Profile drifting buoy heterogeneous digital twinborn model construction method

Abstract

The invention relates to a profile drifting buoy heterogeneous digital twinborn model construction method, which belongs to the technical field of model construction, and comprises the following steps: constructing a mechanical motion model of a profile drifting buoy, acquiring actual operation data, defining an error function according to the mechanical motion model, and calculating an error value between the mechanical motion model and the actual operation data according to the error function. Calculating a mean value and a standard deviation of the error values through a neural network, and performing error correction on the error values according to the mean value to obtain a heterogeneous digital twinborn model; a model error result is calculated according to the predicted operation state and actual operation data; an error threshold value is defined according to the standard deviation; a loss function is calculated according to the error threshold value and the model error result; and parameters of the heterogeneous digital twin model are updated according to the loss function. A deviation quantification mechanism between a physical model and a real system can be established, and high-precision modeling and prediction of the running state of the profile drifting buoy are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and in particular to a method for constructing a heterogeneous digital twin model of a profile drifting buoy. Background Art

[0002] In the field of ocean observation, profile drifting buoys are in a complex and changeable deep-sea environment for a long time. Their motion state, structural health and environmental adaptability directly affect the accuracy of data collection. Traditional monitoring methods rely on limited sensor data and it is difficult to fully evaluate the real-time status of the buoy. Digital twin technology can dynamically simulate the mechanical behavior of the buoy, predict potential failures, and provide visual support for maintenance decisions by constructing a high-fidelity virtual model, thereby significantly improving the reliability and efficiency of ocean observations.

[0003] At present, the construction of heterogeneous digital twin models is mainly based on the establishment of analytical models based on the mechanical equations of the buoy. It has strong interpretability and robustness, is suitable for scenarios with scarce data, or uses deep learning to learn system behavior from historical data, adapts to nonlinear and coupling relationships, and has high modeling efficiency.

[0004] However, the physical modeling method relies on precise knowledge of the mechanism, but the complex working conditions of the buoy in the deep-sea environment cause the model parameters to drift easily and are difficult to update in real time, resulting in insufficient adaptability; the pure data-driven method lacks physical interpretability, has extremely high requirements on the amount and quality of training data, has poor generalization ability when data is insufficient, and has high computational complexity, making it difficult to deploy on lightweight edge devices of buoys. Summary of the Invention

[0005] In response to the shortcomings in the relevant technologies, the purpose of the present invention is to provide a method for constructing a heterogeneous digital twin model of a profile drifting buoy, so as to solve the technical problems that the physical modeling method in the existing technology is difficult to adapt to complex working conditions due to its reliance on precise mechanisms, and the data-driven method is difficult to deploy in lightweight equipment due to its poor interpretability, strong data dependence and high computational complexity.

[0006] The present invention provides a method for constructing a heterogeneous digital twin model of a profile drifting buoy, comprising the following steps: Error calculation step: constructing a mechanical motion model of the profile drifting buoy, obtaining actual operation data of the profile drifting buoy based on the mechanical motion model, defining an error function according to the mechanical motion model, and calculating an error value between the mechanical motion model and the actual operation data according to the error function; Model construction step: calculating the mean and standard deviation of the error value through a neural network, performing error correction on the error value according to the mean, and obtaining a heterogeneous digital twin model according to the error-corrected mechanical motion model; Parameter updating step: obtaining the predicted operating state of the profile drifting buoy according to the heterogeneous digital twin model, calculating the model error result according to the predicted operating state and the actual operating data, defining the error threshold according to the standard deviation, calculating the loss function according to the error threshold and the model error result, and updating the parameters of the heterogeneous digital twin model according to the loss function. The heterogeneous digital twin model is used to generate decisions based on the profile drifting buoy.

[0007] By constructing a mechanical motion model and defining an error function to calculate the error value with the actual operating data, a deviation quantification mechanism between the physical model and the real system can be established, providing an accurate data basis for subsequent error correction; then, the mean and standard deviation of the error are calculated through a neural network and the error is corrected, compensating for the complex nonlinear errors that are difficult to accurately describe in the physical model, and finally forming a heterogeneous digital twin model that integrates physical mechanisms and data-driven, realizing high-precision modeling and prediction of the operating status of the profile drifting buoy.

[0008] In some embodiments of the present invention, the parameter updating step specifically includes: The prediction error between the error value and the mean is obtained through the neural network, the maximum mean difference between the prediction error and the error value is calculated according to the loss function, and the parameters of the heterogeneous digital twin model are updated according to the maximum mean difference.

[0009] The prediction error is calculated through a neural network and combined with the maximum mean difference to update the model parameters, which quantifies the distribution difference between the prediction error and the actual error. By forcing the distribution consistency to be maintained, performance degradation during the model update process is avoided. The parameter optimization process not only considers the numerical size of the error but also pays more attention to the statistical characteristics of the error, thereby improving the scientific nature of the parameter update and the stability of the model update, and ensuring the reliability of the heterogeneous digital twin model in long-term operation.

[0010] In some embodiments of the present invention, the method further comprises: Neural network updating step: If there is a preset deviation between the parameters of the updated heterogeneous digital twin model and the parameters of the mechanical motion model, the parameters of the mechanical motion model are updated, the error value and the mean and standard deviation of the error value are recalculated, and the neural network is updated.

[0011] By triggering the retraining process when a preset deviation is detected between the model parameters and the physical model parameters, and tracking the dynamic changes of the system by periodically refreshing the entire modeling process, the model drift problem caused by environmental changes or equipment aging is solved, thereby achieving the full life cycle adaptability of heterogeneous digital twin models, extending the effective use period of the model, and ensuring the consistency of simulation accuracy with the physical entity.

[0012] In some embodiments of the present invention, the calculation model of the loss function is:

[0013] in, is the loss function; is the model error result; is the error threshold.

[0014] By defining the calculation model of the loss function, the loss calculation process is standardized, ensuring that the optimization behavior under large and small errors meets the design expectations, providing a clear and differentiable optimization target for parameter updating, thereby ensuring the convergence and stability of the entire adaptive system.

[0015] In some embodiments of the present invention, the dynamics of the mechanical motion model is expressed as:

[0016] in, is the mass of the profile drifting buoy; For depth the density of the water below; is the volume of the glass shell; is the compressibility coefficient of the glass shell; is the volume of the outer oil sac; is the compression coefficient of hydraulic oil; is an unmodelable item; is the compression coefficient corresponding to the unmodelable item; is the drag coefficient of the profile drifting buoy; is the movement speed of the profile drifting buoy.

[0017] By defining the dynamic equations of the mechanical motion model, an accurate physical description of the buoy motion was established based on the principles of fluid mechanics and material mechanics, providing a reliable theoretical basis for the heterogeneous digital twin model.

[0018] In some embodiments of the present invention, the error function is expressed as:

[0019] in, is the error function.

[0020] By defining the mathematical expression of the error function, the difference between the mechanical motion model and the actual observation value is formalized into a computable functional relationship, which provides a clear learning goal for the neural network. At the same time, the physical interpretability of each parameter is retained, so that error analysis can be traced back to specific physical factors, providing directional guidance for system optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work. Figure 1 A flowchart of a method for constructing a heterogeneous digital twin model of a profile drifting buoy provided in an embodiment of the present invention; Figure 2 A flowchart of another method for constructing a heterogeneous digital twin model of a profile drifting buoy provided in an embodiment of the present invention; Figure 3 A flowchart of another method for constructing a heterogeneous digital twin model of a profile drifting buoy provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is described and illustrated below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application. Based on the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts are within the scope of protection of this application. It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. Digital Twin technology is a modeling and simulation method that integrates the virtual and real worlds. By constructing a virtual model consistent with the physical entity, it achieves real-time perception, predictive analysis, health management and behavioral decision-making of equipment operating status. It has good visualization, predictability and explainability, and provides new solutions for condition monitoring, fault diagnosis and intelligent decision-making of marine equipment.

[0023] By constructing a virtual model that is highly consistent with the operating status of physical equipment, digital twins can not only achieve real-time perception, predictive analysis, and health assessment of the system, but also have good visualization, explainability, and scalability, providing new solutions for status monitoring, fault diagnosis, and autonomous decision-making of complex systems such as marine equipment.

[0024] Currently, digital twin construction methods mainly fall into two categories: physical modeling-based methods and artificial intelligence-based methods. Physical modeling-based methods rely on prior knowledge of the equipment and mechanical equations, and can provide high interpretability and strong robustness. They are suitable for scenarios with scarce data and complex environments. However, the modeling process is cumbersome and difficult to adapt to changes in system operating conditions. Artificial intelligence-based methods automatically learn system behavior through data-driven methods such as deep learning. They have high modeling efficiency and strong adaptability, especially in dealing with nonlinear and coupled systems. However, they also have problems such as poor interpretability, high dependence on data, high computational complexity, and difficulty in deployment on lightweight equipment.

[0025] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other. The technical solution of the present invention is described in detail below in conjunction with specific embodiments and the accompanying drawings.

[0026] like Figure 1 As shown, the present invention provides a method for constructing a heterogeneous digital twin model of a profile drifting buoy, comprising: Error calculation step S1: constructing a mechanical motion model of the profile drifting buoy, obtaining actual operation data of the profile drifting buoy based on the mechanical motion model, defining an error function according to the mechanical motion model, and calculating the error value between the mechanical motion model and the actual operation data according to the error function; In some embodiments, the dynamics of the mechanical motion model is represented as:

[0027] in, is the mass of the profile drifting buoy; For depth the density of the water below; is the volume of the glass shell; is the compressibility coefficient of the glass shell; is the volume of the outer oil sac; is the compression coefficient of hydraulic oil; is an unmodelable item; is the compression coefficient corresponding to the unmodelable item; is the drag coefficient of the profile drifting buoy; is the velocity of the profile drifting buoy; Specifically, is the volume other than the glass shell volume and the outer oil bladder volume. The other volume is the unmeasurable volume in the process of balancing the mass and buoyancy of the profile drifting buoy in the laboratory, that is, the unmodelable item.

[0028] By defining the dynamic equations of the mechanical motion model, an accurate physical description of the buoy motion was established based on the principles of fluid mechanics and material mechanics, providing a reliable theoretical basis for the heterogeneous digital twin model.

[0029] In some embodiments, historical operating data of the profile drifting buoy is obtained, and the parameters of the mechanical motion model are updated by the least squares method (OLS) according to the historical operating data and preset laboratory parameters to obtain an updated mechanical motion model.

[0030] In some embodiments, the error function is used to represent changes in volume error and compressibility of different materials at different depths; The expression of the error function is:

[0031] in, is the error function; although the error function is composed of linear terms of error, considering the existence of unmodelable terms and neglected terms in the mechanical motion model and different materials, the error function is expressed as express.

[0032] By defining the mathematical expression of the error function, the difference between the mechanical motion model and the actual observation value is formalized into a computable functional relationship, which provides a clear learning goal for the neural network. At the same time, the physical interpretability of each parameter is retained, so that error analysis can be traced back to specific physical factors, providing directional guidance for system optimization.

[0033] Error calculation step S2: Calculate the mean and standard deviation of the error value through the neural network, perform error correction on the error value according to the mean, and obtain the heterogeneous digital twin model based on the error-corrected mechanical motion model; Optionally, the neural network is a Monte Carlo Dropout neural network. Compared with traditional methods that require complex Bayesian frameworks or integrated learning, the Monte Carlo Dropout technology does not need to increase the number of models or significantly change the network structure. Multiple sampling can be performed by activating Dropout only in the test phase. The computational overhead is small, which is very suitable for deployment in edge devices with limited computing resources such as buoys. The Dropout technology randomly discards neurons during the training process, so that the final model does not overly rely on any specific neurons or features, enhances the robustness to noise and outliers, and improves the generalization ability on unseen data.

[0034] Parameter updating step S3: obtain the predicted operating status of the profile drifting buoy according to the heterogeneous digital twin model, calculate the model error result according to the predicted operating status and the actual operating data, define the error threshold according to the standard deviation, calculate the loss function according to the error threshold and the model error result, and update the parameters of the heterogeneous digital twin model according to the loss function.

[0035] Specifically, the profile drifting buoy may face the problem of parameter drift in actual use, that is, the parameters of the updated heterogeneous digital twin model and the parameters of the mechanical motion model have preset deviations. Therefore, in order to always ensure the prediction accuracy of the heterogeneous digital twin model, when the parameters of the heterogeneous data twin model drift, the robust least squares method RLS is used for parameter identification; However, the defect of the robust least squares method (RLS) is that the threshold setting of outliers is fixed. Therefore, the error threshold is defined according to the standard deviation through the 3-sigma criterion, and the error threshold is used as the threshold of the Huber loss function in the robust least squares method to achieve threshold adaptation in the parameter identification process.

[0036] By dynamically defining the error threshold and calculating the Huber loss function through the 3-sigma criterion, adaptive processing of abnormal data is achieved. When the error is within a reasonable range, the square loss is used to ensure the accuracy of parameter update. When an abnormally large error occurs, it switches to linear loss to suppress outlier interference. In this way, the parameter update process maintains the optimization characteristics of the least squares method and has the robustness against data noise.

[0037] In some embodiments, the calculation model of the Huber loss function is:

[0038] in, is the Huber loss function; is the model error result; is the error threshold.

[0039] Specifically, the Huber loss function is a piecewise function. When the absolute value of the model error result is greater than the error threshold, a linear loss calculation is used to ensure robustness. When the error is extremely large, the loss value grows linearly with the error instead of quadratically, reducing the weight of large errors and preventing them from excessively affecting parameter updates. When the absolute value of the model error result is less than or equal to the error threshold, the square loss calculation is used to ensure accuracy, so that when the error is small, its behavior is consistent with the ordinary least squares method, which can provide accurate and fast convergence.

[0040] By defining the calculation model of Huber loss, the loss calculation process is standardized, ensuring that the optimization behavior under large and small errors meets the design expectations, providing a clear and differentiable optimization target for parameter updating, thereby ensuring the convergence and stability of the entire adaptive system.

[0041] In some embodiments, the parameter updating step S3 specifically includes: The prediction error between the error value and the mean is obtained through a neural network, the maximum mean difference between the prediction error and the error value is calculated according to the loss function, and the parameters of the heterogeneous digital twin model are updated according to the maximum mean difference. The heterogeneous digital twin model is used to generate decisions based on the profile drift buoy.

[0042] The prediction error is calculated through a neural network and combined with the maximum mean difference to update the model parameters, which quantifies the distribution difference between the prediction error and the actual error. By forcing the distribution consistency to be maintained, performance degradation during the model update process is avoided. The parameter optimization process not only considers the numerical size of the error but also pays more attention to the statistical characteristics of the error, thereby improving the scientific nature of the parameter update and the stability of the model update, and ensuring the reliability of the heterogeneous digital twin model in long-term operation.

[0043] Specifically, although the influence of the error value is taken into account when updating the parameters, the adaptability of the error value output by the updated mechanical motion model and the neural network still needs to be considered. That is, although the updated mechanical motion model has a smaller error, due to the inconsistent error distribution, the output of the mechanical motion model and the neural network will have a larger error.

[0044] Therefore, in order to avoid this situation, the maximum mean difference is introduced to measure the distribution deviation between the error value between the mechanical motion model and the actual operation data and the prediction error output by the neural network, and the error value and the prediction error are forced to have distribution consistency. At the same time, the maximum mean difference between the errors is used as the loss term for parameter update to constrain the parameter update process.

[0045] Specifically, if Figure 2 As shown in the figure, the process of building a heterogeneous digital twin model is as follows: Construct a mechanical motion model of the profile drifting buoy and obtain the actual operation data of the profile drifting buoy based on the mechanical motion model; An error function is defined according to the mechanical motion model, and an error value between the mechanical motion model and actual operation data is calculated according to the error function; The mean and standard deviation of the error values ​​are calculated through a neural network, the error values ​​are corrected according to the mean, and a heterogeneous digital twin model is obtained based on the error-corrected mechanical motion model. The predicted operating status of the profile drifting buoy is obtained based on the heterogeneous digital twin model, and the model error result is calculated based on the predicted operating status and actual operating data; The error threshold is defined by the 3-sigma criterion and the standard deviation, and the model error result is judged whether it exceeds the threshold and the Huber loss function is calculated; The prediction error between the error value and the mean is obtained through the neural network, and the maximum mean difference (MMD) between the prediction error and the error value is calculated by combining the Huber loss function; The Huber loss function and the maximum mean difference (MMD) are combined as the optimization objective, and the data noise constraint and error distribution constraint are implemented through the robust recursive least squares method (RRLS), so as to update the parameters of the heterogeneous digital twin model.

[0046] Specifically, the Huber loss function is used to dynamically weight the error value to suppress the influence of outliers, and the maximum mean difference (MMD) is used to enforce the distribution consistency of the error value and the prediction error.

[0047] Based on the above-mentioned heterogeneous digital twin model construction method, by constructing a mechanical motion model and defining an error function to calculate the error value with the actual operating data, a deviation quantification mechanism between the physical model and the real system can be established, providing an accurate data basis for subsequent error correction; then, the mean and standard deviation of the error are calculated through a neural network and the error is corrected, compensating for the complex nonlinear errors that are difficult to accurately describe in the physical model, and finally forming a heterogeneous digital twin model that integrates physical mechanisms and data-driven, realizing high-precision modeling and prediction of the operating status of the profile drifting buoy.

[0048] In some embodiments, as Figure 3 As shown, the heterogeneous digital twin model construction method also includes: Neural network updating step S4: If there is a preset deviation between the parameters of the updated heterogeneous digital twin model and the parameters of the mechanical motion model, the parameters of the mechanical motion model are updated, the error value and the mean and standard deviation of the error value are recalculated, and the neural network is updated.

[0049] Specifically, if the parameters of the updated heterogeneous digital twin model after error correction by the Monte Carlo Dropout neural network still have a preset deviation from the parameters of the mechanical motion model, the Monte Carlo Dropout neural network needs to be updated online. By repeatedly updating the parameters of the mechanical motion model, the error value between the mechanical motion model and the actual operation data, as well as the mean and standard deviation of the error value, is recalculated, and the Monte Carlo Dropout neural network is updated to fine-tune the parameters to adapt to the new state. Freeze the feature extraction layer of the neural network according to fixed parameters and fine-tune the output layer to prevent overfitting and reduce the amount of computation; Define the loss function of the neural network; Enable Dropout for training. During fine-tuning, keep the Dropout layer turned on. Each forward propagation will still randomly drop neurons, so that the training process can continue to maintain the model's uncertainty estimation ability.

[0050] Use the optimization algorithm to iterate on the new dataset for a small number of epochs, adjusting the parameters of the unfrozen layers.

[0051] Evaluate the performance of the fine-tuned network on a validation set that was not used in training; If the performance meets the requirements, the fine-tuned neural network parameters are used to replace the original parameters to complete the online update; if the performance does not meet the requirements, the hyperparameters need to be adjusted and fine-tuned again, or a more complex update process needs to be triggered.

[0052] By triggering the retraining process when a preset deviation is detected between the model parameters and the physical model parameters, and tracking the dynamic changes of the system by periodically refreshing the entire modeling process, the model drift problem caused by environmental changes or equipment aging is solved, thereby achieving the full life cycle adaptability of heterogeneous digital twin models, extending the effective use period of the model, and ensuring the consistency of simulation accuracy with the physical entity.

[0053] It should be noted that the above is a reference method for constructing a heterogeneous digital twin model of a profile drifting buoy, and the present invention is not limited to this.

[0054] The embodiments of the present invention achieve the complementary advantages of physical mechanisms and data-driven, so that the heterogeneous digital twin model not only retains the interpretability and robustness of the physical model, but also compensates for the complex nonlinear errors that are difficult to accurately describe the physical model through neural networks, ultimately achieving the purpose of high-precision modeling. At the same time, the parameter update mechanism ensures that the heterogeneous digital twin model can adapt to the complex and changeable marine environment, providing a more reliable virtual simulation platform for buoy status monitoring, and solving the technical problems that the physical modeling method in the existing technology relies on precise mechanism knowledge, but the complex working conditions of the buoy in the deep-sea environment cause the model parameters to drift easily, and are difficult to update in real time, and the adaptability is insufficient; the pure data-driven method lacks physical interpretability, has extremely high requirements on the amount and quality of training data, has poor generalization ability when data is insufficient, and has high computational complexity, making it difficult to deploy on lightweight edge devices of buoys.

[0055] Finally, it should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other. The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the same. Although the present invention has been described in detail with reference to preferred embodiments, persons skilled in the art should understand that the specific implementation methods of the present invention may still be modified or some technical features may be replaced by equivalents without departing from the spirit of the technical solutions of the present invention, and all of these should fall within the scope of the technical solutions claimed for protection by the present invention.

Claims

1. A method for constructing a heterogeneous digital twin model of a profile drifting buoy, characterized in that: The steps include: Error calculation step: constructing a mechanical motion model of the profile drifting buoy, obtaining actual operation data of the profile drifting buoy based on the mechanical motion model, defining an error function according to the mechanical motion model, and calculating an error value between the mechanical motion model and the actual operation data according to the error function; Model construction step: calculating the mean and standard deviation of the error value through a neural network, performing error correction on the error value according to the mean, and obtaining a heterogeneous digital twin model according to the error-corrected mechanical motion model; Parameter updating step: obtaining the predicted operating state of the profile drifting buoy according to the heterogeneous digital twin model, calculating the model error result according to the predicted operating state and the actual operating data, defining the error threshold according to the standard deviation, calculating the loss function according to the error threshold and the model error result, and updating the parameters of the heterogeneous digital twin model according to the loss function. The heterogeneous digital twin model is used to generate decisions based on the profile drifting buoy.

2. The method for constructing a heterogeneous digital twin model of a profile drifting buoy according to claim 1, characterized in that: The parameter updating step specifically includes: The prediction error between the error value and the mean is obtained through the neural network, the maximum mean difference between the prediction error and the error value is calculated according to the loss function, and the parameters of the heterogeneous digital twin model are updated according to the maximum mean difference.

3. The method for constructing a heterogeneous digital twin model of a profile drifting buoy according to claim 1, characterized in that: The method further comprises: Neural network updating step: If there is a preset deviation between the parameters of the updated heterogeneous digital twin model and the parameters of the mechanical motion model, the parameters of the mechanical motion model are updated, the error value and the mean and standard deviation of the error value are recalculated, and the neural network is updated.

4. The method for constructing a heterogeneous digital twin model of a profile drifting buoy according to claim 1, characterized in that: The calculation model of the loss function is: in, is the loss function; is the model error result; is the error threshold.

5. The method for constructing a heterogeneous digital twin model of a profile drifting buoy according to claim 1, characterized in that: The dynamics of the mechanical motion model is expressed as: in, is the mass of the profile drifting buoy; For depth the density of the water below; is the volume of the glass shell; is the compressibility coefficient of the glass shell; is the volume of the outer oil sac; is the compression coefficient of hydraulic oil; is an unmodelable item; is the compression coefficient corresponding to the unmodelable item; is the drag coefficient of the profile drifting buoy; is the movement speed of the profile drifting buoy.

6. The method for constructing a heterogeneous digital twin model of a profile drifting buoy according to claim 5, characterized in that: The expression of the error function is: in, is the error function.

Images