Automatic welding robot based on artificial intelligence and welding system thereof

Abstract

The invention relates to the technical field of artificial intelligence welding, and discloses an automatic welding robot based on artificial intelligence and a welding system thereof. The system comprises a welding feature acquisition module, a process parameter generation module, a real-time regulation and control module and a motion planning module. The welding characteristic acquisition module is used for acquiring three-dimensional contour data, material component information and welding seam geometric parameters of a workpiece to be welded, and generating a welding characteristic spectrum through characteristic fusion; a process parameter generation module retrieves a matching template from a welding knowledge base according to the parameters, and outputs a reference welding process parameter combination in combination with an environment temperature and humidity compensation coefficient; the real-time regulation and control module dynamically corrects the reference parameters to generate an optimization instruction according to the molten pool form, thermal radiation distribution and electric arc voiceprint characteristics in the welding process; and the motion planning module calculates a motion track, attitude parameters and a speed curve of the welding execution mechanism according to the optimization instruction to form a robot control instruction set. The system can improve the intelligent level of welding, guarantees stable welding quality, and is suitable for welding scenes of multiple manufacturing industries.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence welding, in particular to an automatic welding robot based on artificial intelligence and a welding system thereof. BACKGROUND

[0002] In modern manufacturing, welding as a key connection process is widely used in mechanical manufacturing, automobile production, shipbuilding, aerospace and other fields. With the continuous improvement of the requirements of industrial production on product precision, quality stability and production efficiency, the traditional welding method gradually exposes many shortcomings. The traditional welding operation highly depends on the experience and skills of the operator, and the difference in technical level of different operators will directly lead to the fluctuation of welding quality, especially when dealing with complex structure workpieces or special material workpieces, it is difficult to guarantee the consistency and reliability of the weld. In the preparation stage before workpiece welding, the acquisition of workpiece three-dimensional profile, material composition and weld geometry parameters in the traditional way is mostly manual measurement or single detection equipment, the measurement efficiency is low and the data accuracy is easily affected by environmental factors, it is difficult to form comprehensive and accurate workpiece welding feature information, which further affects the rationality of the subsequent process parameter setting. In the process parameter determination link, the traditional method usually relies on the preset fixed parameter template, or adjusts the parameters through multiple trial weldings, which not only consumes a lot of manpower and material cost, but also cannot dynamically compensate according to the real-time environmental temperature and humidity changes, resulting in poor stability of welding quality under different environmental conditions. In the welding process, the traditional system lacks effective monitoring and analysis means for key real-time features such as molten pool shape, heat radiation distribution and arc soundprint, and cannot timely find abnormal conditions in the welding process and adjust the process parameters, which is easy to cause weld defects such as pores, cracks, incomplete fusion and other problems. At the same time, the motion trajectory planning of the welding execution mechanism is mostly based on the pre-set fixed path, which is difficult to flexibly optimize the motion trajectory, posture and speed according to the dynamically adjusted process parameters, further limiting the improvement of welding precision and efficiency. With the transformation of manufacturing industry to intelligent and automatic direction, the traditional welding technology has been unable to meet the production demand of high precision, high efficiency and high stability, and an automatic welding system integrating multi-dimensional feature acquisition, intelligent process parameter generation, real-time dynamic regulation and control and flexible motion planning is needed to solve the problems of low efficiency, large quality fluctuation and strong dependence on manual work in the traditional welding process. SUMMARY

[0003] The present application aims to provide an automatic welding robot based on artificial intelligence and a welding system thereof to solve the problems raised in the background art.

[0004] To achieve the above-mentioned purpose, the present application provides an automatic welding system based on artificial intelligence, which comprises: The welding feature acquisition module is configured to acquire three-dimensional contour data, material composition information, and weld geometry parameters of a workpiece to be welded, and generate a workpiece welding feature map through feature fusion technology. The process parameter generation module is configured to retrieve a matched process parameter template from a welding knowledge base based on the workpiece welding feature map, combine an ambient temperature and humidity compensation coefficient, and output a reference welding process parameter combination. The real-time regulation module is configured to dynamically correct the reference welding process parameter combination according to a molten pool shape feature, a thermal radiation distribution, and an arc voiceprint feature in a welding process, and generate an optimized welding process parameter instruction. The motion planning module is configured to calculate a motion trajectory, a posture adjustment parameter, and a speed control curve of a welding execution mechanism according to the optimized welding process parameter instruction, and form a robot control instruction set.

[0005] Preferably, the generation process of the workpiece welding feature map comprises: curvature features and normal vectors of each curved surface segment are extracted through surface segmentation processing of the three-dimensional contour data, and a geometric feature matrix is established; an element content ratio in the material composition information is analyzed, and a material welding performance index is calculated; a width variation rate and an angle deviation value of the weld geometry parameters are measured, and a weld feature vector is constructed; the geometric feature matrix, the material welding performance index, and the weld feature vector are fused to generate the workpiece welding feature map.

[0006] Preferably, the determination process of the reference welding process parameter combination comprises: a material welding performance index in the workpiece welding feature map is used to filter process parameter templates in the welding knowledge base that have a material matching degree reaching a preset threshold; a width variation rate of the weld feature vector is used to adjust a current fluctuation range in the process parameter template; a voltage reference value in the process parameter template is corrected in combination with an ambient temperature and humidity compensation coefficient to form the reference welding process parameter combination.

[0007] Preferably, the generation process of the optimized welding process parameter instruction comprises: a contour change feature of the molten pool is captured through a high-speed camera system, and a molten pool stability index is calculated; temperature gradient data of a thermal radiation distribution are acquired using an infrared temperature measurement system, and a heat affected zone model is established; an energy distribution spectrum of an arc voiceprint feature is collected, and an arc abnormal state is identified; a key parameter in the reference welding process parameter combination is adjusted according to the molten pool stability index, the heat affected zone model, and the arc abnormal state.

[0008] Preferably, the calculation process of the molten pool stability index further comprises: analyzing the width fluctuation rate and length consistency in the molten pool profile change characteristics; extracting the distribution rule and amplitude characteristics of the molten pool surface ripples; combining the width fluctuation rate, length consistency and surface ripple characteristics to calculate the molten pool stability index.

[0009] Preferably, the forming process of the robot control instruction set comprises: analyzing the welding speed requirement in the optimized welding process parameter instruction and decomposing it into an axial movement component and a radial compensation amount; planning the approach path of the welding execution mechanism according to the three-dimensional profile data in the workpiece welding feature map; dynamically adjusting the posture control parameters based on the real-time changes of the weld geometry parameters; integrating the movement component, approach path and posture parameters to form the robot control instruction set.

[0010] Preferably, the dynamic adjustment process of the posture control parameters further comprises: monitoring the real-time width change of the weld geometry parameters; calculating the included angle deviation between the welding torch axis and the weld centerline; determining the compensation amount of the posture adjustment according to the width change and the included angle deviation; updating the pitch angle and yaw angle parameters of the welding execution mechanism.

[0011] Preferably, the system further comprises: a quality prediction module, which establishes a quality risk prediction model based on the correlation between defect features and process parameters in historical welding data; real-time analysis of the molten pool stability index and the heat affected zone model, and triggering the process parameter emergency intervention instruction when the risk coefficient exceeds the set value.

[0012] Preferably, the establishment process of the quality risk prediction model comprises: extracting the pore distribution characteristics and incomplete fusion area characteristics in the historical welding data; analyzing the corresponding relationship between the defect features and the molten pool stability index; establishing a quality risk prediction model based on a multi-dimensional feature space.

[0013] Preferably, the present application further comprises an automatic welding machine based on artificial intelligence, which comprises all modules and functions of the above-mentioned automatic welding system based on artificial intelligence.

[0014] Compared with the prior art, the present application has the following advantages: The welding feature acquisition module realizes comprehensive acquisition of multi-dimensional information of the workpiece, which is different from the traditional single detection or manual measurement method. The feature fusion technology is used to integrate three-dimensional profile data, material composition information and weld geometry parameters to generate a workpiece welding feature map, which can more comprehensively and accurately reflect the welding-related characteristics of the workpiece, provide more reliable basic information for subsequent process parameter setting, effectively avoid process parameter setting deviation caused by incomplete or inaccurate feature information, and improve the rationality of welding process design from the source. In the process parameter generation link, the system does not rely on fixed templates or repeated trial welding, but retrieves matching process parameter templates from the welding knowledge base based on the generated workpiece welding feature map, and outputs the reference welding process parameter combination combined with the environmental temperature and humidity compensation coefficients. This process makes full use of the large amount of welding experience data accumulated in the knowledge base, and also considers the influence of environmental factors on the welding process, so that the output reference process parameters are more suitable for the actual welding scene requirements, reducing the welding quality fluctuations caused by environmental changes, without the need for parameter adjustment through multiple trial weldings, greatly reducing the consumption of manpower and materials, and shortening the process preparation time. The real-time regulation module is set to enable the system to continuously monitor the molten pool shape feature, heat radiation distribution and arc sound feature during the welding process. Through analysis of these real-time data, subtle changes or potential abnormalities in the welding process can be found in time, and the reference welding process parameter combination can be dynamically modified to generate optimized welding process parameter instructions. Compared with the lack of real-time monitoring and adjustment in the traditional welding process, this module can effectively avoid welding defects caused by various uncertain factors in the welding process, ensure that the welding process is always in a stable and ideal state, and improve the consistency and reliability of the welding quality.

[0015] The motion planning module accurately calculates the motion trajectory, pose adjustment parameters and speed control curve of the welding execution mechanism according to the optimized welding process parameter instructions, forming a robot control instruction set. This motion planning method based on dynamically optimized process parameters breaks the limitations of traditional fixed path planning, so that the motion of the welding execution mechanism can match the real-time process parameters. Whether it is the welding of complex structure workpieces or the adjustment of welding speed and pose to adapt to changes in process parameters, more accurate welding operations can be achieved, further improving welding precision, avoiding welding efficiency reduction or quality problems caused by mismatch between motion trajectory and process parameters, and promoting the development of welding operations towards higher efficiency, higher quality and lower cost. This reduces the dependence on operator experience and reduces human factors interference in the welding process, which is more in line with the development needs of modern manufacturing industry intelligence and automation. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1A timing diagram for the artificial intelligence-based automatic welding system described herein; Figure 2 A flowchart for generating a workpiece welding feature map; Figure 3 A flowchart for generating an optimized welding process parameter instruction. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0018] Please refer to Figure 1 The present application provides an artificial intelligence-based automatic welding system, which comprises a welding feature acquisition module, a process parameter generation module, a real-time control module and a motion planning module.

[0019] The welding feature acquisition module is responsible for acquiring the three-dimensional contour data, material composition information and weld geometry parameters of the workpiece to be welded, and generating a workpiece welding feature map using feature fusion technology; the map comprehensively reflects the geometric characteristics, material properties and weld shape of the workpiece. The process parameter generation module retrieves a matched process parameter template from a pre-constructed welding knowledge base based on the workpiece welding feature map, and outputs a reference welding process parameter combination in combination with real-time environmental temperature and humidity compensation coefficients; the combination provides initial parameter settings for the welding process. The real-time control module dynamically monitors the molten pool shape feature, heat radiation distribution and arc soundprint feature during the welding process, corrects the reference welding process parameter combination by analyzing these real-time data, and generates an optimized welding process parameter instruction; ensuring the stability of the welding quality. The motion planning module calculates the motion trajectory, pose adjustment parameters and speed control curve of the welding execution mechanism according to the optimized welding process parameter instruction, and finally forms a robot control instruction set; driving the welding robot to complete accurate operation. The modules are interconnected through a data bus, realizing real-time information sharing and collaborative control, and the overall system adopts a layered processing architecture, with the bottom layer collecting raw data, the middle layer performing feature extraction and decision-making through artificial intelligence algorithms, and the upper layer executing action output.

[0020] Embodiment 1: Please refer to Figure 2The generation process of the workpiece welding feature map starts with the accurate capture of the three-dimensional spatial information of the workpiece. A high-precision laser three-dimensional scanner is used to scan the workpiece to be welded which is placed on the welding fixture. The laser beam emitted by the scanner forms a light strip on the surface of the workpiece. The deformed light strip image is captured by a binocular vision sensor, and dense three-dimensional point cloud data is generated through point cloud reconstruction algorithm, which records the spatial coordinate information of the workpiece surface. In the preprocessing stage of point cloud data, statistical filtering algorithm is used to remove outlier noise points, and then voxel grid downsampling is used to reduce the data volume while maintaining the model accuracy, creating good conditions for subsequent surface segmentation processing. The surface segmentation processing uses a region growing algorithm based on curvature features. The algorithm selects seed points from the point cloud and expands the region based on the normal vector and curvature continuity between adjacent points. Each generated surface segment is assigned an independent identifier. The curvature feature is derived by calculating the Gaussian curvature and average curvature of all points in the segment. The normal vector is determined by fitting the local tangent plane through principal component analysis. The construction of the geometric feature matrix takes the surface segment as the basic unit. The rows of the matrix correspond to different segments, and the columns contain the average curvature, curvature rate of change, azimuth and inclination of the normal vector, etc. of the segment, forming a multi-dimensional numerical matrix to quantitatively describe the geometric feature of the workpiece surface.

[0021] The acquisition of material composition information relies on a mobile laser-induced breakdown spectroscopy analyzer. The analyzer is driven by a mechanical arm to approach the workpiece welding area. Laser pulses are emitted to produce plasma on the material surface. The wavelength and intensity of the plasma emission spectrum are analyzed to determine the content ratio of elements such as carbon, silicon, manganese, chromium, and nickel in the material. The calculation of the material welding performance index considers the influence of each element on the welding hot crack sensitivity, hardening tendency, and toughness. A multi-factor weighted model is used, with carbon equivalent calculation as the basis and the content of harmful elements such as phosphorus and sulfur for correction. Finally, a standardized index ranging from zero to one is obtained. The higher the index value, the better the material welding performance. The measurement of weld geometry parameters is completed by an industrial digital camera equipped with a high-resolution lens. The camera collects two-dimensional images of the weld area from multiple angles, and the three-dimensional morphology of the weld is reconstructed through stereo vision algorithm. The measurement parameters include the width, depth, angle of the weld bevel, and the variation of these parameters along the weld length. The width variation rate is calculated using the sliding window method. The width values are measured at fixed intervals along the weld length, and then the standard deviation of the width variation percentage between adjacent measurement points is calculated. The angle deviation value is obtained by comparing the absolute difference between the actual measured weld bevel angle and the design standard angle. The construction of the weld feature vector includes the statistical features of the width sequence such as mean, variance, skewness, and the distribution characteristics of the angle deviation, forming a numerical sequence that describes the continuous change of the weld geometry.

[0022] The feature fusion process adopts a multi-layer perceptron neural network structure, which includes an input layer, multiple hidden layers, and an output layer. The geometric feature matrix is flattened and sent to the input layer of the network, the material welding performance index is input as a scalar, and the weld feature vector is standardized and input simultaneously. The network performs nonlinear transformation and weighted combination of features from different sources in the hidden layer, introduces nonlinearity through the activation function, and enables the system to learn the complex correlation between different features. The trained network can map heterogeneous input features to a unified low-dimensional feature space and generate a workpiece welding feature map. The workpiece welding feature map is transmitted and stored in the system in the form of a feature tensor, with each dimension of the tensor corresponding to different feature types and spatial position information. The map not only contains static features, but also records the distribution of features at different weld positions, providing comprehensive and detailed input for subsequent process parameter decision-making. The entire generation process uses a pipeline architecture, with data buffer zones between each processing stage to ensure data processing continuity and real-time performance even when processing time fluctuates. The feature fusion process also introduces an attention mechanism, allowing the network to adaptively focus on feature dimensions that have a greater impact on welding quality, improving the feature map's representation ability. The system regularly uses newly collected workpiece data to incrementally train the feature fusion network, continuously optimizing feature extraction capabilities over time and gradually adapting to more diverse workpiece types and welding conditions.

[0023] Example 2: The determination process of the reference welding process parameter combination is a decision-making process based on knowledge retrieval and dynamic correction. The welding knowledge base, as the core data resource of the system, stores a large amount of welding case data verified by practice. Each case contains a complete process parameter template. These templates are indexed and classified according to material type, thickness range, and weld form, forming a structured parameter database. The knowledge base uses a distributed architecture for storage, supporting fast parallel retrieval. When receiving the workpiece welding feature map from the feature acquisition module, the system starts the multi-condition matching retrieval process.

[0024] The retrieval process first focuses on the calculation of material matching degree. The system extracts the material welding performance index in the workpiece welding feature spectrum and compares it with the material characteristic data recorded in each template in the knowledge base. The similarity algorithm adopts a distance measurement method based on feature weighting, considering the closeness of multiple indicators such as material chemical composition and mechanical properties. The system will filter out all candidate templates with a material matching degree exceeding a preset threshold, which is usually dynamically adjusted according to the welding quality requirements. A higher matching threshold is set for high-quality welding applications. The candidate templates are arranged in descending order of matching degree score to form a preliminary parameter set. Next, the system enters the parameter adjustment stage, and the width change rate in the weld feature vector becomes the key adjustment basis. The original standard current fluctuation range in the process parameter template needs to be adjusted according to the geometric characteristics of the actual weld. When the weld width change rate is detected to be large, it indicates that the weld gap fluctuates significantly along the welding direction, and the system will accordingly expand the allowed fluctuation range of the current. This adjustment is achieved by looking up the preset parameter adjustment rule table. The rule table records the current adjustment coefficients corresponding to different width change rate intervals. These coefficients are empirical values based on a large amount of process test data. The adjustment process keeps the current reference value unchanged and mainly modifies its upper and lower fluctuation ranges.

[0025] The environmental compensation link collects environmental data in real time through temperature and humidity sensors installed in the welding workstation. The temperature reading is accurate to 0.1 degrees Celsius, and the humidity measurement accuracy reaches ±2%RH. These environmental parameters are converted into temperature compensation coefficients and humidity compensation coefficients. The modification of the voltage reference value is the core link of environmental compensation. The process parameter template in the knowledge base contains the voltage reference value under standard environmental conditions, which now needs to be modified according to the actual environmental conditions. The temperature compensation coefficient mainly affects the arc stability, and as the environmental temperature rises, the voltage required to maintain a stable arc decreases accordingly. Humidity compensation focuses on the effect of protective gas. When humidity increases, the voltage needs to be appropriately increased to compensate for the change in protective gas density. The two compensation coefficients work together through a multiplication model to affect the voltage reference value. The final stage of forming the reference welding process parameter combination requires consistency checking of all parameters. The system checks whether the logical relationship between the parameters is reasonable, such as the matching of welding speed and heat input, and the correspondence between wire feed speed and current value. After passing the verification, the system generates a complete reference welding process parameter combination document. This document records all necessary parameters such as current range, voltage value, welding speed, wire feed speed, and protective gas flow in a standardized format. The parameter combination also includes a confidence score, which is based on the reliability of matching degree calculation and environmental compensation. This score provides a reference basis for the parameter adjustment range of the subsequent real-time control module.

[0026] The updating and maintaining mechanism of the knowledge base is an important guarantee for the continuous optimization of the system. After each welding task is completed, the system will store the actual used process parameters and the welding quality evaluation results as new cases in the knowledge base. Before the new cases are stored, they need to go through quality audit, and only the cases with qualified welding quality will be adopted. The knowledge base also has a self-learning function, which can analyze the parameter rules of successful cases and automatically optimize the coefficient settings in the parameter adjustment rule table, so that the decision-making ability of the system is continuously improved with the increase of use time. For welding tasks of special materials or new joint forms, the system supports manual import of verified process parameter templates to enrich the coverage of the knowledge base. The determination process of the entire reference welding process parameter combination adopts fault-tolerant design. When the number of retrieved candidate templates is insufficient, the system will gradually relax the matching threshold to ensure that a feasible parameter scheme can be given in any case. Multiple safety checks are also provided during the parameter determination process to prevent the occurrence of obviously unreasonable parameter combinations. All parameter adjustments are made within the preset safe range to avoid damage to the welding equipment caused by extreme parameters. The system will record the detailed log of each parameter decision, including the retrieval conditions, matching results, adjustment process and other complete information. These logs are used for subsequent process analysis and system optimization. The output of the reference welding process parameter combination adopts a hierarchical structure, with core parameters as the main output content, and derivative parameters and auxiliary parameters as supplementary information output simultaneously. All parameters are marked with data sources and adjustment basis to facilitate the subsequent module to understand the logical basis of parameter decision. The parameter combination also contains version control information. When the knowledge base is updated, the system will mark the knowledge base version based on which the parameter combination is based on, to maintain the traceability of the process parameters. The entire determination process is completed within seconds, meeting the high requirements of modern automated welding on process preparation efficiency.

[0027] Example 3: see Figure 3The generation of optimized welding process parameter instructions is based on multi-sensor information fusion and dynamic analysis. A high-speed camera system continuously acquires image sequences of the welding pool area at a fixed frame rate. The image acquisition process is coordinated with a filter of a specific wavelength to suppress the interference of arc light. After preprocessing, each frame of image enters the profile extraction link, and an edge detection algorithm is used to identify the boundary line between the pool and the base material. The analysis of the pool profile change characteristics focuses on the time sequence evolution of the geometric shape, and the instantaneous fluctuation of the pool width is calculated by comparing the position changes of the profile points between consecutive frames. The evaluation of the length consistency is realized by measuring the distance changes from the tail of the pool to the center of the arc, and these dynamic parameters collectively reflect the stability state of the pool. The quantification of the pool stability index requires the comprehensive consideration of multiple characteristic parameters. The system uses a sliding time window statistical method to calculate the ratio of the standard deviation to the average value of the pool width within a set time segment when calculating the width fluctuation rate. The evaluation of the length consistency is realized by calculating the coefficient of variation of the pool length, which reflects the dispersion degree of the length value. The feature extraction of the pool surface ripples relies on image texture analysis technology. After frequency domain transformation of the pool area, the distribution rule of the ripples is analyzed, and the amplitude feature is obtained by measuring the height difference between the peak and valley values of the ripples. These characteristic parameters are combined through weighted combination to form a comprehensive pool stability index, and the weight coefficients of each feature are trained according to historical welding quality data.

[0028] The deployment position of the infrared temperature measurement system is carefully designed. Multiple infrared sensors are installed at specific angles around the welding area to form a three-dimensional monitoring network of the thermal field. The temperature data collected by the sensors are processed through spatial interpolation to generate a thermal radiation distribution map of the welding area. The extraction of temperature gradient data is based on the isotherm distribution in the thermal image. The heat concentration area is quantified by calculating the spacing changes between adjacent isotherms. The establishment of the heat affected zone model uses the basic principles of heat transfer and the thermal conductivity characteristics of the material to predict the thermal cycle curve. This model can simulate the heat conduction process in the workpiece and predict the width and organizational change trend of the heat affected zone under different welding parameters. The collection of arc voiceprint features uses a directional microphone array. The arrangement of the array takes into account the propagation path of the sound wave and the shielding of environmental noise. The collected audio signals are first filtered through a bandpass filter to remove low-frequency mechanical noise and high-frequency interference, and then processed by frame. The analysis of the energy distribution spectrum uses the short-time Fourier transform method to convert the time-domain signal to the frequency-domain representation. The characteristic frequency components are identified through a spectrum peak detection algorithm. The identification of abnormal arc states is based on the pattern matching of the voiceprint spectrum. The real-time voiceprint is compared with the typical abnormal voiceprints in the knowledge base in terms of similarity. The judgment threshold of abnormal states is adaptively adjusted according to the welding material and the type of protective gas.

[0029] The calculation process of the pool stability index introduces a comprehensive evaluation model that considers the interaction between multiple characteristic parameters: , wherein: represents the final calculation result of the molten pool stability index, the higher the value, the better the stability; represents the molten pool width fluctuation rate, which is obtained by calculating the standard deviation of the width value in the time window; represents the molten pool length consistency coefficient, reflecting the variation degree of the length value; represents the surface ripple amplitude characteristic parameter, the average ripple amplitude value obtained by image analysis; , , respectively correspond to the weight coefficients of the three characteristics, which are determined by a large number of process test data regression analysis. The model can comprehensively reflect the geometric stability characteristics of the molten pool, and provide quantitative basis for process parameter adjustment.

[0030] The adjustment strategy of the reference welding process parameter combination adopts a multi-level decision mechanism. The system first determines the emergency level of adjustment according to the deviation of the molten pool stability index; when the index shows that the molten pool is in a critical unstable state, the system will preferentially adjust the current parameter, because the current has the most direct influence on the stability of the molten pool. The output of the heat affected zone model is used to guide the fine adjustment of the heat input, and when the model predicts that the heat affected zone width exceeds the allowed range, the system will appropriately reduce the welding speed or adjust the voltage value; the detection result of the arc abnormal state is mainly used for protective adjustment, and when the arc drift or short circuit tendency is identified, the system will immediately fine-tune the combination of voltage and wire feeding speed. The coordinated control in the parameter adjustment process is crucial, and the system maintains a parameter influence relationship matrix, which describes the coupling relationship between various process parameters; when adjusting a parameter, the system will simultaneously adjust other parameters that are strongly associated with it, avoiding the introduction of new unstable factors due to single parameter adjustment. All parameter adjustments are carried out within the preset safety boundary, and the adjustment amplitude is dynamically weighted according to the reliability of real-time monitoring data; high confidence sensor data corresponds to a larger adjustment amplitude, while noisy data corresponds to a relatively conservative adjustment.

[0031] The generation frequency of the optimized welding process parameter instruction is matched with the welding speed, a higher instruction update rate is adopted in high-speed welding to ensure that the system can respond to changes in the process state in time; each optimization instruction contains a parameter adjustment reason explanation, recording which sensor data triggered this adjustment. The instruction output is also subjected to rationality verification, checking whether the adjusted parameters are within the allowed working range of the equipment and whether the parameter combination conforms to the basic laws of welding metallurgy. The entire optimization process forms a complete closed-loop control, with real-time monitoring, analysis, decision-making, and execution closely linked, and the optimal state of the welding process is maintained through continuous parameter fine-tuning. The system's learning mechanism records the effect of each parameter adjustment, and evaluates the effectiveness of the adjustment strategy by comparing the changes in the molten pool stability index before and after adjustment; these feedback data are used for regular optimization of the rule library, enabling the system's decision-making ability to evolve continuously with the accumulation of experience. For welding processes of special materials, the system supports the import of expert experience rules, which work together with the automatic optimization algorithm to ensure that reasonable optimization instructions are generated under various working conditions.

[0032] The forming process of the robot control instruction set can be clearly demonstrated through a specific welding case. Suppose the system is handling a longitudinal seam welding task for a large pressure vessel cylinder, the workpiece material is low-alloy high-strength steel, the wall thickness is twenty-five millimeters, and the weld length is about three meters; the welding execution mechanism uses a six-axis articulated robot with a water-cooled welding gun at the end. The optimized welding process parameter instruction requires the welding speed to be dynamically adjusted between three hundred and fifty millimeters per minute and four hundred and twenty millimeters per minute, while maintaining a constant heat input; the system first analyzes these speed requirements and decomposes them into axial movement components along the weld direction and radial compensation amounts perpendicular to the workpiece surface. The calculation of the axial movement component is based on the welding speed reference value, and the system converts the three hundred and fifty millimeters per minute speed into the angular velocity of each joint of the robot; the determination of the radial compensation amount needs to consider the workpiece assembly error and the weld height change caused by thermal deformation, and the compensation amount is dynamically adjusted by real-time measurement of the distance between the welding gun and the workpiece surface through the laser displacement sensor. The planning of the approach path uses a three-dimensional space interpolation algorithm, and the robot controller generates a smooth approach trajectory based on the three-dimensional profile data in the workpiece welding feature map; this trajectory ensures that the welding gun approaches the weld start point from the starting point with the optimal pose, avoiding interference with the fixture.

[0033] The dynamic adjustment process of the posture control parameters is realized through real-time monitoring system. The visual sensor installed on the side of the welding torch collects the weld images at a rate of fifty frames per second, and the image processing algorithm calculates the change of the weld width in real time. When the weld width increases by 10%, the system will adjust the swing amplitude and dwell time of the welding torch accordingly. The angle deviation between the welding torch axis and the weld center line is measured by the binocular vision system. Two cameras shoot the weld area from different angles, and the relative position relationship in three-dimensional space is calculated by triangulation method. According to the measurement results of the width change and the angle deviation, the system determines the compensation amount of the posture adjustment. The update of the posture parameters of the welding actuator adopts a gradual adjustment strategy. The adjustment of the pitch angle follows the cosine function curve, ensuring the smooth transition of the posture change. The adjustment of the deflection angle is synchronized with the welding speed. A smaller adjustment amplitude is adopted in the high-speed welding section, and a larger adjustment range is allowed in the low-speed section. During the whole adjustment process, the system continuously monitors the joint torque and motor current of the robot. When an abnormal load is found, the adjustment amplitude is automatically reduced to prevent overload.

[0034] Table 1: Dynamic adjustment record of welding posture parameters

[0035] The generation of the motion trajectory adopts the spline curve interpolation method. The system sets a series of key points on the weld path, connects these points through cubic spline curves, and ensures the smoothness and continuity of the robot motion. The posture parameters of each key point are optimized and calculated to ensure that the welding torch always maintains the optimal working angle. The planning of the speed control curve considers the dynamics of the robot. The uniform speed motion is adopted in the straight line section, and the speed is appropriately reduced in the curve section to avoid vibration caused by centrifugal force.

[0036] The final integration process of the robot control instruction set involves coordinate transformation and kinematics calculation. The system converts the global coordinate system of the welding path to the robot base coordinate system, and converts the pose of the end effector to the angle values of the six joints through inverse kinematics calculation. The motion trajectory of each joint is processed through velocity planning and acceleration limitation, generating a smooth joint space trajectory. The instruction set adopts a hierarchical structure, with high-level instructions describing the welding task and path planning, middle-level instructions defining the motion trajectory and posture sequence, and low-level instructions directly controlling the pulse output of the servo driver. The anti-collision detection algorithm runs continuously during the instruction generation process. The system calculates the distance field between the robot and the surrounding environment, and adjusts the trajectory in real time to avoid interference. When a potential collision is predicted, the system automatically replans the path to ensure the safe operation of the welding process. The instruction set also contains fault-tolerant processing mechanisms. When a sensor data is abnormal, the system can switch to a backup control mode and use historical data or model prediction values to continue the welding task. Real-time correction during welding is achieved through closed-loop control. The laser tracking system continuously monitors the deviation between the actual position of the welding torch and the planned trajectory. When the deviation exceeds the set threshold, the system immediately generates trajectory correction instructions. These correction amounts are smoothly integrated into the subsequent motion trajectory through the interpolation algorithm of the robot controller, avoiding sudden position jumps. The response time of the entire control system is controlled within ten milliseconds, ensuring timely compensation for various external disturbances.

[0037] The optimization of the instruction set also considers energy consumption factors. The system selects the motion trajectory with the lowest energy consumption to reduce the energy consumption of the robot body while ensuring the quality of the welding. The trajectory planning algorithm avoids unnecessary acceleration and deceleration processes, maintaining the smoothness of the motion. After each welding task is completed, the system analyzes the differences between the actual motion data and the planned trajectory, and uses these data to optimize the trajectory planning parameters for subsequent tasks, continuously improving control accuracy. The verification of the robot control instruction set is carried out in a virtual simulation environment. The system uses a robot dynamics model to simulate the actual motion process and detects whether there are singular points or over-limit positions. After simulation, the instruction set is downloaded to the actual robot controller for execution. The entire instruction generation process maintains strict timing synchronization, with precise coordination of all motion axes, ensuring stable and reliable welding quality. The system records the execution data of each welding task, including actual trajectory, posture parameters, and sensor readings, which are used for subsequent process analysis and system optimization.

[0038] Example 5: The implementation process of the quality prediction module is applied to the welding of a ring seam of a storage tank. The material of the storage tank is austenitic stainless steel, with a diameter of four meters and a plate thickness of twelve millimeters. The welding process uses a double-wire submerged arc welding process, with a welding speed controlled at about four hundred millimeters per minute. The historical welding database stores the welding records of similar structures completed in the past three years, totaling about two hundred complete cases. Each case includes process parameter records, non-destructive testing results, and mechanical property test data. After the module is started, it first retrieves these historical data from the database, performs preprocessing and feature extraction. The extraction of pore distribution features is based on digital analysis of X-ray detection negatives. The system uses image processing algorithms to identify pore images on the negatives, measuring the diameter and position coordinates of each pore. Feature parameters include the number of pores per unit area, pore diameter distribution, and pore aggregation index. The acquisition of un-melted area features relies on ultrasonic phased array detection data. The system analyzes the signal features in the B-scan images to identify the size and orientation of un-melted defects. Feature parameters include the un-melted area ratio, defect length, depth position, and the angle between the defect orientation and the weld centerline. These defect features are associated with corresponding process parameters to form a training sample set.

[0039] The analysis of the correspondence between defect features and molten pool stability indicators uses a time series matching method. The system spatially corresponds the molten pool stability indicator time curve recorded during welding with the subsequent detected defect positions. By analyzing the change pattern of the molten pool stability indicator near the defect occurrence time point, it finds the correlation pattern. Analysis shows that when the molten pool stability indicator experiences sharp fluctuations within a certain time period, the probability of pores appearing in that area significantly increases. Un-melted defects are often associated with abnormal temperature gradients of heat radiation distribution. The quality risk prediction model based on a multi-dimensional feature space uses a deep learning architecture. The input layer of the model includes fifteen feature parameters, including real-time molten pool stability indicators, maximum heat-affected zone temperature gradient, and arc soundprint anomaly index. The three hidden layers contain thirty-two, sixteen, and eight neurons respectively, and the output layer produces a quality risk coefficient between zero and one. The model training uses one hundred and fifty cases from the historical data, with the remaining fifty cases used to verify the model prediction accuracy. The training process uses the adaptive moment estimation algorithm, and the loss function considers both classification accuracy and generalization ability. During real-time risk monitoring, the system collects current welding state data at a frequency of ten times per second, including molten pool stability indicators and heat-affected zone temperature distribution. Each data point is standardized and input into the trained quality risk prediction model, which outputs the risk coefficient of the current welding point in real time. The calculation of the risk coefficient considers the weighted combination of multiple feature parameters, with the molten pool stability indicator having the highest weight, the heat-affected zone temperature gradient second, and the arc soundprint feature as an auxiliary judgment basis. The system sets a dynamic risk threshold that adjusts according to the importance level of the weld, with more stringent control standards for key load-bearing welds.

[0040] When the risk coefficient exceeds the set value, the system triggers the process parameter emergency intervention instruction; the intervention instruction generation adopts a hierarchical response mechanism, and different intensity intervention measures are taken according to the amplitude and duration of the risk coefficient exceeding the limit. The first level intervention is for slight over-limit situation, the system will automatically adjust the welding parameters for compensation, such as fine-tuning the welding current or voltage value; the second level intervention is for continuous over-limit situation, the system will reduce the welding speed and adjust the heat input parameter at the same time; the third level intervention is for serious over-limit situation, the system will immediately stop the welding process, and prompt the operator to intervene and check. The execution of the emergency intervention instruction adopts a gradual strategy to avoid secondary interference on the welding process caused by parameter mutation; after the instruction is issued, the system will continue to monitor the change trend of the risk coefficient, and if the index does not improve, higher level intervention measures will be started. All intervention actions will be recorded in the process log, including intervention time, intervention type, parameter adjustment amount and effect evaluation after intervention; these data are used for subsequent analysis of the effectiveness of the intervention strategy.

[0041] The self-learning function of the quality prediction module is realized through an online updating mechanism. After each welding task is completed, the system compares the actual detection results with the predicted results. When it finds that the prediction deviation is large, it will automatically mark the case. After accumulating a certain number of new cases, the system will start the model parameter optimization process and fine-tune the prediction model using new data. The model updating process uses an incremental learning algorithm to integrate new experience while maintaining existing knowledge, allowing the prediction ability to evolve over time. The reliability of the module is ensured by a multiple check mechanism. The prediction results of important welds are cross-validated by two independent models. Only when the prediction conclusions of the two models are consistent will intervention be taken. For the uncertainty of the prediction results, the system will adopt a conservative strategy, preferring false positives to missing serious defects. The visualization interface of the prediction results displays the risk coefficient change curve in real time and uses color to mark the high-risk section, making it easy for operators to intuitively understand the welding quality status. The management of historical data uses a version control method. When the welding process or material specification changes, the system will create a new data branch to avoid interference between data of different process conditions. Data cleaning is performed every six months to remove invalid data and outliers and maintain the quality of the database. The module also supports expert experience import function, allowing senior engineers to adjust the risk judgment rules based on actual experience. These manual adjustments are stored in the knowledge base as special cases, enriching the system's decision-making basis. The collaboration with other modules is realized through a data bus. The quality prediction module receives real-time molten pool monitoring data from the real-time control module and feeds back the prediction results to the process parameter generation module. When a quality risk is predicted, the module sends a warning signal to the motion planning module, prompting possible adjustments to the welding path or speed. The entire quality prediction process forms a complete monitoring-prediction-intervention closed loop, providing additional protection for the stability of the welding quality. All data generated during the module operation are encrypted and stored to ensure the security and traceability of the process data.

[0042] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0043] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. An automated welding system based on artificial intelligence, characterized in that, include: The welding feature acquisition module is used to acquire the three-dimensional contour data, material composition information and weld geometric parameters of the workpiece to be welded, and generate the workpiece welding feature map through feature fusion technology; The process parameter generation module, based on the workpiece welding feature map, retrieves matching process parameter templates from the welding knowledge base, and outputs a combination of benchmark welding process parameters by combining the ambient temperature and humidity compensation coefficients. The real-time control module dynamically corrects the combination of reference welding process parameters based on the molten pool morphology, heat radiation distribution, and arc acoustic characteristics during the welding process, and generates optimized welding process parameter instructions. The motion planning module calculates the motion trajectory, attitude adjustment parameters, and speed control curve of the welding actuator based on the optimized welding process parameter instructions, forming a robot control instruction set.

2. The artificial intelligence-based automatic welding system according to claim 1, characterized in that, The process of generating the workpiece welding feature map includes: The three-dimensional contour data is processed by surface segmentation, and the curvature features and normal vectors of each surface segment are extracted to establish a geometric feature matrix; Analyze the element content ratio in the material composition information to calculate the material's welding performance index; Measure the width variation rate and angle deviation values ​​of the weld geometry parameters to construct the weld feature vector; The geometric feature matrix, material welding performance index, and weld feature vector are fused to generate a workpiece welding feature map.

3. The artificial intelligence-based automatic welding system according to claim 1, characterized in that, The process of determining the combination of the reference welding process parameters includes: Based on the material welding performance index in the workpiece welding feature map, process parameter templates in the welding knowledge base that meet the preset threshold of material matching degree are selected. Based on the width change rate of the weld feature vector, adjust the current fluctuation range in the process parameter template; By combining the ambient temperature and humidity compensation coefficients, the voltage reference value in the process parameter template is corrected to form a reference welding process parameter combination.

4. The artificial intelligence-based automatic welding system according to claim 1, characterized in that, The process of generating the optimized welding process parameter instructions includes: The contour changes of the molten pool morphology are captured by a high-speed camera system, and the stability index of the molten pool is calculated. Temperature gradient data of thermal radiation distribution are obtained using an infrared thermometry system to establish a model of the heat-affected zone. Collect the energy distribution spectrum of electric arc acoustic signature features to identify abnormal electric arc states; Based on the molten pool stability index, heat-affected zone model, and abnormal arc state, adjust the key parameters in the baseline welding process parameter combination.

5. The artificial intelligence-based automatic welding system according to claim 4, characterized in that, The calculation process for the molten pool stability index also includes: Analyze the width fluctuation rate and length consistency in the characteristics of molten pool profile variation; Extract the distribution pattern and amplitude characteristics of surface ripples in the molten pool; The stability index of the molten pool is calculated by combining the width fluctuation rate, length consistency and surface ripple characteristics.

6. The artificial intelligence-based automatic welding system according to claim 1, characterized in that, The process of forming the robot control instruction set includes: The welding speed requirement in the optimized welding process parameter command is analyzed and decomposed into axial movement component and radial compensation amount; Based on the three-dimensional contour data in the workpiece welding feature map, plan the approach path of the welding actuator; The attitude control parameters are dynamically adjusted based on the real-time changes in weld geometry parameters. The motion component, approach path, and attitude parameters are integrated to form a robot control instruction set.

7. The artificial intelligence-based automatic welding system according to claim 6, characterized in that, The dynamic adjustment process of the attitude control parameters also includes: Monitor the real-time width changes of weld geometry parameters; Calculate the angular deviation between the welding torch axis and the weld centerline; The compensation amount for attitude adjustment is determined based on the width change and the included angle deviation; Update the pitch and deflection angle parameters of the welding actuator.

8. The artificial intelligence-based automatic welding system according to claim 1, characterized in that, Also includes: The quality prediction module establishes a quality risk prediction model based on the correlation between defect characteristics and process parameters in historical welding data. The system analyzes the molten pool stability index and heat-affected zone model in real time. When the risk coefficient exceeds the set value, it triggers an emergency intervention command for process parameters.

9. The artificial intelligence-based automatic welding system according to claim 8, characterized in that, The process of establishing the quality risk prediction model includes: Extract porosity distribution characteristics and non-fusion region characteristics from historical welding data; Analyze the correspondence between defect characteristics and molten pool stability indices; Establish a quality risk prediction model based on a multidimensional feature space.

10. An automated welding machine based on artificial intelligence, characterized in that, It includes all modules and functions of the AI-based automated welding system as described in any one of claims 1 to 9.

Images