A twin control system and control method for a block assembly welding process in a ship

Abstract

The application discloses a twin control system and method for a ship assembly welding process, which comprises a welding physical entity, a real-time data acquisition and analysis module, a ship assembly welding digital twin model construction module, a welding state monitoring and quality prediction module, and a welding process self-adaptive regulation and control module. By constructing the ship assembly welding process control system and method based on the digital twin, the three-dimensional visual real-time monitoring of the welding process is realized, the real-time detection and prediction of the welding quality are realized based on the multi-source sensor data and the digital twin model, and the welding parameters can be dynamically optimized and adjusted according to the prediction results. The problems of low visualization, lagging quality detection and inaccurate process regulation in the existing welding control are solved, the stability of the welding quality and the production efficiency are improved, and the rework cost is reduced.

Description

Technical Field

[0001] This invention relates to a twin control system and method for the assembly welding process in ships, belonging to the field of ship welding control technology. Background Technology

[0002] Assembly welding in shipbuilding is a core process, and its quality directly affects the strength and safety of the hull structure. Traditional shipbuilding welding process control relies primarily on manual inspections and post-weld quality checks, which presents several problems: insufficient real-time monitoring of welding process parameters such as current, voltage, and weld bead formation, leading to undetected parameter drift; difficulty in online identification of welding defects such as incomplete penetration and porosity, typically only discovered after welding, resulting in high rework costs and long cycles; lack of effective mapping between the physical welding scenario and the digital model, hindering visualized monitoring and data traceability of the welding process; and reliance on operator experience for welding parameter adjustments, lacking scientific and systematic data support for precise control. Digital twin technology, by constructing a virtual mapping of physical entities, enables real-time data interaction and full lifecycle management, and has been widely applied in intelligent manufacturing. However, in shipbuilding assembly welding, due to the complex welding environment, heterogeneous multi-source data, and difficulty in dynamically updating the model, the application of digital twin technology remains in the exploratory stage, and a systematic control solution has not yet been formed. Therefore, there is an urgent need to design a ship assembly welding process control system and method based on digital twins to solve the problems of low visualization, lagging quality inspection and inaccurate process control in the traditional control mode. Summary of the Invention

[0003] Purpose of the invention: To address the shortcomings of existing technologies, this invention provides a twin control system and method for the assembly and welding process in ships, which solves the problems of low visualization, lagging quality inspection, and inaccurate process control in the existing assembly and welding process in ships, and achieves comprehensive and precise control of the welding process.

[0004] Technical solution: A twin control system for the ship assembly welding process, including a welding physical entity, a real-time data acquisition and analysis module, a ship assembly welding digital twin model construction module, a welding status monitoring and quality prediction module, and a welding process adaptive control module;

[0005] The physical entities to be welded include the assembled workpieces in the ship, the welding equipment, and the welding environment;

[0006] The real-time data acquisition and analysis module is connected to the welding physical entity, collects welding equipment status data, welding process data, visual information data and environmental data, and processes and stores the data.

[0007] The ship assembly welding digital twin model construction module is connected to the real-time data acquisition and analysis module. Based on the real-time data, a welding digital twin model is constructed and mapped to the welding physical entity.

[0008] The welding status monitoring and quality prediction module is connected to the real-time data acquisition and analysis module and the digital twin model construction module to realize three-dimensional visualization monitoring of the welding process and real-time detection and prediction of welding quality.

[0009] The welding process adaptive control module is connected to the welding status monitoring and quality prediction module and the welding equipment. It optimizes welding parameters and dynamically adjusts the welding trajectory based on the quality prediction results.

[0010] This invention achieves full-element digital expression and closed-loop management of the welding process by constructing a real-time mapping between the physical entity of the weld and the digital twin model, integrating multi-source data acquisition, three-dimensional visualization monitoring, real-time quality prediction and adaptive control, thus solving the problems of low visualization, lagging quality detection and inaccurate process control in traditional welding processes.

[0011] In a preferred embodiment, to provide a comprehensive and accurate data foundation for the digital twin model, the real-time data acquisition and analysis module includes a gantry displacement sensor, a joint displacement sensor, a laser displacement sensor, an arc sensor, an industrial camera, a data processing and storage module, and a database.

[0012] The gantry displacement sensor is installed on the gantry, and the joint displacement sensor is installed at the joints of the welding robot arm to monitor the operating status of the gantry and robot arm in real time. The laser displacement sensor is installed on the side of the welding torch to collect weld quality data. The arc sensor is integrated into the end of the welding torch for weld tracking. The industrial cameras are distributed at the welding site to collect visual information data. The data processing and storage module cleans and analyzes the data collected by each sensor and camera and stores it in the database. By configuring multiple types of sensing devices such as gantry displacement sensors, joint displacement sensors, laser displacement sensors, arc sensors, and industrial cameras, equipment status, welding process, weld quality, and visual information are collected respectively. The data is then cleaned, analyzed, and stored by the data processing and storage module, achieving synchronous acquisition and efficient management of multi-source heterogeneous data.

[0013] In a preferred embodiment, to further improve the accuracy and real-time performance of data acquisition, the gantry displacement sensor is a laser displacement sensor, which acquires the displacement and movement speed of the gantry in the X, Y, and Z axes in real time; the joint displacement sensor is an incremental encoder, which acquires the joint rotation angle, angular velocity, and angular acceleration; the laser displacement sensor measures the weld bevel width and depth, and the height and width of the formed weld; the arc sensor synchronously acquires the welding current, voltage, and arc length; and the industrial camera is an industrial-grade high-speed camera, covering the entire welding area. By employing a laser displacement sensor to achieve high-precision measurement of the gantry's three-axis displacement and velocity, using an incremental encoder to acquire the joint motion parameters of the robotic arm, using a laser displacement sensor and an arc sensor to acquire the weld geometry and electrical parameters respectively, and using an industrial-grade high-speed camera to acquire visual information of the welding area with full coverage, a refined perception of key parameters in the welding process is achieved.

[0014] In a preferred embodiment, to improve the realism and predictive accuracy of the digital twin model, the digital twin model constructed by the ship assembly welding digital twin model construction module includes a geometric model, a physical model, a behavioral model, and a rule model. The geometric model is a parametric 3D model of the assembled workpiece and welding equipment in the ship, dynamically adjusting its shape, size, and relative position based on real-time data. The physical model adds material properties, mechanical properties, assembly constraints, and welding thermodynamic characteristics to the geometric model. The behavioral model describes the actual state of welding equipment operation and weld formation. The rule model is the welding process operation law summarized from data in the analysis database. By constructing a four-layer architecture of geometric, physical, behavioral, and rule models, the digital twin model not only possesses the ability to dynamically update its appearance but also reflects the physical essence of the welding process, equipment operation behavior, and technological laws, achieving a comprehensive upgrade from static description to dynamic simulation.

[0015] In a preferred embodiment, to achieve visualized monitoring of the welding process and accurate prediction of welding quality, the welding condition monitoring and quality prediction module includes a welding condition monitoring module and a welding quality prediction module. The welding condition monitoring module updates the digital twin model based on real-time data, realizing a three-dimensional visualization of the welding process. The welding quality prediction module uses real-time data as initial conditions, and based on the digital twin model, achieves real-time detection and prediction of welding quality through welding thermodynamic calculations and weld formation simulation analysis. It also continuously optimizes the calculation accuracy using data from the database through parameter identification methods. The welding condition monitoring module dynamically presents the real-time data-driven digital twin model in three dimensions, while the welding quality prediction module, combining thermodynamic calculations and simulation analysis, performs real-time detection and prediction of welding quality. Furthermore, it continuously optimizes the model accuracy using historical data through parameter identification methods, making monitoring more intuitive and predictions more reliable.

[0016] In a preferred embodiment, to improve the accuracy of arc ignition point selection and welding trajectory tracking, the adaptive control module for the welding process includes an arc ignition point location module and a welding trajectory planning and tracking module. The arc ignition point location module, based on intelligent optimization algorithms and the prediction results of the welding quality prediction module, inputs weld groove dimensions, initial welding equipment posture, and real-time environmental data stored in the database to calculate the optimal arc ignition point within a preset arc ignition area. The welding trajectory planning and tracking module drives the welding equipment control system according to the optimal welding parameters, combining the three-dimensional geometric features of the weld and feedback from the arc sensor and laser displacement sensor to construct a dynamic trajectory tracking closed loop. By integrating intelligent optimization algorithms and quality prediction results through the arc ignition point location module, the optimal arc ignition position is calculated based on groove dimensions, equipment posture, and environmental factors. Simultaneously, the welding trajectory planning and tracking module, combining weld geometric features and sensor feedback, constructs a dynamic tracking closed loop to ensure a stable welding process and accurate trajectory.

[0017] The control method for implementing a twin control system for the assembly and welding process in ships includes the following steps:

[0018] S1. Construct a welding digital twin model in the welding digital twin model construction module in the ship;

[0019] S2. During the welding process, the welding equipment status data is collected in real time through gantry displacement sensor and joint displacement sensor, the welding process data is collected in real time through laser displacement sensor and arc sensor, and the visual information data is collected in real time through industrial camera, while environmental data is also collected.

[0020] S3, the data processing and storage module cleans and analyzes the collected data and stores it in the database, and drives the updating of the welding digital twin model in real time;

[0021] S4, the welding quality prediction module, is based on processed real-time data and welding digital twin model. Through welding thermodynamic calculations and weld formation simulation analysis, it realizes real-time detection and prediction of welding quality.

[0022] S5, the welding status monitoring module, dynamically displays the operating status of welding equipment, weld formation process and changes in welding parameters based on real-time data, updated welding digital twin model and prediction data, to achieve three-dimensional visualization monitoring of the welding process;

[0023] S6. Based on three-dimensional visualization monitoring, the welding process adaptive control module optimizes and determines the best combination of welding parameters based on intelligent optimization algorithms and welding quality prediction results, and drives the welding equipment control system to adjust the welding parameters.

[0024] S7. Repeat S2 to S6 to achieve continuous monitoring of the entire welding process and store all data in the database to improve the accuracy of welding thermodynamic calculations and weld formation simulation analysis.

[0025] In a preferred embodiment, the specific method for optimizing the calculation accuracy through parameter identification in S4 is as follows: using the equipment status data and welding process parameters at the initial stage of welding for each weld segment stored in the database as input, and the real-time data and final quality inspection results during the welding process as output, the uncertain parameters in the welding thermodynamic calculation are identified; using the welding process data and visual information data stored in the database as input, and the final weld quality inspection data as output, the uncertain parameters in the weld formation simulation analysis are identified.

[0026] In the preferred embodiment, the welding thermodynamic calculation in S4 specifically includes:

[0027] The welding thermodynamics calculation in S4 is specifically as follows:

[0028] The formula is as follows, which combines the three-dimensional transient heat conduction equation with the Gaussian surface heat source model:

[0029] Formula for electric arc heat flux density:

[0030]

[0031] in, Heat flux density at a distance r from the center of the electric arc; Arc thermal efficiency (0.75~0.85); Welding voltage; Welding current; Effective heating radius of the electric arc; : Calculate the distance from the point to the center of the electric arc;

[0032] Transient heat conduction equation:

[0033]

[0034] in, Material density; Specific heat capacity at constant pressure, which varies with temperature; Instantaneous temperature; :time; Thermal conductivity, which varies with temperature; : Laplace operator; Arc heat flux density.

[0035] The intelligent optimization algorithm in the welding arc initiation point location module is as follows:

[0036] Objective function:

[0037]

[0038] in, Comprehensive quality evaluation indicators; : Three-dimensional coordinates of the starting point of the arc; : Deviation between measured and standard values ​​of bevel width; Standard bevel width; : Deviation between measured and standard values ​​of bevel depth; Standard bevel depth; : Weighting coefficients, summing to 1; Defect risk probability;

[0039] Particle update formula:

[0040]

[0041]

[0042] in, : The velocity of the i-th particle in the d-th dimension in the next round; Inertia weight (0.4~0.9); , Learning factor, all set to 2; , : 0~1 random number; The optimal position of an individual particle; The global optimal position of the particle swarm; : The position of the i-th particle in the d-th dimension in the next round.

[0043] Beneficial effects: This invention, by constructing a digital twin-based ship assembly welding process control system and method, realizes three-dimensional visualization and real-time monitoring of the welding process. Based on multi-source sensor data and digital twin models, it achieves real-time detection and prediction of welding quality, and can dynamically optimize and adjust welding parameters according to the prediction results. It solves the problems of low visualization, lagging quality detection, and inaccurate process control in existing welding management, improves the stability of welding quality and production efficiency, and reduces rework costs. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0048] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0049] like Figure 1 As shown, a ship assembly welding process digital twin control system includes a welding physical entity, a real-time data acquisition and analysis module, a ship assembly welding digital twin model construction module, a welding status monitoring and quality prediction module, and a welding process adaptive control module.

[0050] The physical entities to be welded include the assembled workpieces in the ship, the welding equipment, and the welding environment;

[0051] The real-time data acquisition and analysis module is connected to the welding physical entity, collecting welding equipment status data, welding process data, visual information data, and environmental data, and processing and storing the data. The welding equipment status data includes the displacement and movement speed of the gantry along the X, Y, and Z axes, as well as the joint rotation angle, joint angular velocity, joint acceleration, and initial posture angle of the welding torch of the welding robot. The welding process data includes the weld bevel width and depth collected by the laser displacement sensor, as well as the weld height and width after forming, and the welding current, welding arc voltage, welding arc length, and welding speed collected by the arc sensor. The visual information data includes the length and width of the molten pool, the arc center offset, the weld surface flatness, the edge straightness, and the weld bevel alignment deviation. The environmental data includes the temperature and humidity at the welding site, the airflow velocity and dust concentration in the welding area, and the atmospheric pressure at the welding site.

[0052] The ship assembly welding digital twin model construction module is connected to the real-time data acquisition and analysis module. Based on the real-time data, a welding digital twin model is constructed and mapped to the welding physical entity.

[0053] The geometric model parameter settings include the three-dimensional dimensions of the assembled workpiece in the ship, the basic dimensions of the weld groove (width / depth), the coordinates of the workpiece assembly position, and the geometric parameters of the hull structure connection nodes; the travel range of the gantry / XYZ axis, the number of joints / arm length / range of motion of the welding robot arm, the external dimensions / installation position of the welding torch, the installation coordinates and acquisition range of each sensor; the three-dimensional spatial coordinates of the welding station, the initial relative position of the equipment and the workpiece, and the basic coordinates of the preset welding trajectory;

[0054] The physical model parameter settings include the density, specific heat capacity, thermal conductivity, melting point, elastic modulus, and Poisson's ratio of the workpiece base material (ship steel) and welding materials; the thermal efficiency of the welding arc, the solidification coefficient of the molten pool, the thermal expansion coefficient of the heat-affected zone, and the heat loss coefficient (radiation / convection) of the welding process; the moment of inertia of the gantry, the damping coefficient of the robotic arm joint, the welding force feedback threshold of the welding torch, and the sensor acquisition accuracy and error coefficient;

[0055] The behavior model parameter settings include static low-frequency updates for basic equipment operation behavior parameters (speed threshold, rotation angle range), with an update frequency of 12h / time; and dynamic high-frequency updates for real-time equipment operation status (gantry moving speed, robotic arm joint angular velocity) and real-time weld forming behavior, with an update frequency of 10ms / time.

[0056] The parameter settings for the rule model include process rule thresholds and initial parameters for the optimization algorithm, which are statically updated at a low frequency of 7 days / time; parameter correlation coefficients and quality judgment correlation weights are dynamically updated at a low frequency of 1 hour / time, based on the statistical analysis results of the accumulated collected data.

[0057] In addition to scheduled updates, when the data analysis results collected by the real-time data acquisition and analysis module meet the following event triggering conditions, the model will immediately trigger an emergency update to prioritize the matching of the model's state with that of the physical entity and avoid model distortion due to sudden parameter changes:

[0058] Sudden changes in welding equipment status: displacement, speed, and rotation angle parameters of the gantry / robotic arm exceed the preset threshold ±10%; welding torch attitude angle adjustment is ≥5° in a single instance; sensor data acquisition shows jumps / abnormalities;

[0059] Sudden changes in welding process parameters: single change in welding current / voltage ≥20A / 5V; arc length change ≥1mm; deviation of weld groove size / forming size from preset value ≥0.5mm;

[0060] Sudden changes in welding environment parameters: single change in welding site temperature ≥ 5℃; single change in relative humidity ≥ 10%RH; single change in airflow velocity ≥ 0.5m / s; single change in dust concentration ≥ 5mg / m³;

[0061] Abnormal welding quality prediction: The welding quality prediction module outputs prediction results of abnormal weld pool formation and weld defect risk ≥80%; the deviation between the model output and the actual measured value of the physical entity exceeds the error threshold.

[0062] Welding equipment operation intervention: Operators manually adjust welding equipment parameters, correct trajectory, and start / stop the equipment. After manual intervention is triggered, the model records the intervention parameters and reinitializes the simulation state based on these parameters.

[0063] The welding status monitoring and quality prediction module is connected to the real-time data acquisition and analysis module and the digital twin model construction module to realize three-dimensional visualization monitoring of the welding process and real-time detection and prediction of welding quality.

[0064] The quality prediction method, combining the process characteristics of assembly welding in ships, sensor data acquisition frequency, and digital twin model update rate, adopts a real-time prediction mode with a 1-second basic prediction cycle. Simultaneously, it sets an advanced short-cycle prediction dimension to achieve dynamic tracking of welding quality and early risk warning. Specific time scale divisions are as follows:

[0065] Real-time detection and prediction: The prediction cycle is 1 second. Based on the cleaned data from the real-time data acquisition and analysis module (acquisition frequency 5~10ms), combined with the real-time simulation results of the digital twin model, the weld quality status and real-time risk of defects at the current welding moment (every 1 second) are output, with a time lag of ≤5ms from the physical welding process.

[0066] Advanced risk prediction: Based on the current 1-5s real-time welding data, predict the welding quality development trend and potential defect risks in the next 3-5s, and reserve parameter adjustment response time for the welding process adaptive control module to ensure the timeliness of control actions;

[0067] Weld segment quality assessment: For a single continuous weld segment (length ≥ 50 mm), the comprehensive quality prediction result of the weld segment is output every 10 seconds during the welding process, and the final quality prediction assessment of the entire weld segment is output immediately after the welding is completed, with a time difference of ≤ 1 min from physical testing;

[0068] Quality prediction across three time scales—real-time detection prediction, advanced risk prediction, and weld segment quality assessment—is implemented using the same multi-source data fusion prediction model. The model shares a unified input layer, computational kernel, and output dimension specification, achieving multi-scale prediction output only through differentiated settings of time window slicing, prediction step size configuration, and result aggregation rules. The model continuously reads cleaned data and digital twin simulation data at a frequency of 5–10 ms, performing a snapshot calculation every 1 second of accumulated data, directly outputting the instantaneous weld quality status and real-time defect risk within that time window, without time extrapolation, and with a lag of ≤5 ms from the physical process. Using the current 1–5 s real-time data collected by the model as initial conditions, the core logic of welding thermodynamics calculation and weld formation simulation is reused to extrapolate 3–5 s forward according to welding process rules, outputting the quality development trend and potential defect risk for that period. Only the model's prediction time endpoint is adjusted, without changing the calculation rules. The model continuously collects the 1-second real-time detection prediction results, performing weighted aggregation and statistical analysis every 10 seconds to output the comprehensive quality prediction result for the weld segment. After the welding of a single weld segment is completed, all 1-second results for the entire segment are comprehensively evaluated, with a time difference of ≤1 minute from the physical inspection. Only result aggregation is performed throughout the process; model calculations are not repeated. The three timescale predictions share the same data source, computational kernel, and quality judgment rules. Through multi-dimensional output from a single model, dynamic tracking, early warning, and segmented judgment of welding quality are achieved, balancing real-time detection, responsive control, and complete quality evaluation.

[0069] The prediction model employs a multi-source data fusion input mode. Input data comes from cleaned, effective data from the real-time data acquisition and analysis module and real-time simulation output data from the digital twin model construction module. It is divided into a basic input layer and a model simulation layer, with no redundant input dimensions. All input data undergoes normalization processing to meet the calculation requirements of the prediction model. The prediction model's output adopts a three-dimensional output mode of "real-time status + risk level + quantitative indicators," divided into core quality indicator output and defect risk prediction output. All output results are quantitative values ​​plus level judgments, which can be directly integrated with the 3D visualization display of the welding status monitoring and welding quality prediction module and the parameter optimization and adjustment of the welding process adaptive control module. The output results support formatted data storage, facilitating subsequent quality traceability and model optimization.

[0070] The welding process adaptive control module is connected to the welding status monitoring and quality prediction module and the welding equipment. It optimizes welding parameters and dynamically adjusts the welding trajectory based on the quality prediction results.

[0071] The optimization method utilizes a self-developed welding process database (containing over 10,000 process records) and an improved SAPSO-BP neural network prediction model to achieve intelligent parameter matching. SAPSO, or Simulated Annealing-Adaptive Particle Swarm Optimization, is used to optimize the initial weights and thresholds of the BP neural network, overcoming the shortcomings of traditional BP, such as being prone to local optima, slow convergence speed, and low prediction accuracy. It incorporates the simulated annealing Metropolis criterion, allowing for probabilistic acceptance of inferior solutions to escape local optima. Inertial weights and learning factors are adaptively adjusted to balance global search and local optimization. Welding quality error is used as the fitness function to adapt to the multi-parameter coupling characteristics of ship assembly welding. A three-layer BP structure is adopted: the input layer corresponds to welding process and environmental parameters, the hidden layer has 10 nodes with the Sigmoid activation function, and the output layer uses the Purelin activation function to output welding quality indicators. Using historical data from the process database as samples, the optimal initial weights and thresholds are obtained through SAPSO optimization, followed by BP network training. During the online phase, real-time data is received, normalized, and input into the model, outputting prediction and optimization results. Inputs include plate thickness, bevel type, welding current, voltage, speed, environmental parameters, and equipment attitude. Outputs include weld penetration, weld width, angular deformation, transverse shrinkage rate, defect risk probability, and optimal welding parameter combination. The system interfaces with the real-time data acquisition and analysis module to obtain cleaned data; it collaborates with the digital twin model construction module to calibrate parameters; it provides a prediction kernel for the welding quality prediction module; and it outputs parameters and instructions to the welding process adaptive control module, forming a closed-loop management system.

[0072] The described dynamic adjustment method innovatively employs a multi-sensor fusion strategy to dynamically correct manufacturing assembly errors. A high-precision laser displacement sensor scans the workpiece surface point cloud, and combined with robot pose information, the actual assembly model is reconstructed. By comparing the actual assembly model with the theoretical design model, the three-dimensional offset of the weld start point is calculated, generating a global coordinate system offset command to correct the welding code. Simultaneously, a 100Hz high-frequency sampling arc sensor is introduced to capture welding current and voltage fluctuation signals in real time. Wavelet analysis of the signals decouples the weld's lateral / longitudinal sway, dynamically adjusting the welding torch pose to achieve closed-loop control for weld tracking. This technology overcomes the stringent requirements of traditional teach-and-write programming for assembly accuracy. Using the dynamic trajectory tracking technology of this system, the average repair time for typical mid-assembly welds is reduced by approximately 30% to 40% compared to traditional teach-and-write programming methods.

[0073] The real-time data acquisition and analysis module includes a gantry displacement sensor, a joint displacement sensor, a laser displacement sensor, an arc sensor, an industrial camera, a data processing and storage module, and a database.

[0074] The gantry displacement sensor is installed on the gantry, and the joint displacement sensor is installed at the joint of the welding robot arm to monitor the operating status of the gantry and the robot arm in real time, respectively; the laser displacement sensor is installed on the side of the welding torch to collect weld quality data; the arc sensor is integrated into the end of the welding torch for weld tracking; the industrial cameras are distributed at the welding site to collect visual information data; the data processing and storage module cleans and analyzes the data collected by each sensor and camera, and stores it in the database.

[0075] The data cleaning method first deduplicates the data by constructing a unique identifier key based on the acquisition timestamp and sensor number. Then, it sets effective value ranges according to the equipment range and welding process standards for each type of data, eliminating invalid data that exceeds the range. It uses the 3σ criterion combined with a sliding window to identify outliers. Minor anomalies are corrected by linear interpolation of the preceding and following effective data; severe anomalies are replaced by matching similar data from the same process history; and continuous anomalies trigger an early warning and are supplemented with simulation data from a digital twin model. Missing data is supplemented according to duration: short-duration missing data is filled forward, medium-duration missing data is linearly interpolated, and long-duration missing data is supplemented with simulation data from a digital twin. All numerical data is normalized to the [0,1] interval to eliminate dimensional differences. Simultaneously, Gaussian filtering is applied to laser displacement sensor data for noise reduction, 50Hz notch filtering is applied to arc sensor data to remove power frequency interference, and grayscale and median filtering are preprocessed on industrial camera vision data, eliminating invalid frames. Finally, the time axis, spatial coordinates, and logical consistency of all data are verified.

[0076] The data analysis method first performs basic statistical analysis on the cleaned data, calculating indicators such as mean, standard deviation, and coefficient of variation, and plots time-series trend curves to determine parameter stability and distribution characteristics. Then, it extracts time-domain, spatial-domain, and thermodynamic characteristic parameters combined with digital twin simulation to uncover process features strongly correlated with welding quality. Subsequently, it calculates the correlation between parameters and welding quality using the Pearson correlation coefficient, employs the Apriori algorithm to mine association rules between multi-parameter combinations and welding defects, constructs the optimal matching interval for welding process parameters, and calculates the current parameter matching degree. Finally, it performs single-parameter anomaly warnings based on process and statistical thresholds, multi-parameter collaborative anomaly warnings combined with association rules, and proactive anomaly warnings through trend prediction. The output includes warning results containing anomaly type, risk level, cause, and adjustment suggestions. All analysis results are structured and stored in a database, providing data support for digital twin model updates, welding quality prediction, and adaptive control.

[0077] The gantry displacement sensor is a laser displacement sensor, which collects the displacement and speed of the gantry in the X, Y, and Z axes in real time; the joint displacement sensor is an incremental encoder, which collects the joint rotation angle, angular velocity, and angular acceleration; the laser displacement sensor measures the weld bevel width and depth, as well as the weld height and width after forming; the arc sensor synchronously collects the welding current, voltage, and arc length; and the industrial camera is an industrial-grade high-speed camera that covers the entire welding area.

[0078] The digital twin model constructed by the assembly welding digital twin model construction module in the ship includes a geometric model, a physical model, a behavioral model, and a rule model. The geometric model is a parametric three-dimensional model of the assembled workpiece and welding equipment in the ship, which dynamically adjusts its shape, size, and relative position based on real-time data. The physical model adds material properties, mechanical properties, assembly constraints, and welding thermodynamic characteristics to the geometric model. The behavioral model describes the actual state of the welding equipment operation and weld formation. The rule model is the welding process operation law summarized from the data in the analysis database.

[0079] The geometric model is an integrated parametric 3D model of the assembled workpiece and welding equipment in a ship, constructed based on non-uniform rational B-splines (NURBS). The workpiece bevel size, weld formation size, and equipment motion parameters are set as adjustable parameters. The model coordinate system is consistent with the global coordinate system of the welding station. Based on real-time data acquisition, the shape, size, and relative position are dynamically adjusted through parametric drive. The specific adjustment method is calculated as follows:

[0080] Laser displacement sensor collects actual bevel width ,depth Height of formed weld ,width , with model initial parameters , , , In comparison, by , , Calculate the correction value and update the model parameters as follows: , , , Combined with the coefficient of thermal expansion Real-time temperature of the welding area Initial temperature ,according to Compensation for workpiece thermal deformation dimensions; gantry displacement according to collected X, Y, Z axis displacements. , , Combined with initial coordinates , , Real-time spatial coordinates are calculated using PM(X0+Xt,Y0+Yt,Z0+Zt); the welding robotic arm employs the DH parameter method, based on the real-time rotation angles of each joint. Through homogeneous transformation matrix Calculate the real-time coordinates of the welding torch tip; calculate the coordinates of the workpiece bevel center. Coordinates of the arc center of the welding torch relative deviation vector The relative positions of the equipment and workpiece in the model are corrected accordingly to ensure that the mapping error is ≤0.2mm. The model is adjusted at a frequency of 10ms / time and synchronized with the physical entity in real time.

[0081] The physical model adds material properties, mechanical properties, assembly constraints, and welding thermodynamics to the geometric model, as specifically defined below:

[0082] Material properties: The base material of the workpiece is marine steel, and the welding material is matching welding wire. The defined density is 7850 kg / m³, the specific heat capacity at room temperature is 450 J / (kg・K), the thermal conductivity is 50 W / (m・K), the melting point is 1538℃, and the Poisson's ratio is 0.3.

[0083] Mechanical properties: Marine steel has an elastic modulus of 206 GPa, yield strength of 345 MPa, tensile strength of 470 MPa, shear strength of welded joint of 300 MPa, and molten viscosity of welding material of 0.006 Pa·s.

[0084] Assembly constraints: The workpieces in the middle assembly are fixed by tack welding. The constraint types are surface-to-surface contact constraint and shaft-to-hole coaxial constraint. The tensile strength of the tack weld is ≥5kN and the workpiece assembly positioning deviation is ≤0.5mm.

[0085] Welding thermodynamic properties: welding arc thermal efficiency 0.75~0.85, molten pool solidification coefficient 2.5×10⁻ 6 m² / s, thermal expansion coefficient of the heat-affected zone is 1.2×10⁻ 5 / ℃, the radiation heat loss coefficient during welding is 0.1, and the convective heat transfer coefficient is 15W / (m²・K);

[0086] The behavioral model quantifies the actual state of welding equipment operation and weld formation. At the equipment operation level, it describes the uniform / acceleration / deceleration operation characteristics of the gantry X / Y / Z axis moving speed of 0.5~5mm / s, the rotation response characteristics of the welding robot arm joint rotation angle of -180°~180° (response delay ≤2ms), and the following characteristics of the welding torch attitude angle adjustment. At the weld formation level, it describes the dynamic correlation between welding current of 80~300A and voltage of 18~36V and weld penetration depth of 0.5~8mm and weld width of 2~10mm, as well as the real-time forming state of weld pool length of 5~20mm and width of 3~15mm as the welding speed changes, and synchronously maps the dynamic changes of weld reinforcement height and straightness during the welding process.

[0087] The rule model is based on historical welding data mining from the database to summarize the operational patterns of the welding process. Process parameter rules: When welding thick plates for ship assembly, a bevel width of 10-20mm corresponds to a welding current of 200-280A, a voltage of 28-36V, and a welding speed of 1-3mm / s. Environmental impact rules: When the humidity at the welding site is >60%, the welding current needs to be increased by 5%-10% to reduce the risk of porosity defects; when the gas flow rate is >1.5m / s, the shielding gas flow rate needs to be increased. Quality judgment rules: Weld penetration deviation ≤±0.5mm, weld width deviation ≤±0.8mm, and weld straightness ≤0.5mm / m are considered acceptable; an arc center offset >0.8mm triggers a weld deviation warning. Equipment operation rules: When the gantry movement speed fluctuation rate is >5% or the robotic arm joint angular velocity changes abruptly ≥2° / s, it is judged as abnormal equipment operation and parameter adjustments are triggered.

[0088] The welding condition monitoring and quality prediction module includes a welding condition monitoring module and a welding quality prediction module. The welding condition monitoring module updates the digital twin model based on real-time data to achieve a three-dimensional visualization of the welding process. The welding quality prediction module uses real-time data as initial conditions, and based on the digital twin model, it performs welding thermodynamic calculations and weld formation simulation analysis to achieve real-time detection and prediction of welding quality. It also continuously optimizes the calculation accuracy using data in the database through parameter identification methods.

[0089] Using the initial equipment status and process parameters of weld seam welding from the database as input, and real-time welding process data and final quality inspection results as output, this method identifies uncertain parameters x such as arc thermal efficiency and heat loss coefficient in thermodynamic calculations. The objective function is to minimize the sum of squared residuals between simulated and measured values.

[0090]

[0091] In the formula, The temperature value is for thermodynamic simulation. This is the actual measured temperature value. Given the number of data samples, the optimal parameters are solved iteratively using the least squares method. .

[0092] Using welding process and visual information data from the database as input, and final weld quality inspection data as output, this study addresses uncertain parameters such as the molten pool solidification coefficient and groove matching coefficient in the forming simulation. Identification, the objective function is:

[0093]

[0094] In the formula, These are the simulated values ​​for the forming dimensions. The optimal parameters are obtained iteratively using the gradient descent method, based on the measured forming dimensions. ,Will , Substitute the values ​​into the corresponding calculation model to complete the accuracy optimization.

[0095] The adaptive control module for the welding process includes a welding arc initiation point location module and a welding trajectory planning and tracking module. The welding arc initiation point location module calculates the optimal arc initiation point within a preset arc initiation area based on the prediction results of the intelligent optimization algorithm and the welding quality prediction module, and inputs the weld groove size, the initial posture of the welding equipment, and real-time environmental data stored in the database. The welding trajectory planning and tracking module drives the welding equipment control system according to the optimal welding parameters, and constructs a dynamic trajectory tracking closed loop by combining the three-dimensional geometric features of the weld and the feedback from the arc sensor and the laser displacement sensor.

[0096] The arc initiation point location module employs a particle swarm optimization (PSO) algorithm to calculate the optimal three-dimensional coordinates of the arc initiation point within a preset arc initiation area, aiming to achieve the best welding quality and the lowest defect risk. The specific calculations are as follows:

[0097] Objective function construction:

[0098] In the formula, The three-dimensional coordinates of the starting point of the arc; , The measured bevel width and depth at the starting point of the arc. , Standard bevel size; The probability of defect risk predicted for the quality at this location; , , Weighting coefficients ( (as specified by the process rules).

[0099] Particle velocity and position updates:

[0100] Speed ​​updates:

[0101] Location update:

[0102] In the formula, For inertial weights, , As a learning factor, , A random number between 0 and 1; This represents the optimal position for an individual particle. This is the globally optimal position for the particle swarm.

[0103] Constraints and Convergence Criteria

[0104] Preset arc initiation area constraints: , , (Determined by the bevel profile and the range of equipment movement); when ( To improve convergence accuracy, iteration is stopped when the number of iterations reaches 10⁻³ or the maximum number of iterations is reached, and the global optimal position is output. This is the optimal starting point for the arc.

[0105] The process of constructing the closed loop for dynamic trajectory tracking is as follows:

[0106] Trajectory pre-planning: Based on the three-dimensional geometric features of the weld seam from the digital twin model, combined with the optimal welding parameters, the basic motion trajectory of the welding torch (including X / Y / Z axis coordinates and attitude angle sequence) is generated.

[0107] Real-time feedback acquisition: The laser displacement sensor acquires the center coordinates and forming size deviation of the weld groove in real time, and the arc sensor synchronously acquires the arc center offset, with a sampling frequency of 100Hz;

[0108] Deviation calculation: Compare the pre-planned trajectory with the feedback data, calculate the position deviation Δ(x,y,z) and attitude deviation Δ(θ,ϕ), and establish the deviation vector;

[0109] Closed-loop control: The deviation vector is input into the equipment control system, and the adjustment amount is calculated through the proportional-integral-derivative (PID) algorithm to dynamically correct the welding torch position and attitude;

[0110] Continuous iteration: Repeated feedback acquisition - deviation calculation - control execution process to ensure trajectory tracking error ≤0.1mm, realizing dynamic closed-loop control of the welding process.

[0111] The control method for implementing a twin control system for the assembly and welding process in ships includes the following steps:

[0112] S1. Construct a welding digital twin model in the welding digital twin model construction module in the ship;

[0113] S2. During the welding process, the welding equipment status data is collected in real time through gantry displacement sensor and joint displacement sensor, the welding process data is collected in real time through laser displacement sensor and arc sensor, and the visual information data is collected in real time through industrial camera, while environmental data is also collected.

[0114] S3, the data processing and storage module cleans and analyzes the collected data and stores it in the database, and drives the updating of the welding digital twin model in real time;

[0115] S4, the welding quality prediction module, is based on processed real-time data and welding digital twin model. Through welding thermodynamic calculations and weld formation simulation analysis, it realizes real-time detection and prediction of welding quality.

[0116] S5, the welding status monitoring module dynamically displays the operating status of welding equipment, the weld formation process and changes in welding parameters based on real-time data, the updated welding digital twin model and prediction data, to achieve three-dimensional visualization monitoring of the welding process;

[0117] S6. Based on three-dimensional visualization monitoring, the welding process adaptive control module optimizes and determines the best combination of welding parameters based on intelligent optimization algorithms and welding quality prediction results, and drives the welding equipment control system to adjust the welding parameters.

[0118] S7. Repeat S2 to S6 to achieve continuous monitoring of the entire welding process and store all data in the database to improve the accuracy of welding thermodynamic calculations and weld formation simulation analysis.

[0119] The specific method for optimizing the calculation accuracy through parameter identification in S4 is as follows: taking the equipment status data and welding process parameters at the initial stage of welding of each weld segment stored in the database as input, and taking the real-time data and final quality inspection results during the welding process as output, the uncertain parameters in the welding thermodynamic calculation are identified; taking the welding process data and visual information data stored in the database as input, and taking the final weld quality inspection data as output, the uncertain parameters in the weld formation simulation analysis are identified.

[0120] Steps for identifying uncertain parameters in welding thermodynamic calculations

[0121] Data extraction: The initial equipment status data (gantry displacement, robotic arm joint parameters, etc.) and welding process parameters (current, voltage, welding speed, etc.) of the target weld are selected from the database as the input set, and the real-time temperature data of the welding process and the final quality inspection results (penetration depth, heat-affected zone range, etc.) are extracted as the output set.

[0122] Data preprocessing: Denoising, completion, and normalization are performed on the input and output data to remove outliers and ensure data consistency;

[0123] Initial parameter settings: Based on ship welding process standards, set the initial values ​​and reasonable ranges for uncertain parameters such as arc thermal efficiency and heat loss coefficient;

[0124] Simulation calculation: Substitute the input data and initial parameters into the welding thermodynamic model to calculate the simulation output values ​​such as temperature field and thermal cycle curve;

[0125] Error assessment: Calculate the sum of squared residuals between the simulation output value and the measured temperature and quality test results, and use it as an error evaluation index;

[0126] Parameter iterative optimization: The least squares method is used to iteratively adjust the uncertain parameters. The error is recalculated after each iteration until the error is less than the preset threshold (≤3%).

[0127] Parameter verification and solidification: The optimized parameters are substituted into the model for verification. After confirming that the simulation accuracy meets the requirements, the parameters are solidified and stored in the database for subsequent thermodynamic calculations.

[0128] Steps for identifying uncertain parameters in weld formation simulation analysis

[0129] Data extraction: Extract welding process data (groove size, current and voltage, etc.) and visual information data (molten pool size, weld characteristics, etc.) of the target weld from the database as the input set, and extract the final weld quality inspection data (penetration depth, weld width, reinforcement height, etc.) as the output set;

[0130] Data preprocessing: Denoising and completing the input and output data, unifying coordinates and units, and ensuring data validity;

[0131] Initial parameter settings: Based on welding process standards, set the initial values ​​and ranges of uncertain parameters such as molten pool solidification coefficient and groove matching coefficient;

[0132] Simulation calculation: Substitute the input data and initial parameters into the weld formation simulation model to obtain the simulated values ​​of the forming dimensions;

[0133] Error assessment: Calculate the sum of squared residuals between the simulated values ​​and the measured quality data, and use it as an error evaluation index;

[0134] Parameter iterative optimization: The gradient descent method is used to iteratively adjust the parameters until the error is ≤3%;

[0135] Parameter verification and solidification: The optimized parameters are substituted into the model for verification. Once the accuracy meets the standard, the parameters are solidified and stored in the database for subsequent forming simulation analysis.

[0136] The welding thermodynamics calculation in S4 is specifically as follows:

[0137] The formula is as follows, which combines the three-dimensional transient heat conduction equation with the Gaussian surface heat source model:

[0138] Formula for electric arc heat flux density:

[0139]

[0140] in, Heat flux density at a distance r from the center of the electric arc; Arc thermal efficiency (0.75~0.85); Welding voltage; Welding current; Effective heating radius of the electric arc; : Calculate the distance from the point to the center of the electric arc;

[0141] Transient heat conduction equation:

[0142]

[0143] in, Material density; Specific heat capacity at constant pressure, which varies with temperature; Instantaneous temperature; :time; Thermal conductivity, which varies with temperature; : Laplace operator; Arc heat flux density.

[0144] The intelligent optimization algorithm in the welding arc initiation point location module is as follows:

[0145] Objective function:

[0146]

[0147] in, Comprehensive quality evaluation indicators; : Three-dimensional coordinates of the starting point of the arc; : Deviation between measured and standard values ​​of bevel width; Standard bevel width; : Deviation between measured and standard values ​​of bevel depth; Standard bevel depth; : Weighting coefficients, summing to 1; Defect risk probability;

[0148] Particle update formula:

[0149]

[0150]

[0151] in, : The velocity of the i-th particle in the d-th dimension in the next round; Inertia weight (0.4~0.9); , Learning factor, all set to 2; , : 0~1 random number; The optimal position of an individual particle; The global optimal position of the particle swarm; : The position of the i-th particle in the d-th dimension in the next round.

[0152] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0153] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A twin control system for assembly and welding processes in ships, characterized in that: It includes a welding physical entity module, a real-time data acquisition and analysis module, a digital twin model construction module for ship assembly welding, a welding condition monitoring and quality prediction module, and a welding process adaptive control module. The physical entities to be welded include the assembled workpieces in the ship, the welding equipment, and the welding environment; The real-time data acquisition and analysis module is connected to the welding physical entity, collects welding equipment status data, welding process data, visual information data and environmental data, and processes and stores the data. The ship assembly welding digital twin model construction module is connected to the real-time data acquisition and analysis module. Based on the real-time data, a welding digital twin model is constructed and mapped to the welding physical entity. The welding status monitoring and quality prediction module is connected to the real-time data acquisition and analysis module and the digital twin model construction module to realize three-dimensional visualization monitoring of the welding process and real-time detection and prediction of welding quality. The welding process adaptive control module is connected to the welding status monitoring and quality prediction module and the welding equipment. It optimizes welding parameters and dynamically adjusts the welding trajectory based on the quality prediction results.

2. The twin control system for ship assembly and welding processes according to claim 1, characterized in that: The real-time data acquisition and analysis module includes a gantry displacement sensor, a joint displacement sensor, a laser displacement sensor, an arc sensor, an industrial camera, a data processing and storage module, and a database. The gantry displacement sensor is installed on the gantry, and the joint displacement sensor is installed at the joint of the welding robot arm to monitor the operating status of the gantry and the robot arm in real time, respectively; the laser displacement sensor is installed on the side of the welding torch to collect weld quality data; the arc sensor is integrated into the end of the welding torch for weld tracking; the industrial cameras are distributed at the welding site to collect visual information data; the data processing and storage module cleans and analyzes the data collected by each sensor and camera, and stores it in the database.

3. The twin control system for ship assembly and welding processes according to claim 2, characterized in that: The gantry displacement sensor is a laser displacement sensor, which collects the displacement and speed of the gantry in the X, Y, and Z axes in real time; the joint displacement sensor is an incremental encoder, which collects the joint rotation angle, angular velocity, and angular acceleration; the laser displacement sensor measures the weld bevel width and depth, as well as the weld height and width after forming; the arc sensor synchronously collects the welding current, voltage, and arc length; and the industrial camera is an industrial-grade high-speed camera that covers the entire welding area.

4. The twin control system for ship assembly and welding processes according to claim 1, characterized in that: The digital twin model constructed by the assembly welding digital twin model construction module in the ship includes a geometric model, a physical model, a behavioral model, and a rule model. The geometric model is a parametric three-dimensional model of the assembled workpiece and welding equipment in the ship, which dynamically adjusts its shape, size, and relative position based on real-time data. The physical model adds material properties, mechanical properties, assembly constraints, and welding thermodynamic characteristics to the geometric model. The behavioral model describes the actual state of the welding equipment operation and weld formation. The rule model is the welding process operation law summarized from the data in the analysis database.

5. The twin control system for ship assembly and welding processes according to claim 1, characterized in that: The welding condition monitoring and quality prediction module includes a welding condition monitoring module and a welding quality prediction module. The welding condition monitoring module updates the digital twin model based on real-time data to achieve a three-dimensional visualization of the welding process. The welding quality prediction module uses real-time data as initial conditions, and based on the digital twin model, it performs welding thermodynamic calculations and weld formation simulation analysis to achieve real-time detection and prediction of welding quality. It also continuously optimizes the calculation accuracy using data from the database through parameter identification methods.

6. The twin control system for ship assembly and welding processes according to claim 1, characterized in that: The adaptive control module for the welding process includes a welding arc initiation point location module and a welding trajectory planning and tracking module. The welding arc initiation point location module calculates the optimal arc initiation point within a preset arc initiation area based on the prediction results of the intelligent optimization algorithm and the welding quality prediction module, and inputs the weld groove size, the initial posture of the welding equipment, and real-time environmental data stored in the database. The welding trajectory planning and tracking module drives the welding equipment control system according to the optimal welding parameters, and constructs a dynamic trajectory tracking closed loop by combining the three-dimensional geometric features of the weld and the feedback from the arc sensor and the laser displacement sensor.

7. A control method for implementing the twin control system for the assembly and welding process in a ship as described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Construct a welding digital twin model in the welding digital twin model construction module in the ship; S2. During the welding process, the welding equipment status data is collected in real time through gantry displacement sensor and joint displacement sensor, the welding process data is collected in real time through laser displacement sensor and arc sensor, and the visual information data is collected in real time through industrial camera, while environmental data is also collected. S3, the data processing and storage module cleans and analyzes the collected data and stores it in the database, and drives the updating of the welding digital twin model in real time; S4, the welding quality prediction module, is based on processed real-time data and welding digital twin model. Through welding thermodynamic calculations and weld formation simulation analysis, it realizes real-time detection and prediction of welding quality. S5, the welding status monitoring module, dynamically displays the operating status of welding equipment, weld formation process and changes in welding parameters based on real-time data, updated welding digital twin model and prediction data, to achieve three-dimensional visualization monitoring of the welding process; S6. Based on three-dimensional visualization monitoring, the welding process adaptive control module optimizes and determines the best combination of welding parameters based on intelligent optimization algorithms and welding quality prediction results, and drives the welding equipment control system to adjust the welding parameters. S7. Repeat S2 to S6 to achieve continuous monitoring of the entire welding process and store all data in the database to improve the accuracy of welding thermodynamic calculations and weld formation simulation analysis.

8. The control method of the twin control system for ship assembly and welding processes according to claim 7, characterized in that: The specific method for optimizing the calculation accuracy through parameter identification in S4 is as follows: taking the equipment status data and welding process parameters at the initial stage of welding of each weld segment stored in the database as input, and taking the real-time data and final quality inspection results during the welding process as output, the uncertain parameters in the welding thermodynamic calculation are identified. Using welding process data and visual information data stored in the database as input, and the final weld quality inspection data as output, the uncertain parameters in the weld formation simulation analysis are identified.

9. The control method of the twin control system for ship assembly and welding processes according to claim 7, characterized in that: The welding thermodynamics calculation in S4 is specifically as follows: The three-dimensional transient heat conduction equation is combined with the Gaussian surface heat source model, and the formula is as follows: Formula for electric arc heat flux density: ； in, Heat flux density at a distance r from the center of the electric arc; Arc thermal efficiency; Welding voltage; Welding current; Effective heating radius of the electric arc; : Calculate the distance from the point to the center of the electric arc; Transient heat conduction equation: ； in, Material density; Specific heat capacity at constant pressure, which varies with temperature; Instantaneous temperature; :time; Thermal conductivity, which varies with temperature; : Laplace operator; Arc heat flux density.

10. The control method of the twin control system for assembly and welding processes in ships according to claim 7, characterized in that: The intelligent optimization algorithm in the welding arc initiation point location module is as follows: Objective function: ； in, Comprehensive quality evaluation indicators; : Three-dimensional coordinates of the starting point of the arc; : Deviation between measured and standard values ​​of bevel width; Standard bevel width; : Deviation between measured and standard values ​​of bevel depth; Standard bevel depth; Weighting coefficients, summing to 1; Defect risk probability; Particle update formula: ； ； in, : The velocity of the i-th particle in the d-th dimension in the next round; Inertia weight; , Learning factor, all set to 2; , : 0~1 random number; The optimal position of an individual particle; The global optimal position of the particle swarm; : The position of the i-th particle in the d-th dimension in the next round.

Images