Coordinated optimization analysis method for multi-station tower welding

Abstract

The present application relates to the technical field of process optimization, in particular to a kind of collaborative optimization analysis method of multi-station tower cylinder welding.The present application is aimed at the problem that tower cylinder multi-station welding is difficult to be cooperatively controlled, first loads the multimodal perception data of weld visual, arc sensing, heat field distribution, welding voiceprint, constructs fusion evaluation model to obtain weld deviation and sensor confidence score, determines sensor health state according to threshold value, triggers corresponding fault-tolerant reconstruction control mode to obtain welding tracking control parameter, then loads multi-station welding heat source model, decouples heat interference in combination with control parameter and heat field data, cooperatively controls multi-station, outputs and executes collaborative control instruction, realizes welding process optimization.

Description

Technical Field

[0001] This invention relates to the field of process optimization technology, and in particular to a collaborative optimization analysis method for multi-station tower welding. Background Technology

[0002] In the multi-station welding process of large wind turbine towers, the large size of the workpiece, the severe heat accumulation due to thick walls and multiple weld layers, and the simultaneous operation of multiple welding torches leading to the superposition of heat fields, easily cause welding deformation and stress concentration. Existing technologies mainly rely on a single vision or arc sensor for trajectory tracking. For example, patent document CN121361094A discloses a confidence-weighted vision-arc fusion method, but it only addresses trajectory correction for a single robot and does not consider thermal coupling interference between multiple stations. Patent document CN103341685B proposes a fusion strategy that switches between magnetically controlled arc and vision based on weld curvature, improving the accuracy of complex trajectory tracking. However, its fusion decision depends on the geometric features of the workpiece, lacks a fault-tolerant mechanism for sensor failure, and does not involve the coordinated control of multiple heat sources.

[0003] Furthermore, at the tower welding site, intense arc light and fumes often cause sensor signal loss, and existing methods are prone to downtime or welding misalignment because they cannot dynamically reconstruct the control mode. Therefore, there is an urgent need for a collaborative control method that can integrate multimodal information, has sensor failure tolerance, and can decouple thermal interference from multiple workstations. Summary of the Invention

[0004] This invention addresses the problems in existing technologies, such as the inability to integrate multimodal information, lack of sensor failure tolerance, and inability to decouple multi-station thermal interference, by providing a collaborative optimization analysis method for multi-station tower welding.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a collaborative optimization analysis method for multi-station tower welding, comprising:

[0007] S100: Load multi-modal perception data of tower welding in multi-station welding scenarios. The multi-modal perception data of tower welding includes weld visual perception data, arc sensing perception data, thermal field distribution perception data and welding acoustic perception data.

[0008] S200: Construct a multimodal fusion evaluation model, input the multimodal sensing data of tower welding into the multimodal fusion evaluation model, obtain the fusion weld deviation and several sensor confidence scores, and the several sensor confidence scores correspond to the multimodal sensing data of tower welding.

[0009] S300: Based on the confidence scores of several sensors, compare them with the preset confidence warning threshold and the preset confidence failure threshold respectively, dynamically determine the health status of each sensor, and obtain the sensor health status set;

[0010] S400: Based on the sensor health status set and the fused weld deviation, trigger the corresponding fault-tolerant reconfiguration control mode to obtain the reconfigured welding tracking control parameters;

[0011] S500: Loads a multi-station welding heat source model, and based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, decouples and coordinates the thermal interference between multiple stations, obtains multi-station collaborative control commands, and executes collaborative control of the tower welding process.

[0012] Optionally, load multimodal sensing data of tower welding in multi-station welding scenarios, including:

[0013] In the scenario of tower welding, the weld groove edge contour data, groove center point deviation data and groove width data collected by the laser vision sensor are added to the weld visual perception data.

[0014] In the scenario of tower welding, the instantaneous value of welding current, instantaneous value of welding voltage, standard deviation of current, standard deviation of voltage, and short-circuit frequency data collected by the arc sensor are added to the arc sensor sensing data.

[0015] In the scenario of tower welding, the infrared thermal imager collects the center temperature data of the molten pool, the width data of the molten pool, and the temperature gradient data of the heat-affected zone, and adds the thermal field distribution sensing data.

[0016] Acquire the frequency band sound pressure level data of the welding process collected by the microphone array in the tower welding scenario, and add it to the welding acoustic perception data.

[0017] Optionally, a multimodal fusion evaluation model is constructed, which previously included:

[0018] Collect historical welding process datasets for a preset tower model. The historical welding process datasets include historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, historical welding acoustic perception data, and the corresponding historical actual weld deviation.

[0019] Using historical real weld deviation as supervision, and taking historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, and historical welding acoustic perception data as inputs, and using deviation output nodes and the confidence scores of each sensor as outputs, an initial deep confidence network model is trained to obtain a multimodal fusion evaluation model whose output layer includes output nodes.

[0020] Optionally, based on the confidence scores of several sensors, the health status of each sensor is dynamically determined by comparing them with preset confidence warning thresholds and preset confidence failure thresholds, thereby obtaining a sensor health status set, including:

[0021] The confidence score of any sensor is compared one by one with the preset confidence warning threshold and the preset confidence failure threshold.

[0022] When the confidence score of a sensor is greater than or equal to the preset confidence warning threshold, the corresponding sensor is determined to be in a healthy state.

[0023] When the sensor confidence score is less than the preset confidence warning threshold but greater than or equal to the preset confidence failure threshold, the corresponding sensor is determined to be in an interference state.

[0024] When the confidence score of a sensor is less than the preset confidence failure threshold, the corresponding sensor is determined to be in a failure state.

[0025] The results of the health status, interference status, and failure status of all sensors are combined to form a sensor health status set.

[0026] Specifically, based on the sensor health status set and the fused weld deviation, a corresponding fault-tolerant reconfiguration control mode is triggered to obtain the reconfigured welding tracking control parameters, including:

[0027] When all sensors in the sensor health status set are determined to be in a healthy state, the full fusion tracking control mode is triggered, and the fusion weld deviation is directly converted into welding tracking control parameters.

[0028] When the sensor health status is determined to be in an interference state by the visual sensor, and the arc sensor, thermal imager, and acoustic sensor are not determined to be in a failure state, the downgraded fusion tracking control mode is triggered. The weight of the weld visual perception data in the fusion calculation is reduced, while the weight of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data in the fusion calculation is increased. Based on the adjusted weights, the downgraded fusion weld deviation is recalculated and converted into welding tracking control parameters.

[0029] When the sensor health status is determined to be in a failed state by the visual sensor, the vision-free guidance control mode is triggered. Historical weld path trajectory data is loaded, and the weld position is estimated by combining the instantaneous values ​​of welding current, welding voltage, and short-circuit frequency data in the arc sensor data. The estimated weld deviation is then converted into welding tracking control parameters.

[0030] Optionally, a multi-station welding heat source model is loaded. Based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, the thermal interference between multiple stations is decoupled and coordinated to obtain multi-station coordinated control commands, including:

[0031] Input the real-time welding power data and real-time welding torch position data of each station into the multi-station welding heat source model to obtain the prediction data of the overall transient thermal field distribution of the tower.

[0032] When the temperature data of the molten pool at the first station exceeds the preset temperature anomaly threshold, the abnormal temperature data of the molten pool at the first station is input into the multi-station welding heat source model to back-infer the heat source and obtain the heat interference source identification result.

[0033] When the thermal interference source identification result determines that the abnormal temperature of the molten pool originates from the thermal radiation interference of the second station, the cloud-based collaborative controller sends a thermal shielding command to the second station. The thermal shielding command includes welding torch angle adjustment parameters and local cooling activation parameters.

[0034] The cloud-based collaborative controller performs unified calibration of the arc start and end points of the circumferential welds at multiple workstations based on the fusion weld deviation at each workstation at the current moment, thereby obtaining joint consistency calibration parameters.

[0035] By merging the thermal shielding command and the connector consistency calibration parameters, a multi-station collaborative control command is obtained.

[0036] Optionally, loading a multi-station welding heat source model also includes:

[0037] Collect historical welding power data, historical welding torch position coordinates, and corresponding historical tower thermal imaging data for each workstation to construct a heat source model training set.

[0038] Based on the double ellipsoidal heat source distribution function, an initial heat source model including a heat source superposition factor is constructed, where the heat source superposition factor is used to characterize the energy coupling effect of heat sources in adjacent workstations.

[0039] Using historical tower thermal imaging data as supervision, and historical welding power data and historical welding torch position coordinates as input, the parameters of the initial heat source model are identified to obtain a multi-station welding heat source model.

[0040] Optional, also includes:

[0041] When the confidence scores of several sensors are all less than the preset confidence failure threshold, the sensor health status set is determined to be in a fully failed state, triggering the emergency shutdown protection mode, generating a shutdown alarm command and terminating the welding process.

[0042] Optional, also includes:

[0043] The historical welding process of tower welding was used to obtain the welding quality false alarm rate due to sensor failure and the welding quality false alarm rate due to environmental interference.

[0044] Calculate the ratio of the welding quality false alarm rate to the welding quality false alarm rate, and set it as the confidence threshold adjustment factor;

[0045] Based on the ratio of the preset baseline adjustment factor to the confidence threshold adjustment factor, the preset confidence warning threshold and the preset confidence failure threshold are adjusted.

[0046] Optional, also includes:

[0047] To obtain real-time welding progress data and real-time weld height detection data for each workstation in the tower welding scenario;

[0048] When the real-time weld reinforcement detection data exceeds the preset reinforcement height threshold, the corresponding location is identified as a repair welding area;

[0049] Based on real-time welding progress data, adjacent workstations that are idle or ahead in welding progress are scheduled to generate repair welding coordination instructions and perform online repair welding operations on the repair welding area.

[0050] By implementing this invention, it is possible to load multimodal sensing data of tower welding in multi-station welding scenarios. The multimodal sensing data of tower welding includes weld visual sensing data, arc sensing data, thermal field distribution sensing data, and welding acoustic signature sensing data. This overcomes the information limitations of single-type sensing data. Multimodal data can reflect the real state of the welding process from different dimensions, avoiding analytical biases caused by the one-sidedness of single-source data. At the same time, the data collected for multi-station tower welding scenarios is consistent with the working conditions of actual industrial production, providing a matching data source for subsequent multi-station collaborative control.

[0051] By implementing this invention, a multimodal fusion evaluation model can be constructed. Multimodal sensing data of tower welding is input into the multimodal fusion evaluation model to obtain the fused weld deviation and confidence scores of several sensors. These confidence scores correspond to the multimodal sensing data of tower welding. Compared to deviation detection by a single sensor, the fused weld deviation combines multi-dimensional data features, resulting in higher detection accuracy and a more realistic reflection of the actual weld deviation in tower welding. Simultaneously, it enables quantitative determination of the sensor's own operating status, overcoming the shortcoming of existing technologies lacking sensor failure assessment capabilities and providing a quantitative basis for subsequent fault-tolerant control.

[0052] By implementing this invention, it is possible to dynamically determine the health status of each sensor by comparing its confidence scores with preset confidence warning thresholds and preset confidence failure thresholds, thereby obtaining a set of sensor health statuses. This enables refined and dynamic determination of sensor statuses, rather than a simple binary judgment of effective / ineffectiveness, and can accurately identify intermediate states of sensors affected by environmental interference such as arc light and smoke. The resulting set of health statuses provides a clear and explicit basis for the accurate triggering of subsequent fault-tolerant reconfiguration control modes, avoiding false or missed triggering of fault-tolerant control.

[0053] By implementing this invention, it is possible to trigger a corresponding fault-tolerant reconfiguration control mode based on the sensor health status set and the fused weld deviation, and obtain the reconfigured welding tracking control parameters; it has fault tolerance capability for sensor failure and interference, and when the sensor is affected by environmental interference or failure, it can ensure the continuity of welding tracking control without stopping the machine, by dynamically reconfiguring the control mode, adjusting the data fusion weight, or calculating the weld position; it can also customize the generation of control parameters for different sensor failure scenarios, ensuring that welding trajectory tracking can still maintain high accuracy under non-ideal sensing conditions, and avoiding weld deviation problems caused by sensor signal loss.

[0054] By implementing this invention, it is possible to load a multi-station welding heat source model, and based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, decouple and coordinate the thermal interference between multiple stations, obtain multi-station collaborative control commands, and execute collaborative control of the tower welding process. This overcomes the shortcomings of existing technologies that do not consider multi-station thermal coupling interference, accurately identifies thermal interference sources, and decouples thermal interference through thermal shielding, parameter adjustment, etc., effectively avoiding welding deformation and stress concentration caused by the superposition of thermal fields from multiple welding torches. Furthermore, it allows for unified calibration of the arc starting and ending points of multiple stations, and simultaneous scheduling of stations for online supplementary welding, achieving global collaborative control of multi-station welding and improving the consistency and quality stability of the overall tower welding.

[0055] In summary, by implementing this invention, global collaborative control of the multi-station welding process of the tower can be achieved, effectively solving the problems of welding deformation and stress concentration caused by thermal field superposition in multi-station welding of the tower, as well as the problems of shutdown and weld deviation caused by sensor signal loss due to arc light and dust. At the same time, it improves the accuracy of weld trajectory tracking and the quality consistency of multi-station welding, ensuring the continuity, stability and efficiency of the tower welding process, and significantly improving the welding quality and production efficiency of large wind turbine towers. Attached Figure Description

[0056] Figure 1 A flowchart illustrating a collaborative optimization analysis method for multi-station tower welding provided by this invention;

[0057] Figure 2This is a flowchart illustrating the process of obtaining multi-station collaborative control commands in a collaborative optimization analysis method for multi-station tower welding provided by the present invention. Detailed Implementation

[0058] like Figure 1 As shown, this embodiment of the invention provides a collaborative optimization analysis method for multi-station tower welding, including:

[0059] S100: Load multi-modal perception data of tower welding in multi-station welding scenarios. The multi-modal perception data of tower welding includes weld visual perception data, arc sensing perception data, thermal field distribution perception data and welding acoustic perception data.

[0060] S200: Construct a multimodal fusion evaluation model, input the multimodal sensing data of tower welding into the multimodal fusion evaluation model, obtain the fusion weld deviation and several sensor confidence scores, and the several sensor confidence scores correspond to the multimodal sensing data of tower welding.

[0061] S300: Based on the confidence scores of several sensors, compare them with the preset confidence warning threshold and the preset confidence failure threshold respectively, dynamically determine the health status of each sensor, and obtain the sensor health status set;

[0062] S400: Based on the sensor health status set and the fused weld deviation, trigger the corresponding fault-tolerant reconfiguration control mode to obtain the reconfigured welding tracking control parameters;

[0063] S500: Loads a multi-station welding heat source model, and based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, decouples and coordinates the thermal interference between multiple stations, obtains multi-station collaborative control commands, and executes collaborative control of the tower welding process.

[0064] In step S100 of this application embodiment, loading multimodal sensing data of tower welding in a multi-station welding scenario includes:

[0065] In the scenario of tower welding, the weld groove edge contour data, groove center point deviation data and groove width data collected by the laser vision sensor are added to the weld visual perception data.

[0066] In the scenario of tower welding, the instantaneous value of welding current, instantaneous value of welding voltage, standard deviation of current, standard deviation of voltage, and short-circuit frequency data collected by the arc sensor are added to the arc sensor sensing data.

[0067] In the scenario of tower welding, the infrared thermal imager collects the center temperature data of the molten pool, the width data of the molten pool, and the temperature gradient data of the heat-affected zone, and adds the thermal field distribution sensing data.

[0068] Acquire the frequency band sound pressure level data of the welding process collected by the microphone array in the tower welding scenario, and add it to the welding acoustic perception data.

[0069] In step S100 of this application embodiment, the purpose of the above steps is to provide comprehensive and multi-dimensional raw input data for the subsequent construction of the multi-modal fusion evaluation model, realize multi-dimensional perception of weld status, welding process thermal field and sensor working status, lay the data foundation for subsequent fusion weld deviation calculation, sensor health status determination, fault-tolerant reconfiguration control mode triggering and multi-station thermal interference decoupling and collaborative regulation, solve the problem that single sensor data in tower welding site is easily affected by arc light and dust interference, resulting in signal loss, and ensure the accuracy and comprehensiveness of multi-station welding collaborative optimization analysis.

[0070] To achieve the above objectives, it is first necessary to acquire the weld groove edge contour data, groove center point deviation data, and groove width data collected by the laser vision sensor in the tower welding scenario, and add them to the weld visual perception data.

[0071] The purpose of this step is to collect geometric feature data related to the weld bevel using a laser vision sensor, which intuitively reflects the actual position and bevel shape of the weld, provides core visual data for calculating weld deviation, and achieves accurate perception of the weld geometry. The collected data is then categorized into weld visual perception data, and the integration of this type of perception data is completed. For example, if the edge contour of a tower weld bevel is a smooth V-shaped contour, the deviation of the bevel center point from the standard position is 0.2 mm, and the bevel width is 15 mm, all of these data are added to the weld visual perception data.

[0072] Next, the instantaneous values ​​of welding current, instantaneous values ​​of welding voltage, standard deviation of current, standard deviation of voltage, and short-circuit frequency data collected by the arc sensor in the tower welding scenario are acquired and added to the arc sensor sensing data.

[0073] The purpose of this step is to collect electrical parameter data related to the welding arc through an arc sensor, reflecting the stability of the arc and the welding energy input state during the welding process. This provides key arc-dimensional data for subsequent fusion evaluation and weld position estimation, enabling real-time perception of the welding arc's working state. The collected data is then categorized into arc sensor data, completing the integration of this type of perception data. For example, if the instantaneous value of the welding current is 320A, the instantaneous value of the welding voltage is 30V, the current standard deviation is 5A, the voltage standard deviation is 1V, and the short-circuit frequency is 15Hz, all of these data are added to the arc sensor data.

[0074] Then, the infrared thermal imager collects the molten pool center temperature data, molten pool width data, and heat-affected zone temperature gradient data in the tower welding scenario, and adds them to the thermal field distribution sensing data.

[0075] The purpose of this step is to collect thermal parameter data of the weld pool and heat-affected zone using an infrared thermal imager, accurately reflecting the thermal field distribution during the welding process. This provides core data on the thermal field dimension for decoupling thermal interference at multiple workstations and calculating heat source models, enabling real-time monitoring of the welding thermal field. The collected data is then categorized into thermal field distribution sensing data, and this type of sensing data is integrated. For example, if the collected data shows that the center temperature of the weld pool at a certain workstation is 1500℃, the width of the weld pool is 8mm, and the temperature gradient of the heat-affected zone is 200℃ / mm, all of these data are added to the thermal field distribution sensing data.

[0076] Furthermore, the sound pressure level data of the welding process frequency band collected by the microphone array in the tower welding scenario is acquired and added to the welding acoustic perception data.

[0077] The purpose of this step is to collect acoustic signature data during the welding process using a microphone array. This helps to reflect the stability of the welding process from the perspective of acoustic signature, supplements the dimensions of multimodal perception data, and improves the comprehensiveness and accuracy of subsequent multimodal fusion evaluation. The collected data of this type is classified into welding acoustic signature perception data, and the integration of this type of perception data is completed. For example, if the sound pressure level data of the 2kHz band is 85dB and the sound pressure level data of the 5kHz band is 78dB during the welding process, all of these data are added to the welding acoustic signature perception data.

[0078] In step S200 of this application embodiment, the construction of the multimodal fusion evaluation model includes the following prior steps:

[0079] Collect historical welding process datasets for a preset tower model. The historical welding process datasets include historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, historical welding acoustic perception data, and the corresponding historical actual weld deviation.

[0080] Using historical real weld deviation as supervision, and taking historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, and historical welding acoustic perception data as inputs, and using deviation output nodes and the confidence scores of each sensor as outputs, an initial deep confidence network model is trained to obtain a multimodal fusion evaluation model whose output layer includes output nodes.

[0081] In step S200 of this application embodiment, the purpose of the above steps is to prepare sufficient and effective training data for training the multimodal fusion evaluation model, and to train the initial deep belief network model in a supervised manner, so that the model has the ability to accurately calculate the fusion weld deviation and evaluate the confidence scores of each sensor from the multimodal perception data, providing reliable model support for subsequent sensor health status determination and fault-tolerant reconfiguration control mode triggering, and ensuring the accuracy of data fusion evaluation in multi-station tower welding collaborative optimization analysis.

[0082] To achieve the above objectives, it is first necessary to collect a historical welding process dataset for a preset tower model. The historical welding process dataset includes historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, historical welding acoustic perception data, and the corresponding historical actual weld deviation.

[0083] The purpose of this step is to provide real and comprehensive sample data for training the initial deep belief network model, ensuring the effectiveness and relevance of the model training. This is achieved by collecting various historical sensing data generated during the past welding processes of a specified pre-defined tower model, along with the corresponding historical actual weld deviations. The historical welding process dataset must include historical weld visual sensing data, historical arc sensing data, historical thermal field distribution sensing data, historical welding acoustic sensing data, and the corresponding historical actual weld deviations. For example, collecting historical welding data for a 2.5MW wind turbine tower model includes historical weld bevel edge contour data, historical instantaneous welding current data, historical molten pool center temperature data, and historical welding process frequency band sound pressure level data. Simultaneously, the historical actual weld deviation at the corresponding welding position is collected as 0.3mm. These data are then integrated into the historical welding process dataset for that pre-defined tower model.

[0084] Next, using historical actual weld deviation as supervision, and using historical weld visual perception data, historical arc sensing data, historical thermal field distribution sensing data, and historical welding acoustic perception data as inputs, and using deviation output nodes and the confidence scores of each sensor as outputs, an initial deep confidence network model is trained to obtain a multimodal fusion evaluation model whose output layer includes output nodes.

[0085] The purpose of this step is to use supervised deep learning training to enable the initial deep belief network model to learn the correlation between multimodal sensing data, fused weld deviation, and sensor confidence levels. This allows the trained multimodal fusion evaluation model to effectively fuse and evaluate multimodal sensing data. The method involves using collected historical sensing data as input to the model's input layer, using historical real weld deviation as the supervision for model training, and setting the deviation output node and each sensor confidence level as the output of the model's output layer. Deep learning algorithms are then used to iteratively train the initial deep belief network model until its output meets the preset accuracy requirements. Finally, a multimodal fusion evaluation model with output nodes is obtained in the output layer. For example, historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, and historical welding acoustic perception data of a preset tower model are input into the initial deep confidence network model. The corresponding historical real weld deviation of 0.2mm is used as supervision. The calculation result of the fused weld deviation and the confidence scores of the laser vision sensor, arc sensor, infrared thermal imager, and microphone array are set as the model output. After multiple rounds of iterative training, the error between the fused weld deviation output by the model and the historical real weld deviation is controlled within 0.05mm. At this time, the training is completed and the corresponding multimodal fusion evaluation model is obtained.

[0086] For example, the multimodal fusion evaluation model can be obtained by training an initial deep belief network model. The training samples are a historical welding process dataset of a preset tower model, containing historical weld visual, arc sensing, thermal field distribution, and welding acoustic perception data, along with corresponding historical actual weld deviations. The sample size is 5000 sets, and the training rounds are set to 300. The initial deep belief network model takes four classes of historical perception data as input, uses historical actual weld deviations as supervision, and outputs deviation output nodes and the confidence scores of each sensor. Hyperparameters are set as follows: learning rate 0.001, 3 hidden layers, and the number of neurons are 512, 256, and 128 respectively, with ReLU activation function. The convergence criterion is that the average error between the fused weld deviation output by the model and the historical actual weld deviation is consistently below 0.05 mm, and the error does not significantly decrease after 20 consecutive iterations.

[0087] Furthermore, it is necessary to input the multimodal sensing data of tower welding into the multimodal fusion evaluation model to obtain the fused weld deviation and several sensor confidence scores, and the several sensor confidence scores correspond to the multimodal sensing data of tower welding.

[0088] For example, the collected visual perception data of the weld bevel with a V-shaped edge profile, a bevel center point deviation of 0.3 mm, and a bevel width of 14 mm; arc sensing data with an instantaneous welding current of 310 A, an instantaneous welding voltage of 29 V, a current standard deviation of 4 A, a voltage standard deviation of 0.8 V, and a short-circuit frequency of 14 Hz; thermal field distribution data with a molten pool center temperature of 1480 ℃, a molten pool width of 7.5 mm, and a heat-affected zone temperature gradient of 190 ℃ / mm; and sound pressure level of 84 dB in the 2 kHz frequency band. B. Welding acoustic signature sensing data with a sound pressure level of 77dB in the 5kHz frequency band is input into the multimodal fusion evaluation model. After calculation by the multimodal fusion evaluation model, the deviation of the fused weld is obtained as 0.25mm. At the same time, several sensor confidence scores are obtained, namely, 92 points for the laser vision sensor, 95 points for the arc sensor, 93 points for the infrared thermal imager, and 90 points for the microphone array. Each score is matched with the corresponding tower welding multimodal sensing data.

[0089] In step S300 of this embodiment, based on the confidence scores of several sensors, the health status of each sensor is dynamically determined by comparing them with a preset confidence warning threshold and a preset confidence failure threshold, thereby obtaining a sensor health status set, including:

[0090] The confidence score of any sensor is compared one by one with the preset confidence warning threshold and the preset confidence failure threshold.

[0091] When the confidence score of a sensor is greater than or equal to the preset confidence warning threshold, the corresponding sensor is determined to be in a healthy state.

[0092] When the sensor confidence score is less than the preset confidence warning threshold but greater than or equal to the preset confidence failure threshold, the corresponding sensor is determined to be in an interference state.

[0093] When the confidence score of a sensor is less than the preset confidence failure threshold, the corresponding sensor is determined to be in a failure state.

[0094] The results of the health status, interference status, and failure status of all sensors are combined to form a sensor health status set.

[0095] In step S300 of this application embodiment, the purpose of the above step is to accurately identify the real-time working status of each sensor during the tower welding process, so as to provide a direct basis for subsequent triggering of the corresponding fault-tolerant reconfiguration control mode based on the sensor status, solve the problem that the sensors at the welding site are easily affected by arc light and dust interference, resulting in signal abnormalities or failures, and ensure the fault tolerance and stability of welding tracking control.

[0096] To achieve the above objectives, it is first necessary to compare the confidence score of any sensor with the preset confidence warning threshold and the preset confidence failure threshold one by one.

[0097] The purpose of this step is to establish a quantitative comparison basis for the health status determination of each sensor, and to initially distinguish the state range of the sensor through numerical comparison. The method is to extract the confidence score of each sensor output by the multimodal fusion evaluation model, and compare the scores with the pre-set confidence warning threshold and confidence failure threshold in turn. For example, if the pre-set confidence warning threshold is 80 points and the pre-set confidence failure threshold is 50 points, and the extracted confidence score of the laser vision sensor is 75 points, the score of 75 points is first compared with the score of 80 points, and then compared with the score of 50 points, and the comparison operation of each sensor is completed.

[0098] When the confidence score of a sensor is greater than or equal to the preset confidence warning threshold, the corresponding sensor is determined to be in a healthy state.

[0099] The purpose of this step is to clarify the criteria for determining whether the sensor is in a normal working state with no interference and reliable data. This is achieved by comparing the sensor confidence score with two thresholds. If the score meets the condition of being greater than or equal to the confidence warning threshold, the sensor's working state can be determined as healthy. For example, if the preset confidence warning threshold is 80 points and the confidence failure threshold is 60 points, and the arc sensor's confidence score is 95 points, since 95 points is greater than the confidence warning threshold of 80 points, the arc sensor is determined to be in a healthy state.

[0100] When the sensor confidence score is less than the preset confidence warning threshold but greater than or equal to the preset confidence failure threshold, the corresponding sensor is determined to be in an interference state.

[0101] The purpose of this step is to clarify the criteria for determining the state of a sensor that is affected by environmental factors such as arc light and smoke, resulting in a decrease in data reliability but not complete failure. This is achieved by, after completing the threshold comparison, if the sensor's confidence score is between the confidence warning threshold and the confidence failure threshold, the sensor's operating state can be determined as an interference state. For example, if the preset confidence warning threshold is 80 points and the confidence failure threshold is 60 points, and the infrared thermal imager's confidence score is 70 points, 70 points is less than the confidence warning threshold of 80 points and greater than the confidence failure threshold of 60 points, therefore the infrared thermal imager is determined to be in an interference state.

[0102] When the confidence score of a sensor is less than the preset confidence failure threshold, the corresponding sensor is determined to be in a failure state.

[0103] The purpose of this step is to clarify the criteria for determining the state of a sensor where signal loss or data unreliability is caused by severe interference or equipment failure. This is achieved by, after completing the threshold comparison, if the sensor confidence score is less than the confidence failure threshold, the sensor's working state can be determined as a failure state. For example, if the preset confidence warning threshold is 80 points and the confidence failure threshold is 60 points, and the microphone array confidence score is 55 points, since 55 points is less than the confidence failure threshold of 60 points, the microphone array is determined to be in a failure state.

[0104] The results of the health status, interference status, and failure status of all sensors are combined to form a sensor health status set.

[0105] The independent state judgment results of each sensor are integrated to form a unified and complete sensor state set, providing a comprehensive state reference for triggering subsequent fault-tolerant reconfiguration control modes. This is achieved by collecting the state judgment results of all sensors involved in perception, such as laser vision sensors, arc sensors, infrared thermal imagers, and microphone arrays, and associating and integrating each sensor with its corresponding judgment state to form a structured sensor health state set.

[0106] In step S400 of this embodiment, based on the sensor health status set and the fused weld deviation, a corresponding fault-tolerant reconfiguration control mode is triggered to obtain the reconfigured welding tracking control parameters, including:

[0107] When all sensors in the sensor health status set are determined to be in a healthy state, the full fusion tracking control mode is triggered, and the fusion weld deviation is directly converted into welding tracking control parameters.

[0108] When the sensor health status is determined to be in an interference state by the visual sensor, and the arc sensor, thermal imager, and acoustic sensor are not determined to be in a failure state, the downgraded fusion tracking control mode is triggered. The weight of the weld visual perception data in the fusion calculation is reduced, while the weight of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data in the fusion calculation is increased. Based on the adjusted weights, the downgraded fusion weld deviation is recalculated and converted into welding tracking control parameters.

[0109] When the sensor health status is determined to be in a failed state by the visual sensor, the vision-free guidance control mode is triggered. Historical weld path trajectory data is loaded, and the weld position is estimated by combining the instantaneous values ​​of welding current, welding voltage, and short-circuit frequency data in the arc sensor data. The estimated weld deviation is then converted into welding tracking control parameters.

[0110] In step S400 of this application embodiment, the purpose of the above steps is to dynamically match and adapt the welding tracking control strategy according to the actual working state of each sensor, avoid the welding trajectory deviation problem caused by sensor interference or failure through fault-tolerant reconstruction, and ensure that accurate welding tracking control parameters can be output under different sensor states. This provides a reliable control basis for subsequent multi-station thermal interference decoupling and coordinated regulation, avoids tower welding shutdown or welding deviation caused by sensor abnormalities, and ensures the continuity and welding accuracy of the multi-station tower welding process.

[0111] To achieve the above objectives, when all sensors in the sensor health status set are determined to be in a healthy state, the full fusion tracking control mode needs to be triggered, and the fusion weld deviation amount needs to be directly converted into welding tracking control parameters.

[0112] First, the sensor health status set is verified. After confirming that the laser vision sensor, arc sensor, infrared thermal imager, and microphone array are all in good condition, the full fusion tracking control mode is triggered. The fusion weld deviation output by the multimodal fusion evaluation model is directly converted into welding tracking control parameters such as welding torch movement distance and welding torch angle according to the preset parameter conversion rules. For example, if the laser vision sensor, arc sensor, infrared thermal imager, and microphone array are all in good condition in the sensor health status set, and the fusion weld deviation output by the multimodal fusion evaluation model is 0.2mm, the 0.2mm fusion weld deviation is directly converted into welding tracking control parameters such as 0.2mm lateral compensation of the welding torch along the bevel centerline and the welding torch angle being maintained at 30° according to the preset rules.

[0113] The preset rules are standardized mapping rules in the tower welding field for converting weld deviation into welding tracking control parameters. These rules need to be formulated in conjunction with the welding torch movement control logic, bevel size requirements, and welding process specifications for tower welding. The core is to convert the numerical value and direction of the weld deviation into corresponding control parameters such as the welding torch's compensation distance in the lateral / longitudinal direction, the welding torch angle fine-tuning, and the travel speed correction value. Simultaneously, they are matched with preset station welding accuracy standards to ensure that the deviation compensation accurately adapts to the actual working conditions of tower welding. For example, a weld deviation of 0.2mm corresponds to a lateral compensation of 1mm for the welding torch; if the deviation direction is left, the welding torch moves to the left; if it is right, it moves to the right, while the welding torch angle and travel speed remain at their default values.

[0114] Specifically, the formulation of the preset rules needs to be based on the actual working conditions of tower welding. First, it is necessary to match the hardware logic of the welding torch motion control, and determine the parameter conversion basis based on the displacement accuracy, angle adjustment range, and speed adjustment step size of the welding torch servo motor. Second, it is necessary to conform to the design requirements of the tower weld bevel size, and define the deviation compensation range according to the bevel angle, root gap, and bevel width. Third, it is necessary to follow the tower welding industry process specifications and the corresponding tower model welding process documents, and clarify the accuracy standards for different weld positions. It should also meet the requirements of multi-station welding collaborative control, ensure the uniformity of conversion rules for each station, and adapt to the consistency requirements of circumferential weld joints.

[0115] The core mapping logic of the preset rules uses the numerical value and direction of the weld deviation as the core input to establish a precise mapping with the welding tracking control parameters: the deviation value is linearly converted according to the hardware displacement accuracy of the welding torch, directly corresponding to the size of the horizontal and vertical compensation distance of the welding torch, and the larger the value, the greater the compensation amount; the deviation direction clearly defines the welding torch compensation direction, the horizontal left deviation of the bevel causes the welding torch to move to the left, the right deviation causes the welding torch to move to the right, and the vertical forward / backward deviation corresponds to the synchronous forward / backward movement of the welding torch; and the adjustment type can be divided according to the deviation threshold, small deviations only adjust the welding torch displacement compensation, while large deviations simultaneously fine-tune parameters such as the welding torch angle and travel speed.

[0116] The preset rules also include process constraints, which must be strictly followed: the welding torch compensation distance should not exceed 1 / 2 of the single-sided gap of the bevel to avoid welding defects such as weld deviation and lack of fusion; the welding torch angle fine adjustment should be controlled within ±5° to match the weld pool formation requirements of the thick-walled tower weld and ensure that the penetration depth meets the standard; the fluctuation of the welding torch travel speed correction value should not exceed ±20% of the default value to maintain stable welding heat input and prevent heat accumulation or insufficient heat input; all converted control parameters should meet the adjustment accuracy of the welding torch servo system and there should be no hardware incompatibility such as over-range or non-integer step size.

[0117] When the sensor health status is determined to be in an interference state by the visual sensor, and the arc sensor, thermal imager, and acoustic sensor are not determined to be in a failure state, the downgraded fusion tracking control mode is triggered. The weight of the weld visual perception data in the fusion calculation is reduced, while the weight of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data in the fusion calculation is increased. Based on the adjusted weights, the downgraded fusion weld deviation is recalculated and converted into welding tracking control parameters.

[0118] First, the health status set of the sensors is verified. If the laser vision sensor is in an interference state and the arc sensor, infrared thermal imager, and microphone array are all in a healthy or interference state and have not failed, the downgraded fusion tracking control mode is triggered. According to the preset weight adjustment rules, the fusion weight of the weld visual perception data is reduced, while the fusion weight of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data is increased. Then, based on the adjusted weights, the fusion calculation of each modal perception data is performed again to obtain the downgraded fusion weld deviation. Finally, the deviation is converted into welding tracking control parameters according to the parameter conversion rules.

[0119] The preset weight adjustment rule is triggered by interference from the laser vision sensor and the non-failure of the other three types of sensors. The quantification basis is the health status of each sensor. The core rule is to reduce the weight of the interfered weld visual perception data, and to increase the weight of the healthy / non-failed arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data by an equal amount, while ensuring that the total weight of the four types of perception data is always 1.

[0120] For example, in the sensor health status analysis, the laser vision sensor is in an interference state, while the arc sensor, infrared thermal imager, and microphone array are all in a healthy state. In the original fusion calculation, the weight of the weld visual perception data was 0.4, and the weights of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data were each 0.2. After adjustment, the weight of the weld visual perception data was reduced to 0.1, and the weights of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data were each increased to 0.3. Based on the new weights, the degraded fusion weld deviation was recalculated to be 0.22mm. This deviation was then converted into welding tracking control parameters with a lateral compensation of 0.22mm along the bevel centerline of the welding torch and a welding torch travel speed of 5mm / s.

[0121] When the sensor health status is determined to be in a failed state by the visual sensor, the vision-free guidance control mode is triggered. Historical weld path trajectory data is loaded, and the weld position is estimated by combining the instantaneous values ​​of welding current, welding voltage, and short-circuit frequency data in the arc sensor data. The estimated weld deviation is then converted into welding tracking control parameters.

[0122] First, the sensor health status set is checked. After confirming that the laser vision sensor is in a failed state, the vision-free guidance control mode is triggered. Historical weld path trajectory data of the same tower and the same work position are loaded from the database as the basic trajectory reference. The instantaneous values ​​of welding current, welding voltage, and short-circuit frequency data in the arc sensor perception data are extracted and substituted into the preset weld position estimation algorithm. Combined with the historical trajectory data, the deviation between the actual weld position and the estimated weld position of the standard trajectory is calculated. Finally, the deviation is converted into welding tracking control parameters according to the parameter conversion rules.

[0123] The principle of the weld position estimation algorithm is based on the correlation between arc sensing data and weld position. Using historical weld path trajectory data of the same tower and the same work station as a benchmark, the algorithm quantifies the offset of the actual weld position relative to the standard trajectory by real-time changes in instantaneous welding current, instantaneous welding voltage, and short-circuit frequency, thus realizing weld position estimation without visual guidance.

[0124] The weld position estimation algorithm first calibrates the historical weld path trajectory data as a standard weld path, and then establishes a quantitative correspondence between the arc sensing data and the weld position offset: when the arc sensing data deviates from the preset standard welding state threshold, it is converted into the weld lateral / longitudinal offset value according to the arc sensing data-offset conversion ratio; if there is no threshold deviation, the historical weld path trajectory data is used as the benchmark to determine that there is no offset.

[0125] The threshold for the preset standard welding state is obtained based on the arc sensing data corresponding to the historical weld path trajectory data of the same tower type and the same work station. First, historical arc sensing data of the work station with qualified welding quality and qualified weld formation are selected, and the stable fluctuation range of the instantaneous value of welding current, instantaneous value of welding voltage, and short-circuit frequency are extracted. The average value of the range is taken as a standard value, and the threshold is calibrated within the range of ±5% to 10% of the standard value. After verifying the adaptability of the threshold under actual welding conditions on site, it is determined as the final threshold for the preset standard welding state.

[0126] The arc sensing data-offset conversion ratio is obtained based on the historical weld path trajectory data and the corresponding historical arc sensing data. For example, it can be that for every 10A deviation of the current from the standard value, the weld position shifts by 0.05mm; for every 1V deviation of the voltage, the weld position shifts by 0.05mm; and for every 3Hz deviation of the short-circuit frequency, the weld position shifts by 0.05mm.

[0127] For example, if the laser vision sensor is in a failed state and the historical weld path trajectory data of a certain circumferential weld position on a 2.5MW wind turbine tower is loaded as a standard trajectory, and the instantaneous welding current value of 320A, the instantaneous welding voltage value of 30V, and the short-circuit frequency of 15Hz are extracted from the arc sensing data, these electrical parameters are substituted into the weld position estimation algorithm. Combined with the arc sensing data-offset conversion ratio obtained from the historical weld path trajectory data, the weld position deviation is estimated to be 0.3mm. This estimated weld deviation is then converted into welding tracking control parameters with a longitudinal compensation of 0.3mm along the bevel centerline of the welding torch and a wire feed speed of 8m / min.

[0128] like Figure 2As shown, in step S500 of this embodiment, a multi-station welding heat source model is loaded. Based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, the thermal interference between the multiple stations is decoupled and coordinated to obtain multi-station coordinated control commands, including:

[0129] Input the real-time welding power data and real-time welding torch position data of each station into the multi-station welding heat source model to obtain the prediction data of the overall transient thermal field distribution of the tower.

[0130] When the temperature data of the molten pool at the first station exceeds the preset temperature anomaly threshold, the abnormal temperature data of the molten pool at the first station is input into the multi-station welding heat source model to back-infer the heat source and obtain the heat interference source identification result.

[0131] When the thermal interference source identification result determines that the abnormal temperature of the molten pool originates from the thermal radiation interference of the second station, the cloud-based collaborative controller sends a thermal shielding command to the second station. The thermal shielding command includes welding torch angle adjustment parameters and local cooling activation parameters.

[0132] The cloud-based collaborative controller performs unified calibration of the arc start and end points of the circumferential welds at multiple workstations based on the fusion weld deviation at each workstation at the current moment, thereby obtaining joint consistency calibration parameters.

[0133] By merging the thermal shielding command and the connector consistency calibration parameters, a multi-station collaborative control command is obtained.

[0134] In step S500 of this application embodiment, the purpose of the above steps is to predict the overall thermal field of tower welding and accurately identify the sources of thermal interference through a multi-station welding heat source model, take targeted thermal shielding measures to eliminate thermal coupling interference between multiple stations, and uniformly calibrate the arc starting point and arc ending point of the circumferential weld at multiple stations to ensure the consistency of the weld joint. Finally, the multi-station collaborative control commands are integrated to realize the collaborative control of the thermal field and trajectory of multi-station tower welding, solve the problems of thermal field superposition, welding deformation and stress concentration caused by simultaneous operation of multiple welding guns, and ensure the overall quality and stability of multi-station welding.

[0135] To achieve the above objectives, it is first necessary to input the real-time welding power data and real-time welding torch position data of each station into the multi-station welding heat source model to obtain the overall transient thermal field distribution prediction data of the tower.

[0136] This involves collecting real-time welding power data and welding torch position data for each station in a multi-station welding scenario. This data is then synchronously input into a multi-station welding heat source model that has undergone parameter identification. The model uses a double-ellipsoidal heat source distribution function and a heat source superposition factor to perform calculations, outputting predicted transient thermal field distribution data for the entire tower. For example, if the real-time welding power data for three stations in tower welding are 30kW, 28kW, and 32kW, and the real-time welding torch position data are at 0°, 120°, and 240° of the tower circumferential seam, this data is input into the multi-station welding heat source model. After calculation, predicted transient thermal field distribution data for the entire tower is obtained, with predicted temperatures around the molten pool at station 0° being 1550℃, at station 120° 1520℃, and at station 240° 1580℃.

[0137] When the temperature data of the molten pool at the first station exceeds the preset temperature anomaly threshold, the abnormal temperature data of the molten pool at the first station is input into the multi-station welding heat source model to back-infer the heat source and obtain the heat interference source identification result.

[0138] First, thermal field distribution sensing data is collected using an infrared thermal imager to monitor the molten pool temperature data at each workstation in real time. This data is then compared with a preset temperature anomaly threshold. When the molten pool temperature data at the first workstation exceeds the threshold, the abnormal molten pool temperature data for that workstation is extracted and input into a multi-workstation welding heat source model. The model performs heat source back-calculation to analyze the cause of the temperature anomaly and identify the corresponding thermal interference source, outputting the thermal interference source identification result. For example, if the preset molten pool temperature anomaly threshold is 1600℃, and the molten pool temperature data at the first workstation is detected to be 1650℃ based on the thermal field distribution sensing data, exceeding the temperature anomaly threshold, this 1650℃ molten pool temperature anomaly data is input into the multi-workstation welding heat source model for heat source back-calculation. After calculation, the thermal interference source identification result is obtained, determining that the temperature anomaly is caused by thermal radiation interference from the second workstation.

[0139] The first and second workstations are not in a one-to-one correspondence. They are two random, independent workstations in a multi-workstation tower welding scenario, chosen as example workstations to illustrate the thermal interference decoupling and collaborative control logic. The first workstation is the abnormal workstation that detects the molten pool temperature data exceeding a preset threshold. The second workstation is the thermal interference source workstation identified as causing the temperature anomaly by reverse engineering from the multi-workstation welding heat source model. The relationship between the two is determined solely by the actual occurrence of thermal radiation interference and has no fixed one-to-one correspondence. It can be flexibly determined based on the actual occurrence scenario of thermal interference on site.

[0140] When the thermal interference source identification result determines that the abnormal temperature of the molten pool originates from the thermal radiation interference of the second station, the cloud-based collaborative controller sends a thermal shielding command to the second station. The thermal shielding command includes welding torch angle adjustment parameters and local cooling activation parameters.

[0141] Once the thermal interference source identification results determine that the abnormal molten pool temperature is caused by thermal radiation interference from the second station, the cloud-based collaborative controller generates a thermal shielding command based on the degree and scope of the thermal interference. This command includes parameters for adjusting the welding torch angle and activating local cooling. The command is then sent to the welding control system of the second station to guide the second station in performing the thermal shielding operation. For example, if the thermal interference source identification results determine that the abnormal molten pool temperature at the first station originates from thermal radiation interference from the second station, the cloud-based collaborative controller sends a thermal shielding command to the second station. The welding torch angle adjustment parameter is to adjust the welding torch 15° away from the first station, and the local cooling activation parameter is to activate the local cooling fan and adjust the cooling airflow to 5 cubic meters per minute.

[0142] Next, the cloud-based collaborative controller performs unified calibration on the arc initiation and arc termination points of the multi-station circumferential welds based on the fusion weld deviation of each station at the current moment, and obtains joint consistency calibration parameters.

[0143] This involves the cloud-based collaborative controller collecting real-time deviations of the fusion weld at each workstation. Based on the welding process requirements of the tower circumferential weld, and using the standard arc-starting and arc-ending points as benchmarks, it calculates corresponding calibration adjustment values ​​based on the deviations at each workstation. This process uniformly calibrates the arc-starting and arc-ending points of each workstation, ultimately generating joint consistency calibration parameters. For example, if the cloud-based collaborative controller collects fusion weld deviations of 0.2mm, 0.15mm, 0.22mm, and 0.18mm at the four workstations of the tower welding, and the standard arc-starting point of the tower circumferential weld is 0°... Using 90°, 180°, 270° and standard arc termination points of 89°, 179°, 269°, 359° as benchmarks, and combining the deviations of each station, a unified calibration is performed to obtain the joint consistency calibration parameters. The calibration values ​​are: arc termination point +0.2mm for the first station, arc termination point +0.15mm for the second station, arc termination point +0.15mm for the third station, arc termination point +0.22mm for the third station, and arc termination point +0.22mm for the fourth station.

[0144] By merging the thermal shielding command and the connector consistency calibration parameters, a multi-station collaborative control command is obtained.

[0145] All heat shielding commands and calculated joint consistency calibration parameters issued by the cloud-based collaborative controller will be summarized and integrated, categorized and associated according to workstation number, to form a multi-workstation collaborative control command containing heat shielding operation parameters and joint calibration parameters for each workstation. This command will then be sent to the welding control system of each workstation. For example, the heat shielding command sent to the second workstation to adjust the welding torch angle by 15° and the local cooling airflow of 5m³ / min will be merged with the joint consistency calibration parameters of the four workstations to form a multi-workstation collaborative control command. The command will specify the arc starting point and arc ending point calibration parameters for each workstation, and will also specify the operation requirements for the second workstation to adjust the welding torch angle and activate local cooling.

[0146] In step S500 of this embodiment, loading the multi-station welding heat source model further includes:

[0147] Collect historical welding power data, historical welding torch position coordinates, and corresponding historical tower thermal imaging data for each workstation to construct a heat source model training set.

[0148] Based on the double ellipsoidal heat source distribution function, an initial heat source model including a heat source superposition factor is constructed, where the heat source superposition factor is used to characterize the energy coupling effect of heat sources in adjacent workstations.

[0149] Using historical tower thermal imaging data as supervision, and historical welding power data and historical welding torch position coordinates as input, the parameters of the initial heat source model are identified to obtain a multi-station welding heat source model.

[0150] In step S500 of this application embodiment, the purpose of the above steps is to build an accurate and realistic heat source model for multi-station tower welding thermal field prediction, thermal interference source identification, and thermal collaborative control. By collecting historical welding data to construct a training set, and combining the double ellipsoidal heat source distribution function to build an initial model and complete parameter identification, the model can accurately characterize the heat source energy coupling effect between multiple stations, accurately calculate the thermal field distribution of tower welding, and provide reliable model support for subsequent multi-station thermal interference decoupling and collaborative control.

[0151] To achieve the above objectives, it is first necessary to collect historical welding power data, historical welding torch position coordinates, and corresponding historical tower thermal imaging data for each workstation, and construct a heat source model training set.

[0152] This involves collecting historical welding power data and welding torch position coordinate data for each workstation during the multi-station tower welding process. Simultaneously, it collects tower thermal imaging data corresponding to the two sets of data in terms of time and workstation. The three types of data are then organized and integrated according to a preset format to form a structured heat source model training set. For example, historical welding power data for three tower welding workstations are collected as 29kW, 31kW, and 30kW, and historical welding torch position coordinates are at 0°, 120°, and 240° of the tower circumferential seam. At the same time, historical tower thermal imaging data for the corresponding time and workstation are collected as molten pool center temperatures of 1490℃, 1510℃, and 1500℃. These data are then integrated according to workstation association to construct a heat source model training set.

[0153] Next, based on the double ellipsoidal heat source distribution function, an initial heat source model including a heat source superposition factor is constructed, where the heat source superposition factor is used to characterize the energy coupling effect of heat sources in adjacent workstations.

[0154] Based on the double ellipsoidal heat source distribution function, a heat source superposition factor is added to the function to characterize the energy coupling effect of heat sources in adjacent workstations. The numerical range and calculation logic of the factor are defined. Combined with the spatial layout and heat source characteristics of multi-workstation welding, an initial multi-workstation welding heat source model is built. For example, based on the double ellipsoidal heat source distribution function, a heat source superposition factor with a value range of 0-1 is introduced. The factor is set to 0.8 when the distance between welding torches in adjacent workstations is less than 500mm and 0.3 when it is greater than 500mm. This characterizes the energy coupling effect of heat sources in adjacent workstations under different distances, thus completing the construction of the initial heat source model.

[0155] Then, using historical tower thermal imaging data as supervision and historical welding power data and historical welding torch position coordinates as input, the parameters of the initial heat source model are identified to obtain a multi-station welding heat source model.

[0156] The historical welding power data and historical welding torch position coordinates from the heat source model training set are input into the initial heat source model. Historical tower thermal imaging data are used as the supervision basis for model training. The parameters in the model, such as the double ellipsoidal heat source parameters and heat source superposition factor, are continuously adjusted through parameter identification algorithms until the error between the model output thermal field data and the historical tower thermal imaging data is controlled within a preset accuracy range. Finally, a multi-station welding heat source model with optimized parameters is obtained. For example, the historical welding powers of 29kW, 31kW, and 30kW and the historical welding torch position coordinates of 0°, 120°, and 240° from the heat source model training set are input into the initial heat source model. The corresponding historical tower thermal imaging data of molten pool center temperature of 1490℃, 1510℃, and 1500℃ are used as supervision. The model parameters are iteratively adjusted through parameter identification algorithms so that the error between the model output molten pool center temperature and the actual thermal imaging data does not exceed ±10℃. At this point, parameter identification is completed and a multi-station welding heat source model is obtained.

[0157] In step S500 of this application embodiment, the following is also included:

[0158] When the confidence scores of several sensors are all less than the preset confidence failure threshold, the sensor health status set is determined to be in a fully failed state, triggering the emergency shutdown protection mode, generating a shutdown alarm command and terminating the welding process.

[0159] In step S500 of this embodiment, the purpose of the above steps is to prevent quality problems or safety accidents caused by uncontrolled welding when all sensor confidence scores are below the failure threshold and there is no effective sensing data. This is achieved by determining that the sensor health status set is in a completely failed state, triggering an emergency shutdown protection mode, generating a shutdown alarm command, and immediately terminating the welding process. For example, a preset confidence failure threshold of 60 points is set. If the confidence scores of the laser vision, electric arc, infrared thermal imager, and microphone array are all below 60 points, a shutdown is triggered, terminating the welding process.

[0160] In step S500 of this application embodiment, the following is also included:

[0161] The historical welding process of tower welding was used to obtain the welding quality false alarm rate due to sensor failure and the welding quality false alarm rate due to environmental interference.

[0162] Calculate the ratio of the welding quality false alarm rate to the welding quality false alarm rate, and set it as the confidence threshold adjustment factor;

[0163] Based on the ratio of the preset baseline adjustment factor to the confidence threshold adjustment factor, the preset confidence warning threshold and the preset confidence failure threshold are adjusted.

[0164] In step S500 of this application embodiment, the purpose of the above steps is to dynamically adjust the preset confidence warning threshold and confidence failure threshold based on historical welding quality detection data of the tower welding scenario, so that the threshold setting is more in line with the environmental characteristics of actual welding and the working characteristics of the sensor, avoid sensor status misjudgment, welding quality missed or false alarm problems caused by fixed thresholds, improve the accuracy of sensor health status judgment, and thus ensure the rationality of fault-tolerant reconfiguration control mode triggering and the reliability of welding quality detection.

[0165] To achieve the above objectives, it is first necessary to obtain the welding quality false alarm rate caused by sensor failure and the welding quality false alarm rate caused by environmental interference during the historical welding process of the tower welding scenario.

[0166] This involves statistically analyzing past welding process quality inspection records and sensor operating status records for tower welding scenarios. The ratio of cases where welding quality issues were not detected due to sensor failure to the total number of quality inspection cases is used to obtain the welding quality false alarm rate. Simultaneously, the ratio of cases where welding quality issues were mistakenly identified due to environmental interference to the total number of quality inspection cases is used to obtain the welding quality false alarm rate. For example, analyzing nearly 1000 historical welding quality inspection records for a certain tower welding scenario, if there are 20 cases of missed welding quality detection due to sensor failure, the calculated welding quality false alarm rate is 2%, and if there are 10 cases of false welding quality detection due to environmental interference, the calculated welding quality false alarm rate is 1%.

[0167] Next, the ratio of the welding quality false alarm rate to the welding quality false alarm rate is calculated and set as the confidence threshold adjustment factor.

[0168] The obtained welding quality false alarm rate is used as the dividend and the welding quality false alarm rate is used as the divisor. A division operation is performed, and the result is defined as the confidence threshold adjustment factor. For example, if the welding quality false alarm rate is 2% and the welding quality false alarm rate is 1%, the ratio of 2% to 1% is calculated to obtain result 2. This result is set as the confidence threshold adjustment factor.

[0169] Then, based on the ratio of the preset baseline adjustment factor to the confidence threshold adjustment factor, the preset confidence warning threshold and the preset confidence failure threshold are adjusted.

[0170] This involves pre-setting a fixed baseline adjustment factor, dividing it by the calculated confidence threshold adjustment factor to obtain a threshold adjustment coefficient, and then multiplying this coefficient by the original preset confidence warning threshold and confidence failure threshold to obtain the adjusted confidence warning threshold and confidence failure threshold. For example, if the preset confidence baseline adjustment factor is 1, the calculated confidence threshold adjustment factor is 2, and the ratio of 1 to 2 yields a threshold adjustment coefficient of 0.5, and the original preset confidence warning threshold is 80 points and the confidence failure threshold is 60 points, multiplying the adjustment coefficient 0.5 by the two thresholds respectively yields an adjusted confidence warning threshold of 40 points and a confidence failure threshold of 30 points.

[0171] In step S500 of this application embodiment, the following is also included:

[0172] To obtain real-time welding progress data and real-time weld height detection data for each workstation in the tower welding scenario;

[0173] When the real-time weld reinforcement detection data exceeds the preset reinforcement height threshold, the corresponding location is identified as a repair welding area;

[0174] Based on real-time welding progress data, adjacent workstations that are idle or ahead in welding progress are scheduled to generate repair welding coordination instructions and perform online repair welding operations on the repair welding area.

[0175] In step S500 of this application embodiment, the purpose of the above steps is to monitor and control the weld reinforcement height in real time and perform online repair welding during the tower welding process. By acquiring the welding progress and weld reinforcement height data of each station, the repair welding area with excessive reinforcement height is accurately identified. Then, in combination with the welding progress scheduling of idle or advanced adjacent stations, online repair welding is carried out to correct welding quality defects in a timely manner, avoid subsequent rework, and make full use of the operation resources of multiple stations to ensure the overall quality and operation efficiency of tower welding.

[0176] To achieve the above objectives, it is first necessary to obtain real-time welding progress data and real-time weld height detection data for each workstation in the tower welding scenario.

[0177] The system collects data such as the completed welding length and the completed circumferential angle of each workstation through the welding control system to form real-time welding progress data. At the same time, the weld inspection equipment collects the actual value of the weld reinforcement in the welded area of ​​each workstation to form real-time weld reinforcement detection data. For example, the system collects real-time welding progress data for the four workstations of the tower welding: the first workstation has completed 60% of the circumferential weld, the second workstation has completed 70% of the circumferential weld, the third workstation has completed 90% of the circumferential weld, and the fourth workstation is idle. At the same time, the real-time weld reinforcement detection data of a certain section of the weld at the first workstation is 5mm, the weld reinforcement at the second workstation is 3mm, and the weld reinforcement at the third workstation is 2.8mm.

[0178] Next, when the real-time weld reinforcement detection data exceeds the preset reinforcement height threshold, the corresponding location is identified as a repair welding area;

[0179] The real-time weld height detection data collected from each workstation will be compared with the preset weld height threshold. If the weld height detection data at a certain location exceeds the threshold, the location will be marked and identified as a weld repair area that needs to be repaired. For example, the preset weld height threshold for tower welding is 4mm. The real-time weld height detection data for a certain section of weld at the first workstation is 5mm. This value exceeds the preset weld height threshold of 4mm. Therefore, the weld section at the first workstation will be identified as a weld repair area.

[0180] Then, based on real-time welding progress data, adjacent workstations that are idle or ahead in welding progress are scheduled to generate a welding repair coordination instruction and perform online welding repair operations on the repair area.

[0181] Based on the real-time welding progress data of each workstation, adjacent workstations that are idle or have a leading welding progress are selected. The cloud-based collaborative controller, combined with information such as the location and range of the welding area, generates a welding collaborative instruction that includes the welding location, welding process parameters, and the workstation. This instruction is then sent to the scheduled workstation, which performs online welding operations on the welding area. For example, if the first workstation has a welding area, and its adjacent fourth workstation is idle while the third workstation has a 90% lead in welding progress, the idle fourth workstation is scheduled to perform the welding operation. The cloud-based collaborative controller generates a welding collaborative instruction that specifies the welding area as the weld segment at 60% of the circumferential weld at the first workstation, the welding current as 280A, and the welding voltage as 28V. The fourth workstation then performs online welding operations on the welding area according to the instruction.

Claims

1. A collaborative optimization analysis method for multi-station tower welding, characterized in that, include: S100: Load multi-modal perception data of tower welding in multi-station welding scenarios. The multi-modal perception data of tower welding includes weld visual perception data, arc sensing perception data, thermal field distribution perception data and welding acoustic perception data. S200: Construct a multimodal fusion evaluation model, input the multimodal sensing data of tower welding into the multimodal fusion evaluation model, obtain the fusion weld deviation and several sensor confidence scores, and the several sensor confidence scores correspond to the multimodal sensing data of tower welding. S300: Based on the confidence scores of several sensors, compare them with the preset confidence warning threshold and the preset confidence failure threshold respectively, dynamically determine the health status of each sensor, and obtain the sensor health status set; S400: Based on the sensor health status set and the fused weld deviation, trigger the corresponding fault-tolerant reconfiguration control mode to obtain the reconfigured welding tracking control parameters; S500: Loads a multi-station welding heat source model, and based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, decouples and coordinates the thermal interference between multiple stations, obtains multi-station collaborative control commands, and executes collaborative control of the tower welding process. Loading multimodal sensing data of tower welding in multi-station welding scenarios, including: In the scenario of tower welding, the weld groove edge contour data, groove center point deviation data and groove width data collected by the laser vision sensor are added to the weld visual perception data. In the scenario of tower welding, the instantaneous value of welding current, instantaneous value of welding voltage, standard deviation of current, standard deviation of voltage, and short-circuit frequency data collected by the arc sensor are added to the arc sensor sensing data. In the scenario of tower welding, the infrared thermal imager collects the center temperature data of the molten pool, the width data of the molten pool, and the temperature gradient data of the heat-affected zone, and adds the thermal field distribution sensing data. Acquire the frequency band sound pressure level data of the welding process collected by the microphone array in the scenario of tower welding, and add it to the welding acoustic perception data; The construction of a multimodal fusion evaluation model previously included: Collect historical welding process datasets for a preset tower model. The historical welding process datasets include historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data, historical welding acoustic perception data, and the corresponding historical actual weld deviation. Using historical real weld deviation as supervision, and taking historical weld visual perception data, historical arc sensing perception data, historical thermal field distribution perception data and historical welding acoustic perception data as input, and using deviation output nodes and confidence scores of each sensor as output, an initial deep confidence network model is trained to obtain a multimodal fusion evaluation model with output nodes in the output layer. A multi-station welding heat source model is loaded. Based on the reconstructed welding tracking control parameters and thermal field distribution sensing data, thermal interference between multiple stations is decoupled and coordinated to obtain multi-station coordinated control commands, including: Input the real-time welding power data and real-time welding torch position data of each station into the multi-station welding heat source model to obtain the prediction data of the overall transient thermal field distribution of the tower. When the temperature data of the molten pool at the first station exceeds the preset temperature anomaly threshold, the abnormal temperature data of the molten pool at the first station is input into the multi-station welding heat source model to back-infer the heat source and obtain the heat interference source identification result. When the thermal interference source identification result determines that the abnormal temperature of the molten pool originates from the thermal radiation interference of the second station, the cloud-based collaborative controller sends a thermal shielding command to the second station. The thermal shielding command includes welding torch angle adjustment parameters and local cooling activation parameters. The cloud-based collaborative controller performs unified calibration of the arc start and end points of the circumferential welds at multiple workstations based on the fusion weld deviation at each workstation at the current moment, thereby obtaining joint consistency calibration parameters. By merging the thermal shielding command and the connector consistency calibration parameters, a multi-station collaborative control command is obtained.

2. The collaborative optimization analysis method for multi-station tower welding according to claim 1, characterized in that, Based on the confidence scores of several sensors, the health status of each sensor is dynamically determined by comparing them with preset confidence warning thresholds and preset confidence failure thresholds, thus obtaining a sensor health status set, including: The confidence score of any sensor is compared one by one with the preset confidence warning threshold and the preset confidence failure threshold. When the confidence score of a sensor is greater than or equal to the preset confidence warning threshold, the corresponding sensor is determined to be in a healthy state. When the sensor confidence score is less than the preset confidence warning threshold but greater than or equal to the preset confidence failure threshold, the corresponding sensor is determined to be in an interference state. When the confidence score of a sensor is less than the preset confidence failure threshold, the corresponding sensor is determined to be in a failure state. The results of the health status, interference status, and failure status of all sensors are combined to form a sensor health status set.

3. The collaborative optimization analysis method for multi-station tower welding according to claim 2, characterized in that, Based on the sensor health status set and the fused weld deviation, the corresponding fault-tolerant reconfiguration control mode is triggered to obtain the reconfigured welding tracking control parameters, including: When all sensors in the sensor health status set are determined to be in a healthy state, the full fusion tracking control mode is triggered, and the fusion weld deviation is directly converted into welding tracking control parameters. When the sensor health status is determined to be in an interference state by the visual sensor, and the arc sensor, thermal imager, and acoustic sensor are not determined to be in a failure state, the downgraded fusion tracking control mode is triggered. The weight of the weld visual perception data in the fusion calculation is reduced, while the weight of the arc sensor perception data, thermal field distribution perception data, and welding acoustic perception data in the fusion calculation is increased. Based on the adjusted weights, the downgraded fusion weld deviation is recalculated and converted into welding tracking control parameters. When the sensor health status is determined to be in a failed state by the visual sensor, the vision-free guidance control mode is triggered. Historical weld path trajectory data is loaded, and the weld position is estimated by combining the instantaneous values ​​of welding current, welding voltage, and short-circuit frequency data in the arc sensor data. The estimated weld deviation is then converted into welding tracking control parameters.

4. The collaborative optimization analysis method for multi-station tower welding according to claim 1, characterized in that, Loading the multi-station welding heat source model also includes: Collect historical welding power data, historical welding torch position coordinates, and corresponding historical tower thermal imaging data for each workstation to construct a heat source model training set. Based on the double ellipsoidal heat source distribution function, an initial heat source model including a heat source superposition factor is constructed, where the heat source superposition factor is used to characterize the energy coupling effect of heat sources in adjacent workstations. Using historical tower thermal imaging data as supervision, and historical welding power data and historical welding torch position coordinates as input, the parameters of the initial heat source model are identified to obtain a multi-station welding heat source model.

5. The collaborative optimization analysis method for multi-station tower welding according to claim 1, characterized in that, Also includes: When the confidence scores of several sensors are all less than the preset confidence failure threshold, the sensor health status set is determined to be in a fully failed state, triggering the emergency shutdown protection mode, generating a shutdown alarm command and terminating the welding process.

6. The collaborative optimization analysis method for multi-station tower welding according to claim 1, characterized in that, Also includes: The historical welding process of tower welding was used to obtain the welding quality false alarm rate due to sensor failure and the welding quality false alarm rate due to environmental interference. Calculate the ratio of the welding quality false alarm rate to the welding quality false alarm rate, and set it as the confidence threshold adjustment factor; Based on the ratio of the preset baseline adjustment factor to the confidence threshold adjustment factor, the preset confidence warning threshold and the preset confidence failure threshold are adjusted.

7. The collaborative optimization analysis method for multi-station tower welding according to claim 1, characterized in that, Also includes: To obtain real-time welding progress data and real-time weld height detection data for each workstation in the tower welding scenario; When the real-time weld reinforcement detection data exceeds the preset reinforcement height threshold, the corresponding location is identified as a repair welding area; Based on real-time welding progress data, adjacent workstations that are idle or ahead in welding progress are scheduled to generate repair welding coordination instructions and perform online repair welding operations on the repair welding area.

Images