Welding control method, system and equipment of welding robot and medium

Abstract

The invention discloses a welding control method, system and equipment of a welding robot and a medium. The welding control method comprises the steps that an environment state vector is determined based on obtained environment data of an offshore converter; judging whether the marine environment is stable or not based on the environment state vector, and if the marine environment is stable, inputting the historical welding data and the environment state vector into a neural network model to obtain a residual stress prediction sequence; calculating a deviation vector based on the residual stress prediction sequence and a preset stress threshold value, generating a parameter optimization sequence based on the deviation vector, and fusing the attitude adjustment signal and the parameter optimization sequence to generate an optimized welding planning path; and according to the real-time welding data and the welding planning path, an attitude parameter fusion vector is generated, and a welding robot is driven to weld the offshore converter station based on the attitude parameter fusion vector. Therefore, accurate and self-adaptive regulation and control of the welding residual stress of the welding robot can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of offshore converter station, and particularly relates to a welding control method, system, device and medium of a welding robot. BACKGROUND

[0002] The offshore converter station bears dynamic loads such as wind, wave and current for a long time, and the welding joint is a weak link of the structure. If the residual stress generated in the welding process is too high or unevenly distributed, it will evolve into the source of fatigue crack initiation and propagation under the long-term action of alternating loads, and eventually may lead to catastrophic structural failure. Therefore, just completing the welding operation is far from enough, and the welding residual stress must be actively and accurately controlled to improve the fatigue resistance of the joint from the root, which is of great significance to ensure the reliability and life of the marine energy system.

[0003] The prior art mainly relies on welding robots with preset fixed parameters to perform automatic operation. This method may be effective in stable land environment, but has fundamental defects in complex marine environment. The main deficiency of the lack of accuracy is that the existing system lacks real-time sensing and response capability to key environmental parameters such as wind speed, sea wave and platform inclination, and cannot convert dynamic environmental changes into welding decision basis. When the platform is slightly tilted or vibrates due to wave impact, the welding robot with fixed parameters cannot adjust its posture and welding energy input in time, resulting in deviation of the actual welding path, out-of-control of heat input distribution, and thus causing poor weld formation and residual stress concentration. The result of such "blind" welding is the randomness and uncontrollability of residual stress distribution, which cannot meet the stringent requirements of offshore structures on welding quality and high-precision stress control. SUMMARY

[0004] The present application provides a welding control method, system, device and medium of a welding robot, which can accurately and adaptively regulate the welding residual stress of the welding robot in complex marine environment.

[0005] An embodiment of the present application provides a welding control method of a welding robot, comprising:

[0006] determining an environmental state vector based on the obtained environmental data of the offshore converter;

[0007] judging whether the marine environment is stable based on the environmental state vector, if the marine environment is stable, inputting the historical welding data and the environmental state vector into a preset neural network model to predict the residual stress distribution, and obtaining a residual stress prediction sequence;

[0008] a deviation vector is calculated based on the residual stress prediction sequence and a preset stress threshold, a parameter optimization sequence is generated based on the deviation vector, and an optimized welding planning path is generated by fusing a pose adjustment signal and the parameter optimization sequence, wherein the pose adjustment signal is generated according to the environment state vector;

[0009] According to the real-time welding data and the welding planning path, a pose parameter fusion vector is generated, and a welding robot is driven to weld the offshore converter station based on the pose parameter fusion vector.

[0010] The embodiment of the present application fuses environment data into a unified "environment state vector" that can be processed by an algorithm, providing a standardized input for subsequent intelligent decision-making; by judging whether the marine environment is stable based on the environment state vector, it ensures that subsequent precise control is only started under acceptable working conditions, avoiding invalid or harmful welding operations in severe environmental fluctuations. When the environment is stable, historical welding data and the environment state vector are input into a preset neural network model to predict residual stress distribution, fully utilizing the powerful nonlinear mapping capability of artificial intelligence technology to achieve prior and precise prediction of the welding thermal process and the final residual stress; by generating a parameter optimization sequence based on the deviation vector, the abstract stress exceeding problem is converted into a specific and executable welding process parameter correction amount; by fusing a pose adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, the spatial pose compensation of the robot is innovatively optimized in coordination with the welding process parameter adjustment, generating a comprehensive operation instruction that can simultaneously cope with mechanical disturbance and uneven heat input, thus planning a low-stress and high-quality welding trajectory from the source. Finally, by generating a pose parameter fusion vector according to real-time welding data and the welding planning path, and driving the welding robot, precise docking of the planning instruction and the robot body control is achieved at the final execution level, and the dynamic stability of the welding process is ensured in combination with real-time feedback. Compared with the prior art, the present application can accurately and adaptively regulate and control the welding residual stress of the welding robot in a complex marine environment.

[0011] Further, the generating of the parameter optimization sequence based on the deviation vector comprises:

[0012] If the deviation vector exceeds a preset deviation threshold, the deviation vector is input into a preset mapping function to calculate a correction amount of the welding parameters, wherein the welding parameters at least include welding current, welding voltage and welding speed;

[0013] The initial welding parameter set is updated based on the correction amount to generate the parameter optimization sequence containing the optimized parameters.

[0014] Thus, by generating the parameter optimization sequence based on the deviation vector, the abstract stress exceeding problem is converted into a specific and executable welding process parameter correction amount.

[0015] Further, the fusion of the posture adjustment signal and the parameter optimization sequence to generate the optimized welding planning path comprises:

[0016] Based on the platform tilt angle in the environment state vector, a tilt angle correction algorithm is used to generate a posture adjustment signal for compensating for the platform posture change, and based on the tilt angle correction data in the posture adjustment signal, the spatial coordinates of the preset welding path template are corrected to generate a trajectory correction sequence.

[0017] The welding parameters in the parameter optimization sequence are mapped to the corresponding path points of the trajectory correction sequence to obtain a mapping result, and the mapping result is fine-tuned based on the residual stress prediction sequence to generate a peak suppression sequence.

[0018] The trajectory correction sequence and the peak suppression sequence are fused to generate an initial welding planning path, and a prediction model is used to calculate a residual reduction vector prediction value corresponding to the initial welding planning path. If the residual reduction vector prediction value meets a first preset condition, the initial welding planning path is determined as the final welding planning path.

[0019] Thus, by fusing the posture adjustment signal and the parameter optimization sequence to generate the optimized welding planning path, the spatial posture compensation of the robot and the welding process parameter adjustment are synergistically optimized, and a comprehensive operation instruction that can cope with mechanical disturbance and uneven heat input is generated, thereby planning a low-stress and high-quality welding trajectory from the source.

[0020] Further, the calculation of the residual reduction vector prediction value corresponding to the initial welding planning path by the prediction model comprises:

[0021] The trajectory correction sequence and the peak suppression sequence are input into a preset random forest model, wherein the trajectory correction sequence contains spatial coordinate deviation data of the welding path points, and the peak suppression sequence contains intensity suppression data after welding parameter adjustment.

[0022] The coordinate deviation data and the intensity suppression data are fused by the random forest model based on a multi-decision tree ensemble learning algorithm, and the residual reduction vector prediction value corresponding to the initial welding planning path is output, wherein the residual reduction vector prediction value represents the expected reduction amount of the welding node residual stress.

[0023] Thus, by introducing a random forest prediction model, the residual stress control effect of the planned path is intelligently estimated before welding execution. It fuses multi-dimensional adjustment data such as trajectory correction and parameter suppression for analysis, quantitatively predicts the stress reduction, and thus realizes closed-loop verification and optimization screening of the welding path scheme. This effectively improves the predictability and accuracy of the control strategy, ensuring that the generated welding path can actively and reliably suppress residual stress in complex marine environments.

[0024] Further, the driving welding robot based on the posture parameter fusion vector to weld the offshore converter station, comprising:

[0025] Input the posture parameter fusion vector into the actuator of the welding robot to generate a preliminary motion instruction sequence to drive the welding robot to perform a welding action;

[0026] Obtain real-time trajectory data collected by a feedback sensor and spatial coordinate data provided by a node positioning system during welding, and fuse the real-time trajectory data and spatial coordinate data using a Kalman filter algorithm to obtain a real-time trajectory deviation sequence;

[0027] If the real-time trajectory deviation sequence exceeds a preset deviation threshold, update the posture parameter fusion vector according to the real-time trajectory deviation sequence until a second preset condition is met to determine a target posture parameter fusion vector to drive the welding robot to complete the welding work according to a stable trajectory sequence based on the target posture parameter fusion vector.

[0028] Thus, by fusing multiple real-time data sources, accurately perceiving the deviation of the welding trajectory from the planned trajectory, and dynamically updating the control instructions based on this, it ensures that the robot can actively offset dynamic disturbances such as wind and waves, platform inclination, etc. in actual operation, and ultimately makes the welding torch run along a stable trajectory, directly ensuring uniform heat input and controllable residual stress from the execution level.

[0029] Further, the judging whether the marine environment is stable based on the environment state vector, comprising:

[0030] Input the environment state vector into a trained classification model for classification processing to obtain an environment dynamic classification result;

[0031] Compare the environment dynamic classification result with a preset dynamic change threshold, and if the classification result does not exceed the threshold, determine that the marine environment is stable.

[0032] Thus, by judging whether the marine environment is stable based on the environment state vector, it ensures that subsequent precise control is only started under acceptable working conditions, avoiding invalid or harmful welding work in severe environmental fluctuations.

[0033] The welding control system of the welding robot comprises an acquisition module, a prediction module, a planning module and a welding module.

[0034] The acquisition module is configured to determine an environment state vector based on the acquired environment data of the offshore converter station.

[0035] The prediction module is configured to determine whether the marine environment is stable based on the environment state vector, and if the marine environment is stable, input the historical welding data and the environment state vector into a preset neural network model to predict the residual stress distribution and obtain a residual stress prediction sequence.

[0036] The planning module is configured to calculate a deviation vector based on the residual stress prediction sequence and a preset stress threshold, generate a parameter optimization sequence based on the deviation vector, and fuse a posture adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, wherein the posture adjustment signal is generated according to the environment state vector.

[0037] The welding module is configured to generate a posture parameter fusion vector according to real-time welding data and the welding planning path, and drive the welding robot to weld the offshore converter station based on the posture parameter fusion vector.

[0038] Further, the prediction module comprises:

[0039] The correction unit is configured to generate a posture adjustment signal for compensating for the posture change of the platform based on the platform inclination angle in the environment state vector through an inclination correction algorithm, correct the spatial coordinates of a preset welding path template based on the inclination correction data in the posture adjustment signal, and generate a trajectory correction sequence.

[0040] The mapping unit is configured to map the welding parameters in the parameter optimization sequence to the corresponding path points of the trajectory correction sequence to obtain a mapping result, and fine-tune the mapping result based on the residual stress prediction sequence to generate a peak suppression sequence.

[0041] The fusion unit is configured to fuse the trajectory correction sequence and the peak suppression sequence to generate an initial welding planning path, and calculate a residual reduction vector prediction value corresponding to the initial welding planning path by using a prediction model, and if the residual reduction vector prediction value meets a first preset condition, determine the initial welding planning path as the final welding planning path.

[0042] Another embodiment of the present application also provides a terminal device, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the steps of the welding control method of the welding robot are implemented.

[0043] Another embodiment of the present application also provides a computer readable storage medium item, comprising: a stored computer program, when the computer program is running, controlling the device where the computer readable storage medium is located to execute the steps of the welding control method of the welding robot. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the present application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0045] Figure 1 is a flowchart of an embodiment of the welding control method of the welding robot provided by the present application;

[0046] Figure 2 is a flowchart of an embodiment of steps S201 to S203 provided by the present application;

[0047] Figure 3 is a flowchart of an embodiment of steps S301 to S303 provided by the present application;

[0048] Figure 4 is a structural schematic diagram of an embodiment of the welding control system of the welding robot provided by the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above description of drawings are intended to cover non-exclusive inclusion.

[0051] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise explicitly specified and limited.

[0052] Reference herein to "embodiments" means that the particular features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily a separate or alternative embodiment to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0053] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects.

[0054] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two), and similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0055] In the description of the embodiments of the present application, unless otherwise explicitly specified and limited, the technical terms "mounting", "connection", "connection", "fixing" and the like should be understood in a broad sense, for example, it can be fixedly connected, or it can be detachably connected, or it can be integrated; it can be mechanical connection, or it can be electrical connection; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the internal communication of two elements or the interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0056] The welded joint of the offshore converter station bears dynamic loads such as wind, wave and current for a long time, which is a weak link of structural safety. If the residual stress is too high or unevenly distributed during welding, fatigue cracks are easily induced under alternating loads, threatening the overall structural safety. Existing welding robots mostly use fixed parameters for operation, lacking real-time sensing and response capability to dynamic factors such as wind speed, sea waves and platform inclination in complex marine environment, resulting in deviation of welding path, out-of-control of heat input and uncontrollable distribution of residual stress, which cannot meet the quality requirements of high-precision welding of offshore structures.

[0057] Reference is made to Figure 1To realize accurate and adaptive control of welding residual stress of a welding robot in a complex marine environment, an embodiment of the present application provides a welding control method of a welding robot, comprising steps S101 to S104.

[0058] In step S101, an environment state vector is determined based on the acquired environment data of the offshore converter.

[0059] In some embodiments, the environment data of the offshore converter is acquired, specifically, wind speed data, sea wave height and inclination angle in the marine environment are collected in real time through a sensor network, wherein the sensor network comprises a wind speed sensor, a sea wave height sensor and an inclination sensor.

[0060] It should be noted that each sensor is installed at a density of one every 10 meters to form a monitoring area with a coverage radius of 50 meters, and the original signals of instantaneous wind speed, peak wave height and platform inclination angle are collected synchronously at a sampling frequency of 5 Hz.

[0061] In some embodiments, the environment state vector is determined based on the environment data, specifically, the collected environment data is subjected to data fusion processing by using a Kalman filtering algorithm to obtain the environment state vector, wherein firstly, a state vector x_k composed of [wind speed, sea wave height, inclination angle] is initialized, the state transition matrix F is set as an identity matrix to capture the dynamic changes of the environment, the observation matrix H is a full observation matrix, and the process noise w_k and the observation noise v_k are configured, then the state prediction equation x_k=Fx_{k-1}+w_{k-1} and the observation update equation z_k=H x_k+v_k are iteratively executed, wherein k is a time index, z_k is an observation vector for optimal estimation and filtering of the original signal, and finally a unified environment state vector x_{k|k} such as [12.3 m / s, 2.7 m, 3.0 degrees] that eliminates random noise interference and can reliably represent the current marine environment is output, and the specific iteration process is not the focus of the present application and will not be expanded here.

[0062] In step S102, it is judged whether the marine environment is stable based on the environment state vector, if the marine environment is stable, historical welding data and the environment state vector are input into a preset neural network model for residual stress distribution prediction to obtain a residual stress prediction sequence.

[0063] In some embodiments, determining whether the marine environment is stable based on the environmental state vector comprises: inputting the environmental state vector into a trained classification model for classification processing to obtain an environmental dynamic classification result; and comparing the environmental dynamic classification result with a preset dynamic change threshold value, and if the classification result does not exceed the threshold value, determining that the marine environment is stable. Specifically, first, the fused environmental state vector is input into a pre-trained support vector machine classification model for classification processing, and then the model calculates a classification score according to a decision function; subsequently, the classification score is compared with a preset dynamic change threshold value (for example, 0.7): if the classification score does not exceed the threshold value, it is determined that the current marine environment is in a stable state, and a "stable" or "safe" label is output; if the classification score exceeds the threshold value (for example, 0.85>0.7), an alarm signal is triggered, indicating an abnormal environmental condition.

[0064] It should be noted that when the classification model determines that the marine environment is unstable, the system will not perform precise welding control based on neural network prediction. At this time, the control system will adopt one of the following strategies: (1) suspend the operation: send a suspension instruction to the welding robot, so that it stays in a safe position, and restart the perception-decision process after the environment returns to a stable state. (2) enable the degradation mode: switch to a basic welding mode based on preset fixed parameters. Although this mode does not perform real-time optimization of residual stress, it includes the most basic environmental posture compensation to complete the necessary welding task under the premise of ensuring the minimum welding quality.

[0065] Through the above strategies, the integrity of the system's decision-making and the safety of the operation in the complex and variable marine environment are ensured.

[0066] It should be noted that the classification model is constructed using a radial basis function kernel K(x, y) = exp(-γ||x-y||^2), where K(x, y) is a radial basis function (RBF) kernel function, x and y represent two input sample vectors, ||x-y|| is the Euclidean distance, and γ is a kernel function coefficient, for example, 0.5. The penalty parameter C = 10 is determined by grid search optimization, and the model is trained based on a historical data set containing 1000 labeled samples (classified into "stable" and "unstable" categories).

[0067] In this way, by determining whether the marine environment is stable based on the environmental state vector, it is ensured that the subsequent precise control is only started under acceptable working conditions, avoiding invalid or harmful welding operations in severe environmental fluctuations.

[0068] In some embodiments, the historical welding data and the environmental state vector are input into a preset neural network model for residual stress distribution prediction to obtain a residual stress prediction sequence. Specifically, first, historical welding data is called from a database, and the data at least includes welding current, welding voltage, welding speed and other key process parameters; then, the environmental state vector and the historical welding data are normalized and spliced into a comprehensive feature vector (for example, [wind speed value, wave height value, inclination angle, current value, voltage value, speed value]); and then, the comprehensive feature vector is input into a pre-trained deep neural network model for forward inference to output the residual stress prediction sequence.

[0069] It should be noted that the neural network model usually adopts a structure including multiple fully connected layers (for example, three layers, each layer having 256 neurons), and uses a ReLU activation function to introduce nonlinearity, and the output layer corresponds to the residual stress values of multiple key nodes on the welding path. The model is obtained after a large amount of historical data (such as 5000 samples) is trained, and an Adam optimizer is used to minimize the error between the predicted stress and the measured stress. The specific training process is not the focus of the present application, and therefore will not be expanded here.

[0070] In step S103, a deviation vector is calculated based on the residual stress prediction sequence and a preset stress threshold, a parameter optimization sequence is generated based on the deviation vector, and an optimized welding planning path is generated by fusing a posture adjustment signal and the parameter optimization sequence, wherein the posture adjustment signal is generated according to the environmental state vector.

[0071] In some embodiments, the deviation vector is calculated based on the residual stress prediction sequence and a preset stress threshold. Specifically, after obtaining the residual stress prediction value, the peak stress in the residual stress prediction sequence is compared with a preset stress threshold (which can be freely set and is not limited in the present application) to calculate the difference between the peak stress in the residual stress prediction sequence and the ideal state (that is, the preset stress threshold), that is, to calculate the deviation vector.

[0072] In some embodiments, the generating the parameter optimization sequence based on the deviation vector comprises: if the deviation vector exceeds a preset deviation threshold, inputting the deviation vector into a preset mapping function to calculate a correction amount of the welding parameter, wherein the welding parameter at least includes welding current, welding voltage and welding speed; updating an initial welding parameter set based on the correction amount to generate the parameter optimization sequence containing the optimized parameter. Specifically, first, it is judged whether the modulus of the deviation vector or the maximum value (such as 40 MPa) in it exceeds a preset deviation threshold (for example, 20 MPa), if it exceeds, the deviation vector is input into a preset mapping function; then, the correction amount of the welding parameter can be calculated through the mapping function, for example, the current needs to be increased by 0.02*40=0.8A. Finally, based on the calculated correction amount, the initial welding parameter set (such as current 200A, voltage 22V, speed 0.8m / min) is updated, thereby generating the parameter optimization sequence containing the optimized parameter (such as current 200.8A), and the sequence is used for subsequent welding path planning to realize accurate regulation of residual stress.

[0073] It should be noted that the mapping function defines the mathematical relationship between the stress deviation and the welding parameter correction amount, and the function can be fitted by historical experimental data, and the form is not limited to linear, for example, a linear proportional relationship f(Δ)=k*Δ, wherein f is the welding parameter correction amount, Δ is the stress deviation vector, and the proportional coefficient k is not a fixed empirical value, but is learned from historical data through regression analysis. Specifically, the system maintains a historical database containing a large number of "stress deviation vector-optimized parameter correction amount" paired samples, and through linear regression analysis on the data set, the error between the predicted correction amount and the actual optimal correction amount is minimized, thereby fitting the specific value of the coefficient k. The specific determination process is not the focus of the present application, and therefore will not be expanded here.

[0074] It should be noted that the parameter optimization sequence refers to the basic parameter that needs to be adjusted in the welding process, and the purpose is to adjust the overall average heat input of the welding process to a suitable range, so that the predicted residual stress level is below the preset stress threshold, and the peak suppression sequence is more fine and point-to-point parameter compensation for the local stress concentration point (peak) identified in the prediction model, to realize the uniformization of residual stress in space.

[0075] In this way, by generating the parameter optimization sequence based on the deviation vector, the abstract stress exceeding problem is converted into a specific and executable welding process parameter correction amount.

[0076] Please refer to Figure 2In some embodiments, the fusing of the posture adjustment signal and the parameter optimization sequence to generate an optimized welding planning path comprises steps S201-S203:

[0077] In step S201, a posture adjustment signal for compensating for the posture change of the platform is generated by a tilt correction algorithm based on the platform tilt angle in the environment state vector, and the spatial coordinates of the preset welding path template are corrected based on the tilt correction data in the posture adjustment signal to generate a trajectory correction sequence.

[0078] In some embodiments, the platform tilt angle in the environment state vector and other parameters (such as wind speed) in the environment state vector are input into the tilt correction algorithm to calculate the tilt correction coefficients of each axis (for example, for a 2.5-degree tilt, the correction coefficient is calculated as 0.5) by the inverse sine function. This series of calculated correction coefficients constitutes the posture adjustment signal for compensating for the dynamic shaking of the platform. Then, the coordinate values of each path point are dynamically adjusted by a spatial coordinate transformation function according to the correction coefficients, and by performing such real-time and adaptive coordinate transformation on all points in the path template, a trajectory correction sequence is finally generated which can offset the influence of platform tilt and ensure the accurate relative position of the welding torch and the weld.

[0079] It should be noted that the tilt correction algorithm is specifically implemented by constructing a spatial coordinate homogeneous transformation matrix. According to the roll angle and pitch angle of the platform in the environment state vector, the coordinates of any point on the preset welding path template are compensated in real time to calculate the corrected coordinates.

[0080] In step S202, the welding parameters in the parameter optimization sequence are mapped to the corresponding path points of the trajectory correction sequence to obtain a mapping result, and the mapping result is fine-tuned based on the residual stress prediction sequence to generate a peak suppression sequence.

[0081] In some embodiments, first, the parameter optimization sequence is distributed to the corresponding path points of the trajectory correction sequence by a mapping algorithm to establish a corresponding relationship between each path point and a set of optimized welding parameters, and a mapping result is obtained. Then, in order to actively suppress the residual stress peak value of welding, the mapping result is fine-tuned based on the residual stress prediction sequence obtained as described above. Specifically, if the predicted stress value sequence indicates that there is a risk of excessively high stress (i.e., peak value) in a certain area, the welding parameters of the corresponding path points in the risk area will be fine-tuned by a data fusion algorithm (for example, the current is appropriately reduced or the welding speed is increased to reduce the heat input), thereby generating a peak suppression sequence specially used for balancing heat distribution and suppressing stress concentration.

[0082] Step S203, fuse the trajectory correction sequence and the peak suppression sequence to generate an initial welding planning path, and calculate a residual reduction vector prediction value corresponding to the initial welding planning path by using a prediction model, and if the residual reduction vector prediction value meets a first preset condition, determine the initial welding planning path as the final welding planning path.

[0083] In some embodiments, the trajectory correction sequence and the peak suppression sequence are fused to generate an initial welding planning path, specifically: first, create a data structure containing spatial coordinates and process parameters for each path point in the trajectory correction sequence; then, assign the optimized welding parameters (including welding current, voltage and speed) of the corresponding path points in the peak suppression sequence to the process parameter field in the data structure, complete the accurate mapping of welding parameters on the spatial path, and at the same time, to realize the deep coupling of the two sequences, the system uses a data fusion algorithm (such as weighted average method) to synchronize the optimization of the spatial coordinates and the welding parameters of the path points: for the path points with large coordinate offset due to platform tilt correction, the system will adjust the welding current and speed accordingly based on the heat input distribution model to prevent new stress concentration in the area due to uneven heat accumulation; on the contrary, for the path points marked as needing to focus on suppressing stress peaks in the peak suppression sequence, the system will also make a slight smoothness adjustment to their spatial coordinates while applying stronger welding parameter control to ensure the motion stability of the robot when executing this section of path. Through the coordinated adjustment and integrated assignment of coordinates and parameters, a complete initial welding planning path integrating spatial trajectory compensation and process parameter optimization is finally generated.

[0084] In some embodiments, the step of calculating the residual reduction vector prediction value corresponding to the initial welding planning path by using the prediction model comprises: inputting the trajectory correction sequence and the peak suppression sequence into a preset random forest model, wherein the trajectory correction sequence contains spatial coordinate deviation data of the welding path points, and the peak suppression sequence contains intensity suppression data after adjusting the welding parameters; and performing fusion processing on the coordinate deviation data and the intensity suppression data by using the random forest model based on a multi-decision tree ensemble learning algorithm, and outputting the residual reduction vector prediction value corresponding to the initial welding planning path, wherein the residual reduction vector prediction value represents an expected reduction amount of the welding node residual stress. Specifically, first, the trajectory correction sequence and the peak suppression sequence are input as input features into the preset random forest model; the random forest model performs fusion and regression analysis on the input coordinate deviation data and intensity suppression data based on the ensemble learning algorithm of the multi-decision tree, and outputs a multi-dimensional residual reduction vector prediction value through a multi-tree voting mechanism, wherein the prediction value quantifies the expected reduction amount of the residual stress in different regions of the welding node after executing the current initial welding planning path, thereby providing a forward-looking evaluation basis for the optimization effect of the path planning.

[0085] It should be noted that the trajectory correction sequence contains spatial coordinate deviation data of the welding path points caused by platform posture compensation, and the peak suppression sequence contains intensity suppression data formed by adjusting the welding parameters (such as current, speed) for balancing the heat input distribution and suppressing stress concentration.

[0086] It should be noted that the training of the random forest model depends on a high-quality historical data set. The data set is constructed by more than 5000 groups of offshore welding process tests, and each group of data includes but is not limited to: environmental state vector (wind speed, wave height, inclination), welding process parameter sequence (current, voltage, speed), and residual stress value sequence measured at multiple key nodes of the welding joint by X-ray diffraction method or blind hole method. Before model training, the data needs to be normalized. The random forest model takes the features of the trajectory correction sequence and the peak suppression sequence as input, and takes the measured stress reduction amount under the corresponding process as output label for training.

[0087] In this way, by introducing the random forest prediction model, the residual stress control effect of the planning path before welding execution is intelligently estimated. It performs fusion analysis on multi-dimensional adjustment data such as trajectory correction and parameter suppression, quantifies and predicts the stress reduction amount, thereby realizing closed-loop verification and optimization screening of the welding path scheme. This effectively improves the predictability and accuracy of the regulation strategy, and ensures that the generated welding path can actively and reliably suppress the residual stress in the complex marine environment.

[0088] In some embodiments, if the residual reduction vector prediction value meets the first preset condition, the initial welding planning path is determined as the final welding planning path, specifically: since the residual reduction vector prediction value output by the random forest model is a multi-dimensional vector, each dimension of which represents the expected residual stress reduction of a key evaluation point on the welding path; therefore, the vector is compared with a preset qualified threshold vector element by element, and the overall length of the residual reduction vector prediction value is calculated and compared with a preset overall stress improvement threshold. Only when the above two conditions are met at the same time, it is determined that it meets the first preset condition. Once the above conditions are met, the system confirms the initial welding planning path as the final welding planning path and issues it to the welding robot for execution. If either condition is not met, the system takes this result as negative feedback, re-triggers the process of collecting the environmental state vector, and performs a new round of path planning and optimization iteration until a qualified welding path is generated or the maximum number of iterations is reached.

[0089] It should be noted that the first preset condition means that the values of each dimension of the residual reduction vector prediction value are not lower than the corresponding qualified threshold, and the overall length thereof exceeds the preset overall stress improvement threshold.

[0090] It should be noted that the qualified threshold vector is preset according to the safety specifications and material characteristics of the welding node, which is not limited in the present application.

[0091] In this way, the optimized welding planning path is generated by fusing the posture adjustment signal and the parameter optimization sequence, which innovatively cooperates the spatial posture compensation of the robot with the welding process parameter adjustment to generate a comprehensive operation instruction that can cope with mechanical disturbance and uneven heat input, and plans a low-stress and high-quality welding trajectory from the source.

[0092] Step S104, generating a posture parameter fusion vector according to the real-time welding data and the welding planning path, and driving the welding robot to weld the offshore converter station based on the posture parameter fusion vector.

[0093] In some embodiments, the posture parameter fusion vector is generated according to the real-time welding data and the welding planning path, specifically: the optimized welding planning path generated in the foregoing link is weighted and fused with the real-time welding data such as welding current and voltage collected by the sensor in real time during the welding process of the welding robot to generate a unified posture parameter fusion vector V that can be directly analyzed by the execution mechanism, wherein the fusion formula is: V = α·C + β·P, wherein C is a tilt angle correction coefficient sequence, P is a parameter optimization sequence, α and β are preset weight coefficients, and α + β = 1, for example, α = 0.6, β = 0.4, to balance the contributions of posture compensation and process parameter adjustment.

[0094] Please refer to Figure 3 In some embodiments, the welding robot is driven to weld the offshore converter station based on the pose parameter fusion vector, including steps S301 to S303.

[0095] Step S301, input the pose parameter fusion vector to the actuator of the welding robot, generate a preliminary motion instruction sequence to drive the welding robot to perform a welding action;

[0096] In some embodiments, after generating the pose parameter fusion vector, it is input to the bottom layer actuator of the welding robot to decode the vector into a specific preliminary motion instruction sequence to drive the joints of the robot and the welding torch to start the welding action.

[0097] Step S302, obtain real-time trajectory data collected by a feedback sensor during welding and spatial coordinate data provided by a node positioning system, and fuse the real-time trajectory data and spatial coordinate data using a Kalman filtering algorithm to obtain a real-time trajectory deviation sequence;

[0098] In some embodiments, in order to deal with the trajectory deviation caused by the continuous micro-motion of the platform in the marine environment, high-precision closed-loop control needs to be started synchronously during welding. Specifically, the real-time trajectory data of the robot is collected by a feedback sensor (such as an encoder, a vision sensor), and the accurate spatial coordinate data provided by a node positioning system (such as a laser tracker) is combined; then, the Kalman filtering algorithm is used to fuse the above two kinds of data to optimally estimate the real-time trajectory deviation sequence between the actual trajectory of the robot and the planned trajectory, thereby effectively filtering out measurement noise and obtaining accurate deviation information.

[0099] Step S303, if the real-time trajectory deviation sequence exceeds a preset deviation threshold, update the pose parameter fusion vector according to the real-time trajectory deviation sequence until a second preset condition is met to determine a target pose parameter fusion vector, and drive the welding robot to complete the welding work according to a stable trajectory sequence based on the target pose parameter fusion vector.

[0100] It should be noted that the second preset condition means that the real-time trajectory deviation sequence is lower than the preset deviation threshold in a plurality of consecutive control periods.

[0101] In some embodiments, the system will continuously determine whether the above real-time trajectory deviation sequence exceeds the preset deviation threshold, and once it exceeds, it indicates that the current welding trajectory is unstable due to environmental interference. At this time, the system will not continue to execute the original instructions, but will immediately trigger the adjustment mechanism, that is, the pose parameter fusion vector is updated in reverse according to the real-time trajectory deviation sequence. The adjustment is an iterative optimization process, which will recalculate the components of the fusion vector based on the deviation amount, generate and execute new control instructions, then collect data again for evaluation, until the trajectory deviation is stable within the allowed range, meeting the second preset condition (such as the deviation of several consecutive periods being lower than the threshold). At this time, the determined vector is the target pose parameter fusion vector, and the robot will drive the welding torch to complete the welding work with high quality and low residual stress based on the finally optimized vector along the stable trajectory sequence.

[0102] In this way, by fusing multiple real-time data sources, the deviation of the welding trajectory from the planned trajectory is accurately perceived, and the control instructions are dynamically and iteratively updated based on this. This ensures that the robot can actively offset dynamic disturbances such as wind and platform tilt in actual operation, and ultimately makes the welding torch run along a stable trajectory, directly ensuring uniform heat input and controllable residual stress from the execution level.

[0103] The embodiment of the present application fuses environmental data into a unified "environmental state vector" that can be processed by an algorithm, providing a standardized input for subsequent intelligent decision-making; by judging whether the marine environment is stable based on the environmental state vector, it ensures that subsequent precise control is only started under acceptable working conditions, avoiding invalid or harmful welding operations in severe environmental fluctuations. When the environment is stable, by inputting historical welding data and the environmental state vector into a preset neural network model for residual stress distribution prediction, the powerful non-linear mapping capability of artificial intelligence technology is fully utilized to achieve prior and accurate prediction of the welding thermal process and the final residual stress; by generating a parameter optimization sequence based on the deviation vector, the abstract stress exceeding problem is converted into specific and executable welding process parameter correction; by fusing the pose adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, the spatial pose compensation of the robot and the adjustment of the welding process parameters are synergistically optimized, generating a comprehensive operation instruction that can simultaneously cope with mechanical disturbance and uneven heat input, planning a low-stress and high-quality welding trajectory from the source. Finally, by generating a pose parameter fusion vector based on real-time welding data and the welding planning path, and driving the welding robot, the precise docking of the planning instruction and the robot body control is achieved at the final execution level, and the dynamic stability of the welding process is ensured combined with real-time feedback. Compared with the prior art, the present application can accurately and adaptively control the welding residual stress of the welding robot in a complex marine environment.

[0104] As Figure 4As shown, on the basis of the above method embodiment, corresponding device embodiments are provided.

[0105] An embodiment of the present application provides a welding control system of a welding robot, comprising an acquisition module 100, a prediction module 200, a planning module 300 and a welding module 400.

[0106] The acquisition module 100 is used for determining an environment state vector based on acquired environment data of a marine converter.

[0107] The prediction module 200 is used for judging whether the marine environment is stable based on the environment state vector, and if the marine environment is stable, inputting historical welding data and the environment state vector into a preset neural network model to perform residual stress distribution prediction and obtain a residual stress prediction sequence.

[0108] The planning module 300 is used for calculating a deviation vector based on the residual stress prediction sequence and a preset stress threshold, generating a parameter optimization sequence based on the deviation vector, and fusing a posture adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, wherein the posture adjustment signal is generated according to the environment state vector.

[0109] The welding module 400 is used for generating a posture parameter fusion vector according to real-time welding data and the welding planning path, and driving the welding robot to perform welding on the marine converter station based on the posture parameter fusion vector.

[0110] It can be understood that the above device embodiment is corresponding to the method embodiment of the present application, and can realize the welding control method of the welding robot provided by any one of the above method embodiments.

[0111] It should be noted that the device embodiments described above are only schematic, and part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the device embodiment provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which can be realized as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0112] On the basis of the above welding control method of the welding robot, another embodiment of the present application provides a terminal device, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the welding control method of the welding robot of any one of the embodiments of the present application is realized.

[0113] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more module elements can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the terminal device.

[0114] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The terminal device can include, but is not limited to, a processor and a memory.

[0115] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and connects all parts of the terminal device through various interfaces and lines.

[0116] On the basis of the above-mentioned method embodiment, another embodiment of the present application provides a computer readable storage medium, including a stored computer program, wherein when the computer program runs, the device where the computer readable storage medium is located executes the welding control method of the welding robot as described in any one of the above-mentioned method embodiments of the present application.

[0117] The modules / units integrated in the device / terminal equipment, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0118] The above is the preferred embodiment of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. A welding control method of a welding robot, characterized by, The method comprises the following steps: determining an environment state vector based on the obtained environmental data of the offshore converter station; judging whether the marine environment is stable based on the environment state vector, if the marine environment is stable, inputting the historical welding data and the environment state vector into a preset neural network model to predict the residual stress distribution and obtain a residual stress prediction sequence; calculating a deviation vector based on the residual stress prediction sequence and a preset stress threshold, generating a parameter optimization sequence based on the deviation vector, and fusing a posture adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, wherein the posture adjustment signal is generated according to the environment state vector; generating a posture parameter fusion vector according to real-time welding data and the welding planning path, and driving the welding robot to weld the offshore converter station based on the posture parameter fusion vector.

2. The welding control method of the welding robot according to claim 1, characterized by, The method further comprises the following steps: if the deviation vector exceeds a preset deviation threshold, inputting the deviation vector into a preset mapping function to calculate a correction amount of the welding parameters, wherein the welding parameters at least include welding current, welding voltage and welding speed; updating the initial welding parameter set based on the correction amount to generate the parameter optimization sequence containing optimized parameters.

3. The welding control method of the welding robot according to claim 1, characterized by, The method further comprises the following steps: based on the platform inclination angle in the environment state vector, generating a posture adjustment signal for compensating the platform posture change through an inclination correction algorithm, correcting the spatial coordinates of a preset welding path template based on the inclination correction data in the posture adjustment signal to generate a trajectory correction sequence; mapping the welding parameters in the parameter optimization sequence to the corresponding path points of the trajectory correction sequence to obtain a mapping result, and fine-tuning the mapping result based on the residual stress prediction sequence to generate a peak suppression sequence; fusing the trajectory correction sequence and the peak suppression sequence to generate an initial welding planning path, and using a prediction model to calculate a residual reduction vector prediction value corresponding to the initial welding planning path, if the residual reduction vector prediction value meets a first preset condition, the initial welding planning path is determined as the final welding planning path.

4. The welding control method of the welding robot according to claim 3, characterized by, The method further comprises the following steps: inputting the trajectory correction sequence and the peak suppression sequence into a preset random forest model, wherein the trajectory correction sequence contains spatial coordinate deviation data of the welding path points, and the peak suppression sequence contains intensity suppression data of the welding parameters after adjustment; integrating and processing the coordinate deviation data and the intensity suppression data based on a multi-decision tree ensemble learning algorithm through the random forest model to output the residual reduction vector prediction value corresponding to the initial welding planning path, wherein the residual reduction vector prediction value represents the expected reduction amount of the welding node residual stress.

5. The welding control method of the welding robot according to claim 1, characterized by, The method further comprises the following steps: Input the posture parameter fusion vector into an execution mechanism of the welding robot to generate a preliminary motion instruction sequence to drive the welding robot to perform the welding action; Obtain real-time trajectory data collected by a feedback sensor and spatial coordinate data provided by a node positioning system during the welding process, and fuse the real-time trajectory data and the spatial coordinate data by using a Kalman filtering algorithm to obtain a real-time trajectory deviation sequence; If the real-time trajectory deviation sequence exceeds a preset deviation threshold, update the posture parameter fusion vector according to the real-time trajectory deviation sequence until a second preset condition is met, determine a target posture parameter fusion vector, and drive the welding robot to complete the welding work according to a stable trajectory sequence based on the target posture parameter fusion vector.

6. The welding control method of the welding robot according to any one of claims 1 to 5, characterized by, The method further includes: Input the environment state vector into a trained classification model for classification processing to obtain an environment dynamic classification result; Compare the environment dynamic classification result with a preset dynamic change threshold, and if the classification result does not exceed the threshold, determine that the marine environment is stable.

7. A welding control system for a welding robot, characterized in that The method further includes: An obtaining module, a prediction module, a planning module, and a welding module; The obtaining module is configured to determine an environment state vector based on obtained environment data of the offshore converter station; The prediction module is configured to determine whether the marine environment is stable based on the environment state vector, and if the marine environment is stable, input historical welding data and the environment state vector into a preset neural network model for residual stress distribution prediction to obtain a residual stress prediction sequence; The planning module is configured to calculate a deviation vector based on the residual stress prediction sequence and a preset stress threshold, generate a parameter optimization sequence based on the deviation vector, and fuse a posture adjustment signal and the parameter optimization sequence to generate an optimized welding planning path, wherein the posture adjustment signal is generated according to the environment state vector; The welding module is configured to generate a posture parameter fusion vector based on real-time welding data and the welding planning path, and drive the welding robot to weld the offshore converter station based on the posture parameter fusion vector.

8. The welding control system of the welding robot according to claim 7, characterized in that, The prediction module includes: A correction unit configured to generate a posture adjustment signal for compensating for a change in the posture of the platform based on a platform inclination angle in the environment state vector by using an inclination correction algorithm, correct spatial coordinates of a preset welding path template based on inclination correction data in the posture adjustment signal to generate a trajectory correction sequence; A mapping unit configured to map welding parameters in the parameter optimization sequence to corresponding path points of the trajectory correction sequence to obtain a mapping result, and fine-tune the mapping result based on the residual stress prediction sequence to generate a peak suppression sequence; A fusion unit configured to fuse the trajectory correction sequence and the peak suppression sequence to generate an initial welding planning path, and calculate a residual reduction vector prediction value corresponding to the initial welding planning path by using a prediction model, and if the residual reduction vector prediction value meets a first preset condition, determine the initial welding planning path as the final welding planning path.

9. A terminal device, comprising: The method further includes: one or more processors; a memory coupled to the processors for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement steps of the welding control method of the welding robot according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, comprising: a stored computer program, wherein when the computer program runs, the device where the computer readable storage medium is located performs steps of the welding control method of the welding robot according to any one of claims 1-7.

Images