A method and system for construction optimization control of offshore structures based on digital twinning

Abstract

The application relates to the technical field of data processing, and particularly discloses a marine structure construction optimization control method and system based on digital twinning. The application obtains measured strain data by establishing a stress sensing network, simultaneously constructs a first coupled finite element reference model on the digital side to obtain simulation strain data, screens sensitive parts based on the difference between the measured strain and the simulation strain, constructs an error function, and optimizes welding heat source parameters by using an inversion algorithm to obtain a second coupled finite element reference model that is corrected and calibrated, generates residual stress response samples under various welding sequences by using the second reference model, trains an AI agent model, realizes rapid prediction of residual stress under different construction paths, calls the AI agent model to evaluate the remaining welding sequence according to the current construction state, and selects the path with the smallest residual stress increment as the target construction path, so that stress active control and path optimization in the construction process are realized, and construction safety and efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and system for optimizing the construction control of marine engineering structures based on digital twins. Background Technology

[0002] Deep-sea floating wind turbine platforms are characterized by their large size, enormous weight, complex structure, and numerous connections, and their construction process is extremely complex, posing significant challenges to construction technology. Throughout the entire construction process (such as critical manufacturing, welding, and hoisting), the combined effects of the structure's internal self-weight load, construction load, and environmental load generate complex stresses and deformations, directly affecting subsequent construction. In particular, wind power foundation equipment, represented by substations and jacket structures, are complex, irregular, large spatial structures with numerous pipe nodes. During the welding and manufacturing process, improper assembly and welding sequences can generate enormous internal residual stresses.

[0003] If these residual stresses cannot be accurately predicted and controlled, they will significantly affect the fatigue life and service performance of wind power foundation equipment. On the one hand, existing construction control methods for large marine structures mainly rely on finite element analysis to simulate and evaluate the welding process, and combine engineering experience to pre-determine the assembly and welding sequence. However, traditional finite element models are usually built based on idealized assumptions, and their welding heat source parameters, material properties, and boundary conditions are mostly derived from experience. They are difficult to accurately reflect the real stress state caused by assembly errors, welding process fluctuations, and other factors in the actual construction site, resulting in a large deviation between the simulation results and the actual structural state. On the other hand, large jacket structures and booster station structures have highly nonlinear and large-scale characteristics. The thermo-elastoplastic finite element analysis of their welding process is computationally intensive, and a single simulation usually takes several hours or even days, which is difficult to meet the needs of real-time decision-making and path adjustment in the construction site. Therefore, traditional construction path optimization methods based on high-precision finite element calculations have problems of response lag and insufficient operability in practical engineering applications.

[0004] Therefore, it is necessary to provide a digital twin-based method and system for optimizing the construction control of marine engineering structures to solve the above-mentioned technical problems. In order to solve the above problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of existing technologies, this invention provides a digital twin-based method and system for optimizing and controlling the construction of marine engineering structures. This method addresses the problem that traditional finite element simulation is unable to accurately and quickly predict residual stress during the construction of deep-sea floating wind turbine platforms due to the complexity of the structure, the numerous welding processes, and the superposition of loads, thereby affecting construction optimization and structural service performance.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A digital twin-based method for optimizing and controlling the construction of marine engineering structures includes the following steps:

[0008] By acquiring the structural parameters and construction process information of large marine structures, strain sensors are deployed at the key stress-bearing parts of the marine structure to be built to form a stress sensing network to obtain measured strain data. On the digital side, a first coupled finite element reference model corresponding to the marine structure to be built is established to obtain simulated strain data, and the initial parameters of the welding heat source model are preset.

[0009] Based on measured strain data and simulated strain data, sensitive stress-bearing parts of marine structures are screened, and anomaly detection is carried out for sensitive stress-bearing parts. Then, an error function reflecting the difference between the two is constructed. With the goal of minimizing the error function, the input parameters of the welding heat source model are reverse-calibrated and corrected using an inversion algorithm to obtain the second coupled finite element reference model.

[0010] Based on the second coupled finite element benchmark model, the construction process of marine structures under different welding sequences is simulated, and an AI proxy model is used to predict the residual stress response under different construction paths.

[0011] Based on the current construction status information of the marine structure, the AI ​​agent model is invoked to evaluate the remaining optional welding sequences, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results.

[0012] As a further aspect of the present invention, the key stress-bearing parts include K-type pipe nodes and / or T-type pipe nodes where welding stress is concentrated; the initial parameters of the welding heat source model include welding heat source power and heat source distribution radius.

[0013] As a further aspect of the present invention, the welding heat source model is used in numerical simulation to equivalently describe the distribution characteristics of welding heat input in the weld area along the spatial and temporal directions. By converting welding process parameters into thermal loads and inputting them into the first coupled finite element model, the influence of the welding process on the temperature field and stress field of the marine structure is simulated.

[0014] As a further aspect of the present invention, sensitive stress-bearing components of marine engineering structures are screened based on measured strain data and simulated strain data, and anomaly detection is performed on these sensitive stress-bearing components. The specific steps are as follows:

[0015] Based on the measured strain data within a preset sampling period and the simulated strain data output in real time, the measured strain time series is obtained by arranging the measured strain data in time sequence, and the simulated strain time series is obtained by arranging the simulated strain data in time sequence.

[0016] The first strain fluctuation value is calculated by adjacent sampling points within the measured strain time series, and the second strain fluctuation value is obtained by calculating adjacent sampling points within the simulated strain time series.

[0017] The sensitive stress-bearing parts of the marine structure are identified by using the fluctuation difference between the first strain fluctuation value and the second strain fluctuation value.

[0018] As a further aspect of the present invention, the measured strain time series is as follows: ,in, This represents the measured strain data corresponding to the i-th sampling point. This refers to the number of samples.

[0019] The simulated strain time series is ,in, This represents the simulated strain data corresponding to the i-th sampling point.

[0020] As a further aspect of the present invention, the specific method for screening and determining the sensitive stress-bearing parts of the marine structure is as follows: the fluctuation difference is compared with a preset fluctuation threshold. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the corresponding key stress-bearing part is marked as a sensitive stress-bearing part. At this time, an anomaly is detected in the sensitive stress-bearing part, triggering an alert. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the screening steps are repeated.

[0021] As a further aspect of the present invention, an error function reflecting the difference between the two is constructed. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model to obtain a second coupled finite element reference model. The specific steps are as follows:

[0022] Based on the measured strain data output by the stress sensing network during the construction of marine structures in real time, the measured strain data is compared with the simulated strain data of the corresponding virtual measuring points in the finite element benchmark model to obtain the strain difference.

[0023] An error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model, thereby obtaining the second coupled finite element reference model.

[0024] As a further aspect of the present invention, the measured strain data is compared with the simulated strain data of the corresponding virtual measuring points in the finite element reference model to obtain the strain difference value. The specific steps are as follows:

[0025] Through a stress sensing network, measured strain data of marine structures are collected in real time according to a preset sampling period. The measured strain data is then processed for time synchronization and outlier removal to form a measured strain time series that progresses with the construction process of the marine structure.

[0026] In the first coupled finite element reference model, virtual measuring points corresponding one-to-one with the spatial positions of each strain sensor in the stress sensing network are extracted. Based on the current welding process and the initial parameters of the welding heat source model, the simulated strain data of the virtual measuring points are calculated to form a simulated strain time series.

[0027] By aligning the measured strain time series with the simulated strain time series on the time axis and comparing them, the strain difference at corresponding moments is calculated.

[0028] As a further aspect of the present invention, an error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model to obtain a second coupled finite element reference model. The specific steps are as follows:

[0029] An error function is constructed based on the strain difference to quantify the overall deviation between measured strain data and simulated strain data.

[0030] Input parameters that significantly affect the temperature and stress fields are selected from the welding heat source model as calibration and correction parameters, and these parameters are used as optimization variables for the inversion algorithm.

[0031] With minimizing the error function as the optimization objective, the parameters to be calibrated and corrected are iteratively updated using an inversion algorithm. In each iteration, the simulation strain data of the finite element model is recalculated based on the updated welding heat source model parameters, and the error function value is recalculated until the error function converges or the preset termination condition is met.

[0032] When the error function meets the preset convergence condition, the corresponding welding heat source model parameters are output as the calibrated and corrected welding heat source parameters. The calibrated and corrected welding heat source parameters are then updated in the first coupled finite element model to obtain a second finite element reference model that can reflect the actual welding heat input and structural stress state.

[0033] A digital twin-based construction optimization and control system for marine engineering structures includes a structural data acquisition module, a sensitive part screening and anomaly detection module, a welding heat source inversion and correction module, a construction path simulation and AI agent model construction module, and a welding sequence optimization decision module.

[0034] The structural data acquisition module is used to acquire structural parameters and construction process information of large marine structures, deploy strain sensors at key stress-bearing parts of the marine structure to be built to form a stress sensing network to acquire measured strain data, establish a first coupled finite element reference model corresponding to the marine structure to be built on the digital side to acquire simulated strain data, and preset the initial parameters of the welding heat source model.

[0035] The sensitive component screening and anomaly detection module is used to screen sensitive stress components of marine structures based on measured strain data and simulated strain data, and to perform anomaly detection on the sensitive stress components.

[0036] The welding heat source inversion and correction module is used to construct an error function that reflects the difference between the two. With the goal of minimizing the error function, the inversion algorithm is used to perform reverse calibration and correction on the input parameters of the welding heat source model to obtain the second coupled finite element reference model.

[0037] The construction path simulation and AI proxy model are used to build modules based on the second coupled finite element benchmark model to simulate the construction process of marine structures under different welding sequences and predict the AI ​​proxy model of residual stress response under different construction paths.

[0038] The welding sequence optimization decision module is used to evaluate the remaining optional welding sequences based on the current construction status information of the marine structure, call the AI ​​agent model, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results.

[0039] The technical effects and advantages of this invention, a digital twin-based optimization control method and system for marine structure construction, are as follows: This invention achieves refined control of the welding construction process of marine structures through a closed-loop mechanism of perception, modeling, and intelligent prediction. By deploying strain sensors at key stress-bearing locations and forming a stress-sensing network, the actual stress state of the structure is acquired in real time, improving the ability to perceive stress changes during construction. An error function is constructed based on the difference between measured and simulated strain, and welding heat source parameters are calibrated and corrected using an inversion algorithm, enabling the second coupled finite element benchmark model to more accurately reflect the actual welding heat input and structural stress distribution, thereby improving the accuracy of simulation prediction. The second benchmark model is then used to generate residual stress response samples under different welding sequences and train an AI proxy model, enabling rapid residual stress prediction of the construction path, avoiding computationally intensive repeated finite element simulations, and saving time and computing resources. By combining the current construction status with the AI ​​proxy model to evaluate the remaining welding sequence, the path with the smallest residual stress increment is selected as the target construction path, achieving proactive stress control and optimization during the construction process, improving construction safety and structural reliability, while reducing trial-and-error costs and construction risks, and realizing intelligent, refined, and efficient construction management. Attached Figure Description

[0040] Figure 1 A flowchart illustrating an optimized control method for marine structure construction based on digital twins, provided as an embodiment of the present invention;

[0041] Figure 2 This is a system block diagram of a marine structure construction optimization control system based on digital twin, provided for an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described technical solutions are only a part of this invention, and not all of it. All other technical solutions obtained by those skilled in the art based on the technical solutions of this invention without inventive effort are within the scope of protection of this invention.

[0043] like Figure 1 The diagram shown is a flowchart of a digital twin-based optimization control method for marine structure construction, provided by an embodiment of the present invention. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. Steps S1 to S4 are detailed as follows:

[0044] Step S1: By acquiring the structural parameters and construction process information of the large marine structure, strain sensors are arranged at the key stress-bearing parts of the marine structure to be built to form a stress sensing network to acquire measured strain data. On the digital side, a first coupled finite element reference model corresponding to the marine structure to be built is established to acquire simulated strain data, and the initial parameters of the welding heat source model are preset.

[0045] Step S2: Based on the measured strain data and simulated strain data, the sensitive stress-bearing parts of the marine structure are screened, and anomaly detection is performed on the sensitive stress-bearing parts. Then, an error function reflecting the difference between the two is constructed. With the goal of minimizing the error function, the input parameters of the welding heat source model are reverse-calibrated and corrected using an inversion algorithm to obtain the second coupled finite element reference model.

[0046] Step S3: Based on the second coupled finite element benchmark model, simulate the construction process of marine structures under different welding sequences, and predict the AI ​​proxy model of residual stress response under different construction paths.

[0047] Step S4: Based on the current construction status information of the marine structure, call the AI ​​agent model to evaluate the remaining optional welding sequences, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results.

[0048] Preferably, the key stress-bearing parts include K-type pipe nodes and / or T-type pipe nodes where welding stress is concentrated; the initial parameters of the welding heat source model include the welding heat source power and the heat source distribution radius.

[0049] The steps for building a stress sensing network are as follows: Based on the structural topology and construction process information, stress analysis is performed on the marine structure to be constructed to identify key stress-bearing parts with high stress concentration risk during welding. Key stress-bearing parts include K-type pipe nodes, T-type pipe nodes, and high welding density connection areas. Strain sensor arrays are arranged at the key stress-bearing parts, and structural strain data is collected according to the preset sampling frequency to form a stress sensing network covering the key nodes.

[0050] Preferably, the welding heat source model is used in numerical simulation to equivalently describe the distribution characteristics of welding heat input in the weld area along the spatial and temporal directions. By converting welding process parameters into thermal loads and inputting them into the first coupled finite element model, the influence of the welding process on the temperature field and stress field of the marine structure can be simulated. The welding heat source model can be a double ellipsoidal heat source model, a disk heat source model, a conical heat source model, or a combination thereof.

[0051] The numerical simulation using a welding heat source model in this invention has significant technical advantages. First, by describing the distribution characteristics of welding heat input along the spatial and temporal directions within the weld region in an equivalent manner in finite element analysis, the complex and difficult-to-measure actual welding physical process is transformed into a calculable and adjustable form of thermal load. This effectively solves the problems of uneven heat input and difficulty in accurately characterizing transient changes in the heat-affected zone during welding, significantly improving the controllability and stability of the welding temperature and stress field simulation.

[0052] Secondly, process parameters such as welding current, voltage, and welding speed are mapped as input parameters of the welding heat source model and loaded into the first thermo-structural coupled finite element model. This enables the simulation model to realistically reflect the differences in heat input under different welding process conditions and their impact on structural deformation and residual stress evolution. This is beneficial for realizing the quantitative correlation between welding process parameters and structural response, and provides a reliable physical basis for subsequent process optimization and path adjustment.

[0053] Furthermore, by employing double ellipsoidal heat source models, disk heat source models, conical heat source models, or combinations thereof, models can be flexibly selected or combined for different welding methods, weld morphologies, and plate thickness conditions. This allows the heat source distribution to more closely resemble the energy transfer characteristics of the actual weld pool, thereby improving the accuracy of simulation results in terms of temperature gradient, weld depth prediction, and residual stress distribution, and enhancing the applicability of the model in complex large-scale marine engineering construction scenarios.

[0054] Finally, the welding heat source model parameters have good adjustability and inverseability, which facilitates online calibration, correction and iterative update of the model parameters by combining measured strain data. This allows the numerical simulation model to evolve dynamically with the construction process and maintain a high degree of consistency with the actual structural state. This provides a stable and reliable foundation for subsequent welding sequence optimization and stress control based on digital twins, and improves the overall safety and quality control of the construction process of large marine engineering structures.

[0055] Preferably, the specific steps for establishing the first coupled finite element reference model corresponding to the marine structure to be constructed are as follows:

[0056] Based on the design drawings of the marine engineering structure to be built, the geometric parameters of the marine engineering structure to be built are extracted, and the geometric parameters are defined parametrically. The geometric parameters include component length, pipe diameter, wall thickness and node connection form. A three-dimensional geometric model of the structure is established based on the geometric parameters.

[0057] Material properties are set for each structural component, and the mechanical and thermal parameters of the material are parameterized. The material parameters include at least the elastic modulus, Poisson's ratio, density, thermal conductivity, specific heat capacity, and coefficient of thermal expansion. According to the welding process requirements, some material parameters are set as functions that change with temperature to reflect the nonlinear behavior of the material during the welding process.

[0058] By introducing a welding heat source model into the three-dimensional geometric model, the distribution characteristics of heat input in the weld area along the spatial and temporal directions during the welding process are equivalently described. The initial parameters of the welding heat source model are parameterized, and the welding sequence is used as a time control parameter to drive the movement of the heat source in the model.

[0059] By discretizing the marine structure using the finite element method, a heat conduction analysis model was established. A welding heat source model was applied as a heat load to the heat conduction analysis model, and convection heat transfer and radiation heat transfer boundary conditions were set to simulate the heat exchange process between the marine structure and the surrounding environment during the welding process.

[0060] A structural mechanics analysis model is established based on finite element discretization, and the thermal load obtained from the heat conduction analysis model is input into the structural mechanics analysis model. Combined with the constraints of the marine structure and the construction stage settings, the thermal deformation and stress response of the structure during the welding process are calculated.

[0061] The heat conduction analysis model is coupled with the structural mechanics analysis model to form the first coupled finite element model, and virtual measuring points corresponding one-to-one with the strain sensor placement positions are set to output simulated strain data.

[0062] In this embodiment of the invention, establishing the first coupled finite element reference model corresponding to the marine structure to be constructed has significant technical and engineering application advantages. Firstly, by comprehensively defining the geometric and material parameters of the marine structure, the finite element model can maintain a high degree of consistency with the design drawings, while also possessing good scalability and reconfigurability. When the structural dimensions, component forms, or material schemes change, there is no need to rebuild the model; only parameter adjustments are required to quickly update the model, significantly improving model construction efficiency. This is particularly suitable for the construction of large-scale marine structures with complex configurations and enormous dimensions.

[0063] Secondly, by introducing temperature-dependent functions of mechanical and thermal parameters into the material properties, the model can truly reflect the nonlinear behavior of material properties changing drastically with temperature during welding. This effectively improves the accuracy of calculations of transient temperature field, thermal deformation, and residual stress during welding, avoiding the underestimation or overestimation of stress caused by traditional linear assumptions, and providing a reliable numerical basis for subsequent stress control and risk assessment.

[0064] Furthermore, by introducing a welding heat source model into the three-dimensional geometric model and using the welding sequence as a time control parameter to drive the movement of the heat source, the model can not only characterize the local heat input characteristics of a single weld, but also reflect the overall influence of different welding sequences and welding rhythms on the overall temperature field and stress evolution path of the structure, providing a direct and quantifiable simulation basis for welding path optimization and construction sequence evaluation.

[0065] Furthermore, by establishing separate heat conduction analysis models and structural mechanics analysis models, and achieving heat-structure coupling at the finite element level, heat input, heat exchange, structural constraints, and construction stage effects can be considered simultaneously within the same computational framework. This allows the model to fully describe the coupling effects of thermal expansion, structural deformation, and stress concentration caused by temperature rise during the welding process, significantly improving the consistency between simulation results and actual construction processes.

[0066] Finally, virtual measuring points corresponding one-to-one with the strain sensor placement locations are set in the coupled finite element model, enabling the simulation output to be directly aligned with the field-measured strain data. This provides a clear data interface for subsequent model calibration, heat source parameter inversion, and real-time updates of the digital twin. This "measurement-simulation" comparable modeling method effectively establishes a data loop between the physical structure and the digital model, laying a solid foundation for real-time stress monitoring and intelligent optimization during the construction process of marine engineering structures.

[0067] Preferably, sensitive stress-bearing components of the marine structure are screened based on measured strain data and simulated strain data, and anomaly detection is performed on these sensitive stress-bearing components. The specific steps are as follows:

[0068] Based on the measured strain data within a preset sampling period and the simulated strain data output in real time, the measured strain time series is obtained by arranging the measured strain data in chronological order. ,in, This represents the measured strain data corresponding to the i-th sampling point. To determine the number of samples, the simulated strain time series is obtained by arranging the simulated strain data in chronological order. ,in, This represents the simulated strain data corresponding to the i-th sampling point;

[0069] The first strain fluctuation value was calculated from adjacent sampling points within the measured strain time series. The second strain fluctuation value is obtained by calculating adjacent sampling points within the simulated strain time series. ;

[0070] Using the fluctuation difference between the first strain fluctuation value and the second strain fluctuation value The process involves screening and identifying sensitive stress-bearing parts of the marine structure. Specifically, the fluctuation difference is compared with a preset fluctuation threshold. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the corresponding key stress-bearing part is marked as a sensitive stress-bearing part. At this time, an anomaly is detected in the sensitive stress-bearing part, triggering an alert. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the screening process is repeated.

[0071] This invention employs the aforementioned method for screening sensitive stress components and detecting anomalies based on a combination of measured and simulated strain data, demonstrating significant technical effectiveness and engineering application value. Firstly, by aligning and analyzing the measured and simulated strain time series within a unified sampling period and timeframe, it avoids environmental noise interference or operational fluctuations caused by relying solely on absolute value thresholds. This ensures anomaly identification is no longer dependent on single-point sensor data but is based on a dual "measured-model" reference, significantly improving the reliability and robustness of anomaly detection. Secondly, by calculating and normalizing the strain fluctuation values ​​between adjacent sampling points, it effectively reflects the rate of strain change and fluctuation trend over time, rather than focusing solely on static strain levels. This fluctuation-based analysis method enables the system to sensitively capture sudden local stress changes during welding or construction, particularly suitable for identifying abnormal responses caused by uneven welding heat input, changes in constraint conditions, or sudden changes in structural stiffness, thus enhancing the ability to identify early anomalies.

[0072] By quantifying the difference between measured and simulated strain fluctuations and using this fluctuation difference as a sensitivity criterion, the degree of deviation between the actual structural behavior and the predictions of the digital twin model can be directly reflected. When the fluctuation difference exceeds a preset threshold, it indicates that the actual response of the stressed part has exceeded the normal evolution range of the model, thereby achieving accurate location of potential risks and effectively avoiding false alarms or missed alarms caused by model errors or measurement errors alone.

[0073] In addition, by screening key stress points point by point in an iterative manner, and automatically marking them as sensitive stress points and triggering alerts when conditions are met, it is possible to realize the real-time detection and continuous tracking of abnormal states. This allows construction personnel or control systems to take timely intervention measures such as process adjustment, path optimization, or local reinforcement to prevent abnormal stress from accumulating and expanding further.

[0074] The embodiments of this invention implement anomaly determination in a simple and clear mathematical form. The calculation process is clear and easy to embed into online monitoring and edge computing systems, which not only ensures real-time performance but also has good engineering feasibility. It provides reliable support for intelligent monitoring, risk warning and active stress control of key stress parts in the construction of large marine engineering structures.

[0075] Preferably, an error function reflecting the difference between the two is constructed. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model to obtain the second coupled finite element reference model. The specific steps are as follows:

[0076] Based on the measured strain data output by the stress sensing network during the construction of marine structures in real time, the measured strain data is compared with the simulated strain data of the corresponding virtual measuring points in the finite element benchmark model to obtain the strain difference.

[0077] An error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to perform reverse calibration and correction on the input parameters of the welding heat source model, resulting in a second coupled finite element reference model.

[0078] This invention employs the aforementioned welding heat source model inversion calibration and correction method based on error function minimization, which significantly improves the consistency and reliability between the coupled finite element model and the actual construction process. By comparing the measured strain data collected in real time by the stress-sensing network with the simulated strain data of the corresponding virtual measuring points in the finite element benchmark model point by point, the deviation between the actual structural response and the numerical model prediction is directly quantified. This eliminates reliance on empirical parameter tuning and instead establishes the model based on real working condition data, thereby improving the objectivity and traceability of model parameter calibration and correction from the source. An error function is constructed using the strain difference, and minimizing the error function is used as the optimization objective, giving the adjustment process of the welding heat source model input parameters clear mathematical constraints and optimization directions. The inversion algorithm automatically searches for the optimal parameter combination, enabling systematic calibration and correction of key parameters such as heat source intensity, distribution pattern, and temporal evolution characteristics. This avoids the inefficiency and uncertainty caused by repeated manual calculations, significantly improving model calibration efficiency.

[0079] By reversing the calibration and correction of the welding heat source model parameters, the second coupled finite element benchmark model can dynamically absorb the actual heat input characteristics and structural response characteristics generated during on-site construction. This more realistically reflects the combined impact of welding process fluctuations, changes in the construction environment, and the evolution of structural constraints on the temperature and stress fields, effectively narrowing the gap between simulation and actual measurement. Once the second coupled finite element benchmark model is formed, it can serve as a highly reliable data source for subsequent welding sequence simulation, residual stress prediction, and AI proxy model training. This provides a more accurate physical basis for subsequent construction path optimization and stress control, preventing errors from accumulating and amplifying in the digital twin system.

[0080] The embodiments of the present invention constitute a closed-loop update process of "measured data - model calibration and correction - re-prediction", which enables the finite element model to continuously evolve with the construction process and remain synchronized with the actual structural state, providing key technical support for realizing real-time digital twins, accurate stress assessment and proactive risk management in the construction of large marine engineering structures.

[0081] Preferably, the measured strain data is compared with the simulated strain data of the corresponding virtual measuring points in the finite element reference model to obtain the strain difference value. The specific steps are as follows:

[0082] Through a stress sensing network, measured strain data of marine structures are collected in real time according to a preset sampling period. The measured strain data is then processed for time synchronization and outlier removal to form a measured strain time series that progresses with the construction process of the marine structure.

[0083] In the first coupled finite element reference model, virtual measuring points corresponding one-to-one with the spatial positions of each strain sensor in the stress sensing network are extracted. Based on the current welding process and the initial parameters of the welding heat source model, the simulated strain data of the virtual measuring points are calculated to form a simulated strain time series.

[0084] This invention compares the measured strain time series with the simulated strain time series on the time axis to calculate the strain difference at corresponding moments. This method of obtaining the strain difference by comparing measured and simulated strain data effectively improves the accuracy and engineering applicability of monitoring and model calibration during the construction process of marine structures. First, a stress-sensing network continuously collects strain data from the marine structure according to a preset sampling period. Time synchronization and outlier removal of the measured data effectively eliminate interference from sensor noise, communication delays, and occasional abnormal readings, ensuring that the resulting measured strain time series accurately and stably reflects the actual stress state of the structure as construction progresses. Virtual measuring points, corresponding one-to-one with the spatial locations of strain sensors, are set in the first coupled finite element reference model, ensuring strict consistency between the simulation output and the on-site measured data in both spatial location and physical meaning, thus avoiding comparison errors caused by mismatched measuring points. This "virtual-real alignment" modeling method provides a unified data foundation for subsequent strain difference calculation and model inversion, significantly improving the accuracy and interpretability of the comparative analysis.

[0085] The simulated strain time series calculated based on the initial parameters of the current welding process and welding heat source model can fully reflect the numerical model's prediction of the structural response under given process conditions. By precisely aligning this with the measured strain time series on the time axis and comparing them moment-by-moment, the dynamic deviation between the actual construction process and the model prediction can be clearly depicted. This helps identify response differences caused by model parameter mismatch or fluctuations in the construction process. By calculating the strain difference at corresponding moments, not only can the static deviation be obtained, but the evolution of the deviation over time and construction stages can also be analyzed. This provides a direct basis for determining the source of the deviation, identifying abnormal stress locations, and triggering model parameter inversion calibration and correction. This time series difference-based analysis method is beneficial for early warning and preventing the gradual accumulation of stress anomalies in the structure.

[0086] The embodiments of this invention have a clear process, low implementation cost, and are easy to embed into digital twin systems and online monitoring platforms. They can serve as a key interface between measured data and simulation models, continuously supporting subsequent calibration and correction of welding heat source models, updating of coupled finite element models, and optimization of construction paths, providing a reliable data foundation for intelligent management and control of the construction process of large marine engineering structures.

[0087] Preferably, an error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to perform reverse calibration and correction on the input parameters of the welding heat source model to obtain the second coupled finite element reference model. The specific steps are as follows:

[0088] An error function is constructed based on the strain difference to quantify the overall deviation between measured and simulated strain data. The formula for calculating the error function is as follows: ;

[0089] Input parameters that significantly affect the temperature and stress fields are selected from the welding heat source model as calibration and correction parameters, which are then used as optimization variables for the inversion algorithm; these calibration and correction parameters are the initial parameters of the welding heat source model.

[0090] With minimizing the error function as the optimization objective, the parameters to be calibrated and corrected are iteratively updated using an inversion algorithm. In each iteration, the simulation strain data of the finite element model is recalculated based on the updated welding heat source model parameters, and the error function value is recalculated until the error function converges or the preset termination condition is met.

[0091] When the error function meets the preset convergence condition, the corresponding welding heat source model parameters are output as the calibrated and corrected welding heat source parameters. The calibrated and corrected welding heat source parameters are then updated in the first coupled finite element model to obtain the second finite element reference model that can reflect the actual welding heat input and structural stress state.

[0092] The method of calibrating and correcting the input parameters of the welding heat source model by constructing an error function based on strain difference and minimizing it, through an inversion algorithm, has significant advantages. First, by quantifying the overall deviation between measured and simulated strain using the error function, the differences between heat input and structural response during welding can be systematically captured, enabling a comprehensive evaluation of model accuracy, rather than relying solely on local data or single-point comparisons. Using input parameters that significantly affect the temperature and stress fields as optimization variables allows the inversion algorithm to focus on adjusting key factors, significantly improving the model's prediction accuracy while maintaining computational efficiency. Second, the simulation strain is calculated in real-time and the error function is re-evaluated during iterative updates, allowing the model parameters to gradually approximate the actual heat input state during welding, thus ensuring that the second coupled finite element reference model can more accurately reflect the stress and deformation characteristics of the actual structure. Finally, this method can directly update the calibrated welding heat source parameters into the first finite element model, forming the second finite element reference model. This provides a reliable and accurate numerical basis for subsequent welding residual stress analysis, structural strength assessment, and welding process optimization, while reducing experimental costs and trial-and-error risks, and improving the feasibility and reliability of engineering applications.

[0093] Preferably, based on the second coupled finite element benchmark model, the construction process of marine structures under different welding sequences is simulated, and an AI proxy model for predicting residual stress response under different construction paths is obtained. Specifically, the residual stress distribution results corresponding to each welding sequence are obtained, and a sample dataset between the welding sequence and the residual stress distribution results is constructed. The welding sequence is used as the input feature, and the residual stress distribution parameters are used as the output feature to train the artificial intelligence model, thereby obtaining an AI proxy model for predicting residual stress response under different construction paths. The residual stress distribution parameters include the peak value and mean value of the residual stress distribution results.

[0094] Preferably, the specific steps of the AI ​​proxy model for predicting residual stress response under different construction paths are as follows:

[0095] Based on the topological relationship of the marine structure at the current construction stage and the remaining welding tasks, a variety of different welding sequence combinations are generated from the executable welding procedures. Each welding sequence corresponds to a construction path. The welding sequence is used to characterize the sequential execution relationship of welds or welded nodes and to meet structural stability and construction process constraints.

[0096] The generated welding sequence is used as a time-driven condition input into the second coupled finite element reference model to simulate the heat input and structural stress state corresponding to each welding process in sequence.

[0097] Statistical analysis was performed on the residual stress distribution results corresponding to each welding sequence to extract residual stress distribution parameters that can characterize the overall level of residual stress. The residual stress distribution parameters include: the peak value of the residual stress distribution results, which reflects the maximum residual stress level inside the structure; and the mean value of the residual stress distribution results, which reflects the average distribution state of the overall residual stress of the structure.

[0098] By using each welding sequence as input samples and the corresponding residual stress distribution parameters as output samples, a one-to-one correspondence between welding sequences and residual stress distribution parameters is constructed, forming a sample dataset for training the artificial intelligence model. This dataset is then used to train the AI ​​model, resulting in an AI proxy model for rapid prediction of residual stress. The AI ​​model can be a deep neural network or a graph neural network, used to learn the mapping relationship between welding sequences and residual stress responses. The number of different welding sequences in the sample dataset is no less than a preset minimum sample size to ensure the generalization ability of the model training.

[0099] The trained artificial intelligence model is used as an AI proxy model for residual stress prediction, which is used to quickly predict the residual stress response under different construction paths during the actual construction stage, thereby avoiding repeated finite element simulation analysis with high computational load.

[0100] It should be noted that the steps for inputting each welding sequence as a sample are as follows:

[0101] The welding sequence is characterized and encoded, and then converted into an input feature vector for recognition by an artificial intelligence model. The input feature vector includes the order of welding node numbers, the topological relationship between adjacent welding processes, or the time interval between welding processes.

[0102] This invention significantly improves the efficiency and accuracy of the welding construction process by constructing an AI proxy model to predict residual stress response under different construction paths. First, different welding sequences are used as input samples, and residual stress distribution parameters are extracted. This allows the model to learn the complex nonlinear relationship between welding sequence and residual stress, achieving a rapid mapping from construction path to stress response. This method effectively replaces traditional computationally intensive finite element simulation analysis, significantly reducing computation time while maintaining prediction accuracy. Second, the welding sequence is characterized by encoding information such as node numbering order, topological relationships, and time intervals into input feature vectors recognizable by the AI ​​model. This enables the model to capture the sequentiality, spatial correlation, and time-driven characteristics of welding processes, thus more accurately reflecting the dynamic process of residual stress formation. Furthermore, by statistically analyzing the distribution parameters such as peak and mean residual stress, the model can not only predict local maximum stress but also reflect the overall stress distribution of the structure, providing a quantitative basis for construction scheme optimization and safety assessment. Finally, ensuring the size and diversity of the sample dataset helps the model's generalization ability, enabling the AI ​​agent model to be applied to various construction paths in the actual construction phase. This provides a fast, reliable, and scalable tool for predicting residual stress for engineering decisions, while reducing engineering trial-and-error costs and improving construction safety and quality.

[0103] Preferably, based on the current construction status information of the marine structure, an AI proxy model is invoked to evaluate the remaining optional welding sequences, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results. The specific steps are as follows:

[0104] Based on the prediction results output by the AI ​​agent model, the final residual stress parameters corresponding to each welding sequence are compared with the benchmark residual stress parameters under the current construction state. The residual stress increment corresponding to each welding sequence is calculated. Based on the value of the residual stress increment, all candidate welding sequences are sorted to form an evaluation result set of welding sequence-residual stress increment.

[0105] From the evaluation results set, the welding sequence with the smallest residual stress increment is selected as the target welding sequence for the current construction phase, and this welding sequence is determined as the target construction path. The target construction path is then issued to the construction control equipment to guide the actual execution of subsequent welding operations, thereby achieving proactive stress control and path optimization in the construction process of marine structures.

[0106] The method described in this invention, which evaluates the remaining welding sequences based on an AI proxy model and selects the welding sequence with the smallest residual stress increment as the target construction path, has significant advantages. By comparing the predicted residual stress of each welding sequence with the current baseline residual stress, the impact of each construction path on the structural stress state can be quantified, enabling the prediction of potential stress accumulation during the welding process. This not only improves the scientific nature of construction decisions but also effectively reduces the risk of stress concentration within the structure. Ranking all candidate welding sequences and selecting the path with the smallest residual stress increment achieves optimized selection of the construction path, allowing the structure to maintain a more uniform stress distribution during construction, thereby enhancing the overall safety and durability of the structure. Furthermore, distributing the target construction path to the construction control equipment enables real-time guidance and automated execution of the construction process, closely integrating proactive stress control with path optimization, reducing the uncertainty of human decision-making and the cost of trial and error. This invention not only improves the reliability and efficiency of marine structure construction but also provides operable and quantifiable technical means for the refined management of complex welding processes and construction quality control.

[0107] A digital twin-based construction optimization control system for marine engineering structures includes a structural data acquisition module, a sensitive part screening and anomaly detection module, a welding heat source inversion and correction module, a construction path simulation and AI agent model construction module, and a welding sequence optimization decision module. The structural data acquisition module is connected to the sensitive part screening and anomaly detection module, which is connected to the welding heat source inversion and correction module, which is connected to the construction path simulation and AI agent model construction module, and the construction path simulation and AI agent model construction module is connected to the welding sequence optimization decision module.

[0108] The structural data acquisition module is used to acquire structural parameters and construction process information of large marine structures, deploy strain sensors at key stress-bearing parts of the marine structure to be built to form a stress sensing network to acquire measured strain data, establish a first coupled finite element reference model corresponding to the marine structure to be built on the digital side to acquire simulated strain data, and preset the initial parameters of the welding heat source model.

[0109] The sensitive component screening and anomaly detection module is used to screen sensitive stress components of marine structures based on measured strain data and simulated strain data, and to perform anomaly detection on the sensitive stress components.

[0110] The welding heat source inversion and correction module is used to construct an error function that reflects the difference between the two. With the goal of minimizing the error function, the inversion algorithm is used to perform reverse calibration and correction on the input parameters of the welding heat source model to obtain the second coupled finite element reference model after calibration and correction.

[0111] The construction path simulation and AI proxy model are used to build modules based on the second coupled finite element benchmark model to simulate the construction process of marine structures under different welding sequences and predict the AI ​​proxy model of residual stress response under different construction paths.

[0112] The welding sequence optimization decision module is used to evaluate the remaining optional welding sequences based on the current construction status information of the marine structure, call the AI ​​agent model, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results.

[0113] like Figure 2 The diagram shown is a system block diagram of a digital twin-based optimized control system for marine structure construction according to an embodiment of the present invention, which can be used to execute... Figure 1 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0114] Through the above embodiments, this invention achieves refined control of the welding and construction process of marine structures through a closed-loop mechanism of perception, modeling, and intelligent prediction. By deploying strain sensors at key stress-bearing locations and forming a stress-sensing network, the actual stress state of the structure is acquired in real time, improving the ability to perceive stress changes during construction. An error function is constructed based on the difference between measured and simulated strain, and an inversion algorithm is used to calibrate and correct welding heat source parameters, enabling the second coupled finite element benchmark model to more accurately reflect the actual welding heat input and structural stress distribution, thereby improving the accuracy of simulation prediction. The second benchmark model is then used to generate residual stress response samples under different welding sequences and train an AI proxy model, enabling rapid residual stress prediction of the construction path, avoiding computationally intensive repeated finite element simulations, and saving time and computing resources. By combining the current construction status with the AI ​​proxy model to evaluate the remaining welding sequence, the path with the smallest residual stress increment is selected as the target construction path, achieving proactive stress control and optimization during the construction process, improving construction safety and structural reliability, while reducing trial-and-error costs and construction risks, and realizing intelligent, refined, and efficient construction management.

[0115] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0116] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing and controlling the construction of marine engineering structures based on digital twins, characterized in that, Includes the following steps: By acquiring the structural parameters and construction process information of large marine structures, strain sensors are deployed at the key stress-bearing parts of the marine structure to be built to form a stress sensing network to obtain measured strain data. On the digital side, a first coupled finite element reference model corresponding to the marine structure to be built is established to obtain simulated strain data, and the initial parameters of the welding heat source model are preset. Sensitive stress-bearing components of marine structures were identified based on measured and simulated strain data. Anomaly detection was then performed on these sensitive components. The specific steps are as follows: Based on the measured strain data within a preset sampling period and the simulated strain data output in real time, the measured strain time series is obtained by arranging the measured strain data in chronological order. ,in, This represents the measured strain data corresponding to the i-th sampling point. To determine the number of samples, the simulated strain time series is obtained by arranging the simulated strain data in chronological order. ,in, This represents the simulated strain data corresponding to the i-th sampling point; The first strain fluctuation value was calculated using adjacent sampling points within the measured strain time series. The second strain fluctuation value is obtained by calculating adjacent sampling points within the simulated strain time series. ; Using the fluctuation difference between the first strain fluctuation value and the second strain fluctuation value The sensitive stress-bearing parts of the marine structure were screened and identified. Next, an error function reflecting the difference between the two is constructed. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model, resulting in the second coupled finite element reference model. The specific steps are as follows: Based on the measured strain data output by the stress sensing network during the construction of marine structures in real time, the measured strain data is compared with the simulated strain data of the corresponding virtual measuring points in the finite element benchmark model to obtain the strain difference. An error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model to obtain the second coupled finite element reference model. Based on the second coupled finite element benchmark model, the construction process of marine structures under different welding sequences is simulated, and an AI proxy model for predicting residual stress response under different construction paths is obtained. Specifically, the residual stress distribution results corresponding to each welding sequence are obtained, and a sample dataset between the welding sequence and the residual stress distribution results is constructed. The welding sequence is used as the input feature, and the residual stress distribution parameters are used as the output feature to train the artificial intelligence model, thereby obtaining an AI proxy model for predicting residual stress response under different construction paths. Based on the current construction status information of the offshore structure, an AI proxy model is invoked to evaluate the remaining optional welding sequences, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results. The specific steps are as follows: Based on the prediction results output by the AI ​​agent model, the final residual stress parameters corresponding to each welding sequence are compared with the benchmark residual stress parameters under the current construction state. The residual stress increment corresponding to each welding sequence is calculated. Based on the value of the residual stress increment, all candidate welding sequences are sorted to form an evaluation result set of welding sequence-residual stress increment. From the set of evaluation results, the welding sequence with the smallest residual stress increment is selected as the target welding sequence for the current construction phase, and this welding sequence is determined as the target construction path.

2. The method for optimizing and controlling the construction of marine structures based on digital twins according to claim 1, characterized in that, Key stress-bearing components include K-type pipe nodes and / or T-type pipe nodes where welding stress is concentrated; the initial parameters of the welding heat source model include welding heat source power and heat source distribution radius.

3. The method for optimizing and controlling the construction of marine structures based on digital twins according to claim 1, characterized in that, The welding heat source model is used in numerical simulation to equivalently describe the distribution characteristics of welding heat input in the weld area along the spatial and temporal directions. By converting welding process parameters into thermal loads and inputting them into the first coupled finite element model, the influence of the welding process on the temperature and stress fields of the marine structure is simulated.

4. The method for optimizing and controlling the construction of marine structures based on digital twins according to claim 1, characterized in that, The specific steps for identifying sensitive stress-bearing parts of marine engineering structures are as follows: compare the fluctuation difference with a preset fluctuation threshold. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the corresponding key stress-bearing parts are marked as sensitive stress-bearing parts. At this time, an anomaly is detected in the sensitive stress-bearing parts, triggering an alert. If the fluctuation difference is greater than or equal to the preset fluctuation threshold, the screening process is repeated.

5. The method for optimizing and controlling the construction of marine structures based on digital twins according to claim 1, characterized in that, The strain difference is obtained by comparing the measured strain data with the simulated strain data of the corresponding virtual measuring points in the finite element reference model. The specific steps are as follows: Through a stress sensing network, measured strain data of marine structures are collected in real time according to a preset sampling period. The measured strain data is then processed for time synchronization and outlier removal to form a measured strain time series that progresses with the construction process of the marine structure. In the first coupled finite element reference model, virtual measuring points corresponding one-to-one with the spatial positions of each strain sensor in the stress sensing network are extracted. Based on the current welding process and the initial parameters of the welding heat source model, the simulated strain data of the virtual measuring points are calculated to form a simulated strain time series. By aligning the measured strain time series with the simulated strain time series on the time axis and comparing them, the strain difference at corresponding moments is calculated.

6. The method for optimizing and controlling the construction of marine structures based on digital twins according to claim 1, characterized in that, An error function is constructed based on the strain difference. With the goal of minimizing the error function, an inversion algorithm is used to reverse-calibrate and correct the input parameters of the welding heat source model, resulting in a second coupled finite element reference model. The specific steps are as follows: An error function is constructed based on the strain difference to quantify the overall deviation between measured strain data and simulated strain data. Input parameters that significantly affect the temperature and stress fields are selected from the welding heat source model as calibration and correction parameters, and these parameters are used as optimization variables for the inversion algorithm. With minimizing the error function as the optimization objective, an inversion algorithm is used to iteratively update the parameters to be corrected. In each iteration, the simulation strain data of the finite element model is recalculated based on the updated welding heat source model parameters, and the error function value is recalculated until the error function converges or the preset termination condition is met. When the error function meets the preset convergence condition, the corresponding welding heat source model parameters are output as the calibrated and corrected welding heat source parameters. The calibrated and corrected welding heat source parameters are then updated in the first coupled finite element model to obtain a second finite element reference model that can reflect the actual welding heat input and structural stress state.

7. A digital twin-based optimization control system for marine structure construction, applied to the digital twin-based optimization control method for marine structure construction as described in any one of claims 1-6, characterized in that, It includes a structural data acquisition module, a sensitive part screening and anomaly detection module, a welding heat source inversion and correction module, a construction path simulation and AI agent model construction module, and a welding sequence optimization decision module; The structural data acquisition module is used to acquire structural parameters and construction process information of large marine structures, deploy strain sensors at key stress-bearing parts of the marine structure to be built to form a stress sensing network to acquire measured strain data, establish a first coupled finite element reference model corresponding to the marine structure to be built on the digital side to acquire simulated strain data, and preset the initial parameters of the welding heat source model. The sensitive component screening and anomaly detection module is used to screen sensitive stress components of marine structures based on measured strain data and simulated strain data, and to perform anomaly detection on these sensitive stress components. The welding heat source inversion and correction module is used to construct an error function that reflects the difference between the two. With the goal of minimizing the error function, the inversion algorithm is used to perform reverse calibration and correction on the input parameters of the welding heat source model to obtain the second coupled finite element reference model. The construction path simulation and AI proxy model are used to build the module based on the second coupled finite element benchmark model to simulate the construction process of marine structures under different welding sequences and predict the AI ​​proxy model of residual stress response under different construction paths. The welding sequence optimization decision module is used to evaluate the remaining optional welding sequences based on the current construction status information of the marine structure, call the AI ​​agent model, predict the final residual stress response corresponding to each welding sequence, and determine the welding sequence with the smallest residual stress increment as the target construction path based on the prediction results.

Images