per unit capacity is constructed; In a preset statistical period, based on the welding stage length distribution, the exhaust parameter energy consumption distribution and the slide table waiting time length proportion recorded in the corresponding period, the reference parameters of the smoke density-exhaust parameter mapping relationship, the decay factor weight of the assembly efficiency parameter and the preset threshold of the slide table carry window are optimized by the energy efficiency model; The optimized parameter group is applied to the next production batch, and based on the deviation between the real-time working condition data and the optimization target, the weight coefficient of the energy efficiency model is learned and adjusted online.

[0012] The above-mentioned scheme of the present application at least includes the following beneficial effects: The above-mentioned scheme of the present application realizes seamless connection of the welding and assembly processes by dynamically predicting the earliest weldable time of each station and intelligently scheduling the operation sequence of the welding robot, reduces the idle time of the equipment, and further improves the production efficiency; By welding state recognition and smoke density prediction based on vision, the exhaust and smoke extraction are driven to realize accurate on-demand management, which ensures the smoke capture effect while reducing the energy consumption of the device By online optimization of the assembly efficiency of different operators, and adaptive correction of the control parameters according to different busbar models, the stability, energy saving and production quality of the production are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is the overall structure schematic diagram provided by the present application.

[0014] Figure 2 is the slide table scheduling control flowchart in the present application.

[0015] Figure 3 is the schematic diagram of the sliding table in the present application.

[0016] Figure 4 is the schematic diagram of the C-shaped sliding arm in the present application.

[0017] Figure 5 is the schematic diagram of the connecting block in the present application.

[0018] In the figure: 101, workbench; 102, protection box; 103, smoke extraction device; 104, debugging table; 105, sliding table; 301, mounting plate; 302, linear conveying table; 303, C-shaped sliding arm; 304, first lead screw; 305, restraint rod; 306, connecting block; 307, first motor. DETAILED DESCRIPTION

[0019] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0020] As Figures 1 to 5 shown, the embodiment of the present application proposes a busbar welding device, which comprises a workbench 101, comprising: a welding robot arm installed on the surface of the workbench 101 for welding of busbars; a conveying device group composed of a plurality of conveying device groups symmetrically distributed on both sides of the welding robot arm, each conveying device comprising two sliding tables 105 for conveying between welding stations and assembly stations and carrying busbars fixed by tooling; a protection box 102 installed on the surface of the workbench 101 for isolating high-temperature spatter, strong arc light and smoke; a smoke extraction device 103 installed on the surface of the protection box 102 for extracting and filtering smoke generated in the protection box 102; a sensor group for collecting image data of the welding station, position data of the sliding table 105 and state data of the busbar; a controller electrically connected with the welding robot arm, the drive of the conveying device, the smoke extraction device 103 and the sensor group, respectively; the controller pre-stores an updateable smoke concentration-extraction parameter mapping relationship and integrates a pre-trained welding state recognition model, and is configured to: Based on the identity information of the operator and the busbar state data, the assembly efficiency parameter corresponding to the operator is generated and updated; combined with the position data of the sliding table 105, the earliest weldable time of each sliding table 105 is predicted to schedule the moving order of the sliding table; Based on the image data, the welding stage is identified, and based on the morphological features, brightness features and smoke shielding features of the weld fusion zone, the smoke concentration prediction value is output; according to the smoke concentration-exhaust parameter mapping relationship, the target exhaust parameter is generated, and the operating parameter of the exhaust smoke device 103 is dynamically adjusted, so that the operating parameter can track the target exhaust parameter in real time; When the welding stage is identified as the end of welding and the sliding table 105 meets the carry condition, the exhaust smoke device 103 is adjusted to the operating parameter of maintaining exhaust; and based on the model information of the next to-be-welded busbar and the pre-set welding path, the smoke concentration-exhaust parameter mapping relationship is dynamically corrected.

[0021] The mounting plate 301 is fixedly connected to the surface of the workbench 101, the surface of the mounting plate 301 is fixedly connected with a linear conveying table 302, the surface of the sliding block of the linear conveying table 302 is fixedly connected with a C-shaped sliding arm 303, the inside of the C-shaped sliding arm 303 is provided with a first lead screw 304, the first lead screw 304 is rotatably connected with the C-shaped sliding arm 303 through two first bearings, the surface of the first lead screw 304 is threadedly connected with a connecting block 306, the surface of the connecting block 306 is provided with a plurality of constraint rods 305, the constraint rods 305 are slidably connected with the connecting block 306 through a first circular groove formed in the surface of the connecting block 306, the constraint rods 305 are fixedly connected with the C-shaped sliding arm 303, the constraint rods 305 are arranged in an annular array around the central axis of the first lead screw 304, the sliding table 105 is fixedly connected to the upper surface of the connecting block 306, the surface of the C-shaped sliding arm 303 is provided with a first motor 307, and the output shaft of the first motor 307 is fixedly connected with the first lead screw 304 through a first through groove formed in the surface of the C-shaped sliding arm 303.

[0022] In the embodiment of the present application, the workbench 101 is integrally welded by a Q235 steel plate, which is used to fix the welding robot, the conveying device and the exhaust smoke device 103. The welding robot is a six-axis joint type robot, which is fixed on the surface of the workbench 101 through a special mounting base. The distal end of the welding robot is provided with a gas shielded welding gun. The welding gun is connected with the distal end of the robot through a quick-change clamp. This is known in the prior art and will not be described in detail here. The exhaust smoke device 103 is installed on the surface of the protection box 102 and includes a pipeline, a filter and a fan. The working principle and use process of the exhaust smoke device 103 are known in the prior art and will not be described in detail here. The suction port of the exhaust smoke device 103 extends to the inside of the protection box 102 through the pipeline and is arranged near the welding station. The rotating speed of the fan is adjusted by the controller through the frequency converter. The sensor group at least includes: an industrial camera installed above the welding station, the lens is equipped with a narrow-band filter and a controllable light source to suppress arc light saturation and stabilize the imaging of the molten zone; two encoders are arranged on the same conveying device, respectively installed on the output shaft of the driving motor of the linear conveying table 302 and the first motor 307 of the linear conveying table 302, to obtain the position and in-place state of the sliding table 105; The controller is an industrial PLC, which is internally equipped with a pre-trained welding state recognition model and stores an updateable smoke density-exhaust parameter mapping table, and stores process and path data associated with the busbar model.

[0023] The two sliding tables 105 on the conveying device are driven by corresponding linear conveying tables 302, specifically, the linear conveying table 302 drives the corresponding sliding block to move the C-shaped sliding arm 303, and the C-shaped sliding arm 303 moves to drive the sliding table 105 at the top of the connecting block 306 to move horizontally synchronously; As shown in Figure 1 Two conveying devices are arranged on both sides of the welding mechanical arm, and a debugging table 104 is further arranged on one side of the welding mechanical arm and fixed on the surface of the workbench 101; the debugging table 104 is used for sample printing and welding mechanical arm debugging; the two sliding tables 105 on the same conveying device are arranged on two linear conveying tables 302, and the position of the sliding table 105 inside the protection box 102 is a welding station, and the position of the sliding table 105 outside the protection box 102 is an assembly station; the upper surfaces of the sliding tables 105 corresponding to the two stations are coplanar; When the sliding table 105 at the welding station is driven by the first motor 307 on the surface of the corresponding C-shaped sliding arm 303 to rotate the first screw rod 304, the connecting block 306 drives the sliding table 105 to descend along the surface of the constraint rod 305, until the highest point of the busbar on the surface of the sliding table 105 is below the bottom surface of the sliding table 105 in the assembly station, at this time, the position of the sliding table 105 is a welding standby area; When the sliding table 105 at the assembly station is driven by the first motor 307 on the surface of the corresponding C-shaped sliding arm 303 to rotate the first screw rod 304, the connecting block 306 drives the sliding table 105 to descend along the surface of the constraint rod 305, until the highest point of the busbar on the surface of the sliding table 105 is below the bottom surface of the sliding table 105 in the welding station, at this time, the position of the sliding table 105 is an assembly standby area; When there is a sliding table 105 in the assembly standby area, the busbar assembly in the assembly station is completed, and the corresponding sliding table 105 directly advances to the welding station; when there is a sliding table 105 in the welding standby area, the sliding table 105 in the welding station directly advances to the assembly station; when there is no sliding table 105 in both standby areas, the sliding table 105 in the welding station first advances to the welding standby area, and then advances to the assembly standby area, and the sliding table 105 in the assembly station directly advances to the welding station; When two carriages 105 on the same conveyor move towards each other, one of the carriages 105 needs to be lowered to the same level as the upper surface corresponding to the standby area of the carriage 105; during the movement towards each other, the lower carriage 105 passes through the recess of the C-shaped slide arm 303 corresponding to the upper carriage 105 to avoid motion interference.

[0024] The operator completes identity confirmation through the identification device (such as by reading the ID card of the operator through the card reader), the controller calls the historical assembly record of the operator, and generates the reference assembly efficiency parameter in combination with the current busbar state data (model, piece number). During the assembly process, if the busbar state data reflects an increase in assembly deviation (such as the number of positioning deviation exceeding the tolerance), the controller calculates the efficiency correction factor and updates the assembly efficiency parameter of the corresponding operator online. According to the real-time position of the carriage 105 and the preset motion performance parameters of the conveyor (such as the maximum speed of the carriage 105 0.8 m / s, the maximum acceleration 3 m / s²), the controller calculates the predicted transmission time of the carriage 105 from the current state to the welding station, and combines it with the remaining assembly time predicted based on the assembly efficiency parameter, and takes the larger value as the earliest weldable time of the carriage 105; sort and schedule according to the earliest weldable time, and allow the non-welding side carriage 105 to advance to the predetermined standby position when the safety condition is met.

[0025] The controller pre-processes and extracts features from the image sequence collected by the industrial camera, and the pre-trained welding state recognition model outputs the welding stage, smoke density index and confidence. The controller generates the target extraction parameter according to the smoke density-extraction parameter mapping table, and executes real-time tracking through the frequency converter; when the smoke density is higher than the safety density threshold, it is increased to the upper limit of the target interval; when it is lower than the safety density threshold, it falls back to the target value of the mapping table according to the gradual strategy; when the confidence is lower than the confidence threshold, the conservative extraction higher than the density setting interval is used, and the related data is marked for subsequent correction of the mapping relationship and the model.

[0026] When the welding stage is identified as ending and the carry condition is met (such as the target station being idle), the controller switches the extraction parameter from the current target parameter to the maintenance extraction operation parameter, which ensures that the capture effect is not lower than the capture effect threshold to suppress the settlement; at the same time, based on the model information of the next to-be-welded busbar and the preset welding path, in combination with the material and weld set characteristics, the next cycle smoke demand is predicted, and the smoke density-extraction parameter mapping relationship is corrected online (such as increasing the target fan speed of the response interval), and the corrected mapping relationship is used for the next welding cycle.

[0027] The safety concentration threshold and the capture effect threshold are determined according to the production requirements of the enterprise, or are obtained through on-site calibration on the principle that the capture efficiency is not lower than the preset target; the safety concentration threshold and the capture effect threshold are distinguished and maintained according to the type of the busbar and the welding stage; In the embodiment, the pre-trained welding identification model specifically includes: The industrial camera continuously outputs a welding station image sequence to the controller, the controller runs the pre-trained welding state identification model, and outputs three results: Welding stage: one of four categories of arc striking, molten pool stabilization, arc collecting, and abnormality; Smoke concentration prediction value: expressed in mg / m³; Confidence: 0-1 The welding identification model uses an improved MobileNetV2 backbone and contains a double-branch output, specifically including: Input layer: image size 224x224x3 (three-channel color image input); Main body: 12 inverted residual convolution blocks (3x3 convolution, step 1, expansion rate 1, activation function ReLU6); Feature fusion: performing ROI Pooling on the ROI of the weld fusion zone, and splicing with the global feature channel of the main body; Classification branch: fully connected layer 128 nodes + Softmax, outputting four-class welding stage probability; Regression branch: fully connected layer 64 nodes + linear layer, outputting a smoke concentration prediction value in the range of 0-20 mg / m³; Confidence calculation: the confidence threshold of 0.85 is obtained by weighting the maximum probability of Softmax and the regression residual statistics.

[0028] The training process of the welding identification model includes: Take 10,000 groups of images, of which 2,000 are arc striking, 4,000 are molten pool stabilization, 2,000 are arc collecting, and 2,000 are abnormality (abnormality is subdivided into arc breaking 500, welding deviation 1,000, and excessive spatter 500).

[0029] Working condition coverage: busbar material copper, aluminum; typical cross-sectional size 40x4 to 100x10 mm; welding current 150-300 A; smoke plume shielding occupies 10%-60% of the image area.

[0030] Stage annotation criteria: Arc striking: the molten zone has not stabilized, and the arc light brightness fluctuation is greater than 30%; Arc collecting: the area and brightness of the molten zone continue to decrease and tend to disappear; Abnormality: arc breaking, continuous welding deviation (2-5 mm), and excessive spatter.

[0031] Concentration label acquisition: using a smoke monitor (sampling frequency greater than or equal to 1 Hz, measurement error less than or equal to ±0.5 mg / m³) to align with the image timestamp, to establish image-concentration pairs.

[0032] Data style: training set 70%, validation set 15%, test set 15%: data enhancement adopts random cropping, brightness scaling 0.8-1.2, Gaussian noise, and the enhancement ratio is 20% each.

[0033] Loss function: joint loss = cross-entropy + MSE (this embodiment adopts a proportion of cross-entropy weight 0.6 and MSE weight 0.4, the weight can be determined by testing different combinations on the validation set and selecting the proportion that comprehensively optimizes the classification and regression performance, which is an example value in this embodiment); Optimizer and learning rate: Adam, initial 1×10 -3 ; when the validation set loss does not decrease for 5 consecutive epochs, the learning rate is halved; weight decay 1×10 -4 ; Batch and round: Batch Size 32, maximum 100 epochs; Early stopping condition: stop training when the training set loss is less than 0.05 and the validation set loss is less than 0.08; Phase classification accuracy: when the classification accuracy of the welding recognition model on the independent test set reaches the preset performance threshold, it is determined that the recognition model performance meets the standard. In this embodiment, the performance threshold is set such that the classification accuracy of each welding phase is not less than the preset standard, and the average classification accuracy is not less than the overall threshold, thereby ensuring the reliability of the welding phase recognition.

[0034] When the controller performs welding phase recognition and smoke concentration prediction based on image data, it is configured to: analyze the image sequence of the welding station to extract the appearance features of the weld fusion zone and the smoke plume shielding features, input the extracted feature vector into the pre-trained welding state recognition model, and output the current welding phase, smoke concentration prediction value, and corresponding confidence; when the smoke concentration prediction value is higher than the preset safety concentration threshold, increase the extraction parameter to the upper limit of the target interval of the smoke concentration-extraction parameter mapping relationship; when the smoke concentration prediction value is lower than the preset safety concentration threshold, gradually reduce the extraction parameter to the corresponding target extraction parameter in the smoke concentration-extraction parameter mapping relationship; when the confidence is lower than the confidence threshold, use the extraction parameter higher than the set interval corresponding to the smoke concentration prediction value for control, and mark the corresponding data for subsequent online correction.

[0035] In the embodiment of the present application, the controller continuously receives the image sequence output by the camera, and processes each frame of image: removes the edge occlusion area, intercepts the effective color image block centered on the weld with a size of 224x224 pixels, performs brightness correction and denoising to stabilize the grayscale across frames. Then feature extraction is performed: on the preprocessed grayscale image, edge detection is performed by Canny operator to obtain the melting zone profile, and its pixel length, width and area are calculated as morphological features; the mean and variance of the grayscale of the pixels inside the melting zone are calculated as brightness features; the threshold segmentation method is used to separate the smoke area from the background, and the proportion of the pixel area of the smoke area to the total area of the ROI is calculated as the shielding feature.

[0036] The controller inputs the above engineered features and the 224x224x3 ROI color image into the pre-stored welding state recognition model, and a feature fusion layer is arranged at the rear of the feature extraction network of the welding recognition model. After the feature fusion layer completes splicing and fusion, the current welding stage (arc striking, melting, solidification, arc collecting), the smoke concentration prediction value and its corresponding confidence are output through the classification branch and the regression branch respectively; the single inference delay is not greater than 120ms, and the inference frame rate is not less than 10fps, so as to meet the timing requirements of implementation control. The controller queries the smoke concentration-exhaust parameter mapping relationship according to the output prediction value, and then issues control instructions to the exhaust device 103.

[0037] The safe concentration threshold is set according to the busbar model and the welding stage, and is stored in the threshold table of the controller; the confidence threshold is set to 0.85. The control logic is: When the smoke concentration is higher than the safe concentration threshold of the corresponding stage, the controller increases the exhaust parameter (such as the fan speed) to the upper limit of the target interval of the smoke concentration-exhaust parameter; when the prediction value is lower than the safe concentration threshold, the exhaust parameter is gradually reduced to the corresponding target value of the smoke concentration-exhaust parameter mapping table according to the gradual side rate; When the confidence is lower than the confidence threshold of 0.85, the exhaust parameter higher than the set interval corresponding to the prediction concentration is used for conservative control, and the image corresponding to this inference, the prediction value, the actual exhaust parameter and other working condition information are marked as low-confidence data and stored in the historical database for subsequent offline correction of the smoke concentration-exhaust parameter mapping relationship and the welding recognition model.

[0038] To ensure the effectiveness of the smoke concentration-exhaust parameter mapping relationship, the confidence threshold and the safe concentration threshold, a smoke detector is used to compare typical working conditions during debugging or periodic sampling, and the mapping relationship, the confidence threshold and the safe concentration threshold are corrected and versioned according to the comparison results.

[0039] The controller is configured to: The historical data of the corresponding operator at the assembling station is associated to generate a reference efficiency parameter based on the identity of the operator; and an efficiency decay factor is calculated according to the number of times that the assembly position deviation exceeds the tolerance threshold in the current task period, and the reference efficiency parameter is combined with the efficiency decay factor to generate a dynamically updated assembling efficiency parameter. Based on the real-time position of the sliding table 105 and in combination with the preset motion performance parameters of the conveying device, the shortest transmission time of the sliding table 105 to reach the welding station is calculated, the remaining assembling time is predicted in combination with the assembling efficiency parameter, the larger value of the shortest transmission time and the remaining assembling time is taken as the earliest weldable time, and the earliest weldable time is used for the scheduling control of the sliding table 105.

[0040] In the embodiment of the present application, the assembling station is equipped with an operator identity recognition device (such as an RFID card reader). The operator completes identity authentication when starting work. The controller retrieves the recent historical data (such as the record of the last 100 working days) of the operator according to the identity information, calculates the average single-piece assembling time of the same type busbar as the reference efficiency parameter of the operator, and if the operator has no historical data, a preset general reference value is used.

[0041] The sensor (such as a visual detection system) of the assembling station monitors the assembly position deviation at a fixed period. When the deviation exceeds the preset tolerance range, it is recorded as an out-of-tolerance event.

[0042] The controller applies a decay mechanism to the reference efficiency parameter according to the frequency of the out-of-tolerance event to generate an efficiency decay factor, and sets a lower limit constraint in the factor calculation to avoid excessive decay of the efficiency value. The final dynamic assembling efficiency parameter is determined by combining the reference efficiency parameter and the efficiency decay factor, and is continuously updated during the assembling process.

[0043] The controller determines the earliest weldable time of the sliding table 105 by comprehensively considering the transmission time and the remaining assembling time. The transmission time is calculated according to the distance between the current position of the sliding table 105 and the welding station, in combination with the shortest feasible transmission time calculated based on the pre-stored maximum speed and acceleration / deceleration performance parameters of the conveying device; and the remaining assembling time is estimated based on the dynamic assembling efficiency parameter and the current assembling progress. The controller takes the larger value of the two as the earliest weldable time of the sliding table 105, so as to ensure that the assembling operation and the transmission process are completed when the sliding table 105 reaches the welding station.

[0044] After the earliest weldable time is determined, the controller performs the scheduling decision of the sliding table 105: The controller compares the earliest weldable times of the two sliding tables 105, and selects the sliding table 105 with the earlier time as the priority scheduling object. Before scheduling, the controller verifies the receivable conditions of the welding station, including the idle state of the station and the workpiece in-position detection signal, and if all the conditions are met, the priority sliding table 105 is driven to advance to the welding station, and the non-priority sliding table 105 remains on standby until the next round of scheduling.

[0045] The controller further comprises an exception handling mechanism. When a failure of the skid 105 is detected, such as a sensor alarm or a transmission timeout, the controller automatically removes the skid 105 from the current dispatch queue and triggers an alarm prompt, while displaying the updated dispatch sequence on the device human-machine interface to support manual intervention by the operator.

[0046] The controller is configured to, when executing the position dispatch of the skid 105: determine a welding advance window based on the remaining time of the current work of the welding robot and the shortest time required for the target skid 105 located in the welding standby area to move to the welding station; determine an assembly advance window based on the remaining assembly time determined by the operator's assembly efficiency parameter and the busbar state data and the estimated time required for the skid 105 located in the assembly standby area to move to the assembly station; when any advance window meets the preset threshold and the corresponding target station is available, the controller controls the corresponding skid 105 to advance to the target station within the corresponding advance window; when both types of advance windows meet and the corresponding target stations are available, the controller drives two skids 105 to advance to the corresponding target stations in parallel within the same control cycle; when only one side is allowed to advance in one control cycle, the controller selects the advancing side based on the criterion of shortening the idle time of the welding robot, and the other side remains standby until the advance window is re-evaluated in the next control cycle.

[0047] In an embodiment of the present application, the welding advance window comprises: The controller determines based on the remaining time of the current work of the welding robot and the shortest time required for the target skid 105 located in the welding standby area to move to the welding station. The remaining time of the robot is calculated from the work progress feedback by its controller in real time; the skid 105 moving time is calculated from its current position, target position and preset motion performance parameters of the conveying device.

[0048] When the result of the welding robot's remaining time minus the skid 105 moving time is greater than or equal to the preset safety margin threshold, and the target welding station is in an available state (station idle), it is determined that the welding advance window meets.

[0049] The assembly advance window comprises: The controller estimates the remaining assembly time based on the dynamic assembly efficiency parameter of the operator and the remaining assembly workload, and determines the shortest time required for the skid 105 located in the assembly standby area to move to the assembly station. When the result of the remaining assembly time minus the skid 105 moving time is greater than or equal to the safety margin threshold, and the target assembly station is in an available state, it is determined that the assembly advance window meets.

[0050] The safety margin threshold is a configurable parameter (e.g. 0.3s) to compensate for system response delay and positioning error.

[0051] The controller evaluates the two types of carry windows and station status at each control cycle (e.g. 50ms) and executes according to the following strategies: Single-side window satisfied: when only one type of carry window is satisfied and the corresponding target station is available, the controller immediately issues a carry instruction to the slide table 105 on that side to the target station.

[0052] Double-side window satisfied: when both welding and assembly carry windows are satisfied and both sides of the station are available, the controller has the ability to drive both slide tables 105 in parallel. To ensure safety, the controller verifies that the two moving paths do not interfere with each other in space before driving. If the paths do not interfere, the controller issues a carry instruction to both slide tables 105 simultaneously; if there is a risk of interference, the controller drives the slide tables in time sequence.

[0053] When resource (e.g. power) constraints or safety policy constraints limit the controller to drive only one side of the slide table 105 to carry in the same cycle, the controller makes a selection according to the preset decision criteria.

[0054] The decision criteria include: predicting and comparing the possible waiting time of the welding robot after the two sides of the slide table 105 are ready, selecting the side with longer waiting time to carry first to minimize the overall idle time of the welding robot, and keeping the other side of the slide table 105 standby and re-evaluating in the next control cycle.

[0055] When a fault is detected in the slide table 105 (e.g. drive motor alarm, encoder signal timeout), the controller automatically removes the slide table 105 from the dispatch queue and triggers an audible and visual alarm on the human-machine interface. The real-time updated dispatch queue, carry window status, and fault code are displayed clearly on the human-machine interface, providing an entry for the operator to intervene and recover. After troubleshooting, the operator must confirm through the interface before re-adding the slide table 105 to the automatic dispatch sequence.

[0056] When the welding state recognition model determines that the welding phase is complete and the target slide table 105 meets the carry conditions, the smoke extraction device 103 switches from the target extraction parameters during the welding phase to the maintenance extraction operation parameters, which are set to maintain a capture effect in the protection box 102 that is not less than a capture effect threshold to suppress smoke deposition; and based on the next-to-be-welded busbar model information and its associated preset welding path, in combination with the material of the corresponding busbar model and the weld set characteristics, the smoke demand of the next welding cycle is predicted, and the smoke concentration-extraction parameter mapping relationship is corrected online accordingly to generate the target extraction parameters for the next welding cycle.

[0057] In the embodiment of the present application, the controller continuously receives the results output by the welding state recognition model. When the following two conditions are met simultaneously, the switching of the extraction parameters is triggered: The welding recognition model determines that the welding stage is over (the end features include the disappearance of the molten pool area, the extinguishing of the electric arc, and the continuous low brightness of thermal radiation below the set threshold value); The controller confirms that the skid 105 meets the conditions for leaving the welding station (i.e., the target station / target standby area is idle).

[0058] When the two conditions are met, the controller sends a control signal to the smoke extraction device 103 to drive the fan speed to smoothly transition from the target extraction parameter during welding to the preset maintenance extraction operation parameter.

[0059] The maintenance operation parameter is obtained through on-site calibration, and the principle is that under this air volume, there is no obvious smoke escaping at the opening of the protection box 102 (the top opening, the skid 105 import and export), and the smoke capture efficiency in the protection box 102 is not lower than the set capture effect threshold value, so as to suppress the settlement of smoke, and the switching adopts a ramp-down mode to avoid sudden changes in fan pressure.

[0060] After triggering the switching, the controller immediately starts to obtain the model information of the next welding mother rail, retrieves the typical welding parameters of the model from the process database, including material properties, total weld length, and welding current and speed. Combined with historical statistical data, the controller calculates a smoke generation index to represent the expected smoke level under this working condition.

[0061] The controller compares the predicted value with the baseline value of the existing smoke concentration-fan speed mapping relationship: If the predicted index is significantly lower than the baseline value, the target fan speed of the corresponding concentration interval is adjusted proportionally downward; If the predicted index is higher than the baseline value, the target fan speed is adjusted proportionally upward to ensure that the capture effect is not lower than the capture effect threshold value; The correction amplitude is set with an upper limit, such as not more than ±15% for single adjustment, to avoid excessive correction. The corrected mapping relationship version is saved in association with the mother rail model information.

[0062] In the subsequent welding process, the controller records the corresponding relationship between the fan operation parameters and the smoke concentration predicted by the welding recognition model. When a long-term deviation from the corrected value is detected, the controller can automatically initiate re-correction during the idle period of the equipment to gradually approach the optimal energy efficiency and capture balance point.

[0063] The controller is further configured to continuously update the earliest weldable time of each sliding table 105 by using a rolling time window, and set a freezing time and a minimum holding time to avoid frequent rearrangement; when both sides simultaneously meet the carry condition and the target workstations can be accommodated, introduce a stability weight for priority judgment based on the sorting of the earliest weldable time, and the stability weight is determined based on the fluctuation degree of the assembly efficiency parameter of the corresponding operator and the historical accuracy rate; When the deviation between the predicted earliest weldable time and the actual completion time exceeds a threshold value in a plurality of consecutive control periods, the assembly efficiency parameter of the corresponding operator is automatically converged and corrected, and a review prompt is triggered, so as to ensure the stability of the rhythm, reduce the loss of waiting, and improve the scheduling robustness.

[0064] In the embodiment of the application, the controller sets a rolling time window (such as 30s) with a configurable duration, and the data in the time window is used to calculate and update the earliest weldable time of each sliding table 105. In order to suppress the frequent change of scheduling instructions caused by slight fluctuations in data, the controller is configured to: For the earliest weldable time of a single sliding table 105, when the change amplitude of the calculation result does not exceed the set threshold value or does not reach the minimum update interval, the original value is maintained unchanged to avoid frequent changes.

[0065] Once the controller issues a scheduling instruction according to the sorting result, the scheduling instruction remains effective within a short freezing period and does not change due to slight fluctuations in the prediction result, unless a major event (such as a fault alarm or an emergency stop) occurs.

[0066] When both sides of the sliding table 105 simultaneously meet the carry condition and the target workstations can be accommodated, the controller introduces a stability weight for priority arbitration based on the comparison of the earliest weldable time.

[0067] The fluctuation amplitude of the assembly efficiency parameter of the operator in the recent assembly task is normalized to represent the stability of the assembly efficiency of the corresponding operator.

[0068] The accuracy index is obtained by counting the proportion of the deviation between the predicted earliest weldable time and the actual completion time within the allowable error range of the operator in the recent assembly task.

[0069] The stability weight is obtained by inversely weighting the fluctuation degree of the efficiency and the historical prediction accuracy rate, and the weighting coefficients are configurable parameters (such as when the stability weight is more biased towards punctuality, the weight of the accuracy index can be higher than that of the stability index). When the earliest weldable time of both sides of the sliding table 105 is small (such as 2s), the controller preferentially selects the side with higher stability weight to perform carry.

[0070] The controller continuously monitors the predicted performance of each operator. When the deviation between the predicted earliest weldable time and the actual completion time exceeds a set threshold for a continuous number of tasks, the controller triggers an automatic revision of the assembly efficiency parameter of the operator: The revised assembly efficiency parameter is obtained by proportionally weighted average of historical parameters and parameters calculated based on actual completion, where the weighting proportion is set by a convergence coefficient to balance the revision speed and stability. After revision, the controller triggers a prompt on the operation interface to inform the manager to review the status of the relevant workstations or equipment, but the controller will continue to run directly using the revised assembly efficiency parameter.

[0071] The controller is also configured to perform global energy efficiency optimization control: Based on the historical revision record of the smoke density-exhaust parameter mapping relationship, the update record of the assembly efficiency parameter, and the actual operation cycle of the welding robot, an energy efficiency model is constructed with the optimization goal of minimizing energy consumption per unit capacity. Within a preset statistical period, based on the recorded welding phase duration distribution, exhaust parameter energy consumption distribution, and slide table 105 idle time proportion within the corresponding period, the baseline parameters of the smoke density-exhaust parameter mapping relationship, the decay factor weight of the assembly efficiency parameter, and the preset threshold of the slide table 105 carry-in window are optimized by the energy efficiency model. The optimized parameter set is applied to the next production batch, and based on the deviation of real-time working condition data and optimization target, the weight coefficient of the energy efficiency model is learned and adjusted online.

[0072] In the embodiment of the present application, the controller is configured to perform global energy efficiency optimization control. The controller self-tunes and learns key control parameters based on historical data and real-time data, with the optimization goal of minimizing energy consumption per unit capacity. Energy consumption per unit capacity is defined as the total energy consumed to produce each meter of weld or each product in this embodiment.

[0073] The energy efficiency model constructed by the controller is used to quantify the influence of different control parameters on total energy consumption, with multiple evaluation indicators as input: Exhaust efficiency indicator: derived from the baseline parameters of the smoke density-exhaust parameter mapping relationship, which can be represented by the weighted average of the baseline air volume in each concentration interval.

[0074] Assembly efficiency indicator: derived from the assembly efficiency parameters of each operator, usually high efficiency, indicator value month, can take the average level of all operator assembly efficiency parameters or its derivative.

[0075] Idle loss indicator: calculated from the idle time proportion of the slide table 105, the larger the proportion, the higher the indicator value.

[0076] The energy efficiency model also contains a set of weight coefficients for representing the relative influence of each index on the energy efficiency target, and the initial values can be set according to experience and then corrected through learning.

[0077] The data required for constructing the energy efficiency model is uniformly collected by the controller: Different versions of the smoke concentration-exhaust parameter mapping table and their applicable time periods are retrieved from the database; the exhaust energy consumption data are obtained through the fan speed, power or energy consumption interface feedback of the frequency converter; the assembly efficiency parameter update record is retrieved from the database; and the welding robot operation cycle and the air stand 105 idle time are obtained through the communication interface with the welding robot controller and the transmission device and calculated in the controller.

[0078] At the end of each preset statistical period (such as a shift), the controller performs parameter reverse optimization, specifically including: The actual unit energy consumption of the period is counted, and the values of the exhaust energy efficiency index, the assembly efficiency index and the idle loss index are calculated.

[0079] The controller corrects the weight coefficients of the energy efficiency model using a data-driven method (such as a linear regression algorithm), and calculates new control parameters based on this, including the baseline parameters of the smoke concentration-exhaust parameter mapping relationship, the decay factor weight of the assembly efficiency and the preset threshold of the air stand 105 carry-in window.

[0080] The optimized parameters must meet the following conditions: the smoke capture effect after adjusting the exhaust check parameters is still not lower than the set threshold; and the adjustment of the carry-in window threshold cannot damage the safety interlocking logic and path anti-collision requirements.

[0081] The verified parameters form a new version, which is used in the next production batch.

[0082] During the operation of the new production batch, the controller continuously monitors the actual unit energy consumption and compares it with the predicted value of the energy efficiency model: If the deviation is small and stable, the current parameters and weights are kept unchanged; If the deviation is large and persistent, the controller starts online learning, uses adaptive filtering or recursive updating method to fine-tune the weight coefficients of the energy efficiency model, so that the prediction result gradually converges to the actual value; During the learning and adjustment process, if it is found that the modification of a certain parameter leads to a decrease in capture effect or cycle stability, the controller will immediately revert to the previous safe version, and the event will be recorded for manual review.

[0083] The above describes the preferred embodiments of the present application. It should be noted that for ordinary skilled persons in the technical field, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the protection scope of the present application.

Claims

1. A busbar welding device, comprising a workbench, characterized in that, include: A welding robotic arm, mounted on the surface of the workbench, is used for welding busbars; The conveying device group consists of several sets of conveying devices symmetrically distributed on both sides of the welding robot arm. Each conveying device includes two slides for conveying between the welding station and the assembly station and for carrying the busbar fixed by the tooling. A protective enclosure is installed on the surface of the workbench to isolate high-temperature splashes, strong arc light, and smoke. A smoke extraction device is installed on the surface of the protective box to extract and filter the smoke and dust generated inside the protective box. The sensor array is used to collect image data of the welding station, position data of the slide table, and status data of the busbar. The controller is electrically connected to the welding robotic arm, the driver of the conveying device, the fume extraction device, and the sensor group, respectively. The controller has a pre-stored, updatable mapping relationship between fume concentration and extraction parameters, and integrates a pre-trained welding state recognition model, configured as follows: Based on the operator's identity information and busbar status data, the assembly efficiency parameters of the corresponding operator are generated and updated; combined with the slide table position data, the earliest welding time of each slide table is predicted in order to schedule the movement sequence of the slide tables. Based on the image data, the welding stage is identified, and the predicted value of smoke concentration is output based on the morphological features, brightness features and smoke occlusion features of the weld molten zone; the target extraction parameters are generated according to the smoke concentration-extraction parameter mapping relationship, and the operating parameters of the smoke extraction device are dynamically adjusted so that the operating parameters track the target extraction parameters in real time. When the welding stage is identified as the end of welding and the slide meets the advance conditions, the fume extraction device is adjusted to maintain the operating parameters for extraction; and based on the model information of the next busbar to be welded and the preset welding path, the fume concentration-extraction parameter mapping relationship is dynamically corrected.

2. The busbar welding device according to claim 1, characterized in that, When performing welding stage identification and fume concentration prediction based on image data, the controller is configured to: The image sequence of the welding station is analyzed to extract the appearance features of the weld molten zone and the smoke plume obscuration features. The extracted feature vectors are input into a pre-trained welding state recognition model, which outputs the current welding stage, the predicted value of the smoke concentration and the corresponding confidence level. When the predicted smoke concentration is higher than the preset safe concentration threshold, the extraction parameters are increased to the upper limit of the target range of the smoke concentration-extraction parameter mapping relationship. When the predicted smoke and dust concentration is lower than the preset safe concentration threshold, the extraction parameters are gradually reduced to the target extraction parameters corresponding to the smoke and dust concentration-extraction parameter mapping relationship. When the confidence level is lower than the confidence level threshold, the extraction parameters are controlled by a set range that is higher than the predicted dust concentration value, and the corresponding data is marked for subsequent online correction.

3. The busbar welding device according to claim 1, characterized in that, When performing slide table scheduling, the controller is configured as follows: Based on the operator's identity, the baseline efficiency parameter is generated by associating the historical data of the corresponding operator at the assembly station; based on the number of times the assembly position deviation exceeds the tolerance threshold within the current task cycle, the efficiency decay factor is calculated, and the baseline efficiency parameter is combined with the efficiency decay factor to generate dynamically updated assembly efficiency parameters. Based on the real-time position of the slide table and the preset motion performance parameters of the conveying device, the shortest transmission time for the slide table to reach the welding station is calculated. The remaining assembly time is predicted by combining the assembly efficiency parameters. The larger value between the shortest transmission time and the remaining assembly time is taken as the earliest weldable time, and the earliest weldable time is used for the scheduling and control of the slide table.

4. The busbar welding device according to claim 3, characterized in that, The controller is configured to perform slide position scheduling as follows: The welding advance window is determined based on the remaining time of the current operation of the welding robot arm and the time required for the target slide in the welding standby area to advance to the welding station. The assembly advance window is determined based on the remaining assembly time determined by the operator's assembly efficiency parameters and busbar status data, and the estimated time required for the slide table located in the assembly standby area to advance to the assembly station. When any carry window meets the preset threshold and the corresponding target station can accept it, the controller controls the corresponding slide to carry to the target station within the corresponding carry window; When both types of carry windows are satisfied and the corresponding target workstations can accept them, the controller drives the two slides to carry to the corresponding target workstations in parallel within the same control cycle. When only one side is allowed to move within a control cycle, the controller selects the moving side based on the criterion of shortening the idle time of the welding robot arm, while the other side remains on standby until the next control cycle when the move window is re-evaluated.

5. The busbar welding device according to claim 1, characterized in that, The conveying device includes: Mounting plate (301) is fixedly connected to the surface of workbench (101). Linear conveyor (302) is symmetrically fixedly connected to the surface of mounting plate (301). C-shaped sliding arm (303) is fixedly connected to the surface of linear conveyor (302). A first lead screw (304) is provided inside the C-shaped sliding arm (303). The first lead screw (304) is rotatably connected to the C-shaped sliding arm (303) through two first bearings. A connecting block (306) is threadedly connected to the surface of the first lead screw (304). A plurality of constraint rods (305) are provided on the surface of the connecting block (306). 305) The first circular groove on the surface of the connecting block (306) is slidably connected to the connecting block (306). The constraint rod (305) is fixedly connected to the C-shaped sliding arm (303). The constraint rod (305) is arranged in a ring array around the central axis of the first lead screw (304). The slide table (105) is fixedly connected to the upper surface of the connecting block (306). The surface of the C-shaped sliding arm (303) is equipped with a first motor (307). The output shaft of the first motor (307) is fixedly connected to the first lead screw (304) through the first through groove on the surface of the C-shaped sliding arm (303).

6. The busbar welding device according to claim 1, characterized in that, The controller is configured to: When the welding stage is determined to be over based on the welding state recognition model and the target slide meets the advance condition, the fume extraction device is switched from the target extraction parameter during the welding period to the operating parameter for maintaining extraction. The operating parameter for maintaining extraction is set so that the collection effect in the protective box is not lower than the collection effect threshold in order to suppress the settling of smoke and dust. Based on the next busbar model information to be welded and its associated preset welding path, the dust demand for the next welding cycle is predicted by combining the material and weld set characteristics of the corresponding busbar model. Accordingly, the dust concentration-extraction parameter mapping relationship is corrected online to generate target extraction parameters for the next welding cycle.

7. The busbar welding device according to claim 4, characterized in that, When the controller performs slide scheduling, it is also configured to: continuously update the earliest weldable time of each slide using a rolling time window, and set a freeze time and a minimum hold time to avoid frequent rescheduling; when both sides meet the carry conditions and the target station can accept the slide, a stability weight is introduced to determine the priority based on the earliest weldable time sorting, and the stability weight is determined based on the fluctuation of the assembly efficiency parameter of the corresponding operator and the historical accuracy. When the deviation between the predicted earliest welding time and the actual completion time exceeds the threshold within multiple consecutive control cycles, the assembly efficiency parameters of the corresponding operator are automatically converged and corrected, and a review prompt is triggered to ensure cycle stability while reducing idle waiting losses and improving scheduling robustness.

8. The busbar welding device according to claim 7, characterized in that, The controller is also configured to perform global energy efficiency optimization control: Based on the historical correction records of the dust concentration-extraction parameter mapping relationship, the update records of the assembly efficiency parameters, and the actual working cycle of the welding robot, an energy efficiency model is constructed with the goal of minimizing energy consumption per unit of production capacity. Within a preset statistical period, based on the welding stage duration distribution, extraction parameter energy consumption distribution, and slide table idle time ratio recorded within the corresponding period, the baseline parameters of the dust concentration-extraction parameter mapping relationship, the attenuation factor weight of the assembly efficiency parameter, and the preset threshold of the slide table carry window are optimized in reverse through the energy efficiency model. The optimized parameter set is applied to the next production batch, and the weight coefficients of the energy efficiency model are learned and adjusted online based on the deviation between real-time operating data and the optimization target.

Images