A physical consistent wind blade deformation field measurement method, system and device based on PINN and DIC fusion and a storage medium
Patent Information
- Application Number
- CN202611105129.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-24
- Publication Date
- 2026-10-13
AI Technical Summary
该方法在处理叶片大旋转、大位移时面临根本性挑战：大刚体运动易导致匹配失效与误差累积；其位移-应变分步求解流程会放大图像噪声；更重要的是，该方法完全忽略了力学本构关系，导致测量结果存在物理不一致性——即所获得的位移场与应变场之间不满足变形协调方程和平衡方程
[0035]基于力学普适定律引导的物理信息嵌入机制：将圣维南协调方程、静力平衡方程嵌入损失函数，强制位移、应变场满足力学基本规律，解决传统DIC忽略力学关系、测量结果物理不一致的缺陷；通过联合损失校正DIC初始解，抑制图像噪声带来的误差累积。
Smart Images
Figure CN122899666A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of optical structure monitoring technology, and in particular to a method, system, device and storage medium for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC. Background Technology
[0002] As the core component for energy capture, wind turbine blades operate under complex aerodynamic loads for extended periods. Achieving physical consistency in the measurement of their surface displacement and strain field is crucial for structural safety monitoring.
[0003] Traditional digital image correlation (DIC) technology treats shape field measurement as a purely geometric matching problem. It calculates the displacement field by comparing the gray-level correlation of local sub-regions between a reference image and a deformed image, and then calculates the strain field through numerical differentiation. This method faces fundamental challenges when dealing with large blade rotations and displacements: large rigid body motions easily lead to matching failures and error accumulation; its step-by-step displacement-strain solution process amplifies image noise; more importantly, this method completely ignores mechanical constitutive relations, resulting in physical inconsistencies in the measurement results—that is, the obtained displacement field and strain field do not satisfy the deformation compatibility equation and equilibrium equation.
[0004] While existing deep learning methods improve robustness, their "black box" nature can lead to predictions that violate physical laws in data-scarce regions, making it difficult to establish the necessary measurement confidence. Furthermore, the dynamic changes in blade operating conditions (wind speed, pitch angle, rotational speed, etc.) mean that fixed physical constraints cannot adapt to the mechanical response characteristics under different operating conditions. Therefore, a new paradigm for deformation field measurement that unifies "geometric accuracy" and "physical consistency" is urgently needed. Summary of the Invention
[0005] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a method, system, device and storage medium for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0007] This invention provides a method for measuring the deformation field of physically consistent wind turbine blades based on the fusion of PINN and DIC, comprising the following steps:
[0008] S1: Input the initial displacement field and current operating parameters of the wind turbine blade DIC;
[0009] S2: Based on the input DIC initial displacement field and working parameters, a physical information embedding mechanism guided by the universal laws of mechanics is constructed. The deformation compatibility equation is embedded into the network output structure as a strong constraint, and the equilibrium law residual is introduced into the loss function as a regularization term to perform physical correction on the DIC initial displacement field.
[0010] S3: Construct a dual-branch coupled network fusion architecture for adaptive working condition perception. The observation branch extracts the deformation features of the initial displacement field of DIC, and the physical perception branch dynamically generates physical constraint parameters that adapt to the current working condition with the working condition parameters as input. The observed deformation features and the prior physical features of the working condition are weighted and fused through the gated fusion module to obtain the fused features.
[0011] S4: Based on the fusion features, a physical credibility quantification and fusion verification system is constructed. The full-field physical residual norm is calculated to generate a physical credibility evaluation map. Sparse high-fidelity sensor data is used as a hard constraint and integrated into the network training.
[0012] S5: Outputs physically consistent wind turbine blade displacement field, strain field, and corresponding physical reliability assessment graph.
[0013] Furthermore, step S2, which constructs a physical information embedding mechanism guided by universal laws of mechanics, specifically includes:
[0014] The strain field is calculated based on the displacement field predicted by the network. The strain field is then substituted into the Saint-Venant compatibility equation to solve for the compatibility equation residuals. The stress field is calculated from the strain field based on the linear elastic constitutive relation to solve for the internal equilibrium equation residuals. The physical information loss function is constructed by combining the compatibility equation residuals and the equilibrium equation residuals. The total loss is obtained by combining the DIC matching data loss to complete the physical correction.
[0015] Furthermore, step S3, constructing the network fusion architecture for adaptive working condition perception, specifically includes:
[0016] The observation branch adopts an encoder-decoder structure, inputting full-field displacement data from the DIC to extract high-dimensional deformation features; the physical sensing branch is a parameterized PINN, taking working condition parameters as input, and dynamically generating adaptive physical constraint parameters through the working condition coding network; the gating weights are calculated to adaptively fuse deformation features and physical prior features, and the fused features are sent to the decoder to output the predicted displacement field and strain field.
[0017] Furthermore, step S4, constructing a physical trust quantification and fusion verification system, specifically includes:
[0018] The physical residual norm of the entire field and the local residual of a single point in space are calculated. The local residuals are normalized and visualized to generate a physical credibility assessment map. A high-fidelity sensor data loss term is added, and the data loss, physical loss and sensor loss are combined as the total loss for network training.
[0019] The present invention also provides a physically consistent wind blade deformation field measurement system based on the fusion of PINN and DIC, for performing any of the measurement methods described above, including:
[0020] The physical information embedding module is used to embed the deformation compatibility equation as a strong constraint into the network output and to introduce the equilibrium law residual as a regularization term into the loss function.
[0021] The dual-branch coupled network module includes an observation branch and a physical sensing branch, which are used to realize adaptive sensing of working conditions and feature fusion.
[0022] The credibility quantification and evaluation module is used to generate a physical credibility evaluation map and perform multi-fidelity data fusion verification.
[0023] The deformation field output module is used to output physically consistent displacement and strain fields.
[0024] Furthermore, the dual-branch coupled network module includes:
[0025] The observation branch unit adopts an encoder-decoder structure, inputting the DIC displacement field and outputting deformation characteristics;
[0026] The physical sensing branch unit includes a working condition coding network and a parameterized PINN, which takes working condition parameters as input and outputs adaptive physical constraint parameters.
[0027] The gating fusion unit is used to calculate gating weights and fuse observed features with physical prior features.
[0028] Furthermore, the trust quantification and evaluation module includes:
[0029] The residual calculation unit is used to calculate the residuals of the compatibility equations and the equilibrium equations.
[0030] Normalized mapping units are used to generate physical credibility assessment maps;
[0031] Sensor hard constraint units are used to incorporate sparse high-fidelity sensor data into the network loss function.
[0032] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements the physical consistency wind blade deformation field measurement method based on the fusion of PINN and DIC described above.
[0033] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the physical consistency wind blade deformation field measurement method based on the fusion of PINN and DIC as described above.
[0034] Compared with the prior art, the technical solution disclosed in this invention has the following beneficial effects:
[0035] The physical information embedding mechanism guided by the universal laws of mechanics: The Saint-Venant coordination equation and the static equilibrium equation are embedded into the loss function to force the displacement and strain fields to satisfy the basic laws of mechanics, thus solving the defects of traditional DIC that ignore mechanical relationships and the physical inconsistency of measurement results; the initial solution of DIC is corrected by joint loss to suppress the error accumulation caused by image noise.
[0036] The dual-branch coupled network fusion architecture based on adaptive operating condition perception: parameterized PINN dynamically adjusts physical constraint parameters according to real-time operating conditions, and the gated fusion module adaptively balances image observation information and physical priors. This solves the problems of existing deep learning models being unable to adapt to dynamic blade operating conditions due to fixed physical constraints and black-box predictions violating physical laws, thereby improving the robustness of measurements under varying operating conditions.
[0037] Based on the verification system of physical reliability quantification and multi-fidelity data fusion: a reliability map is generated by physical residuals to intuitively identify the measurement area error and confidence level, making up for the lack of quantitative evaluation methods in deep learning; high-fidelity sparse sensing data of strain gauges is fused as hard constraints to further improve the accuracy of full-field deformation reconstruction and realize the unification of multi-source information from geometric observation, mechanical priors and physical sensing.
[0038] The overall solution combines the geometric measurement accuracy of DIC with the physical constraints of PINN. No manual division of fixed constraints is required throughout the process. The measurement results have both geometric accuracy and mechanical consistency, making it suitable for engineering monitoring scenarios of wind turbine blades with large rotation and large displacement. Attached Figure Description
[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0040] Figure 1 This is a flowchart of the method in the first embodiment of the present invention;
[0041] Figure 2-1 This is a general technical roadmap of the present invention;
[0042] Figure 2-2 A schematic diagram of a physical information embedding mechanism guided by universal laws of mechanics;
[0043] Figure 2-3 A schematic diagram of a network fusion architecture for adaptive working condition perception.
[0044] Figure 2-4 A schematic diagram illustrating the construction of a physical credibility quantification and fusion verification system;
[0045] Figure 3 This is a system structure block diagram according to the sixth embodiment of the present invention. Detailed Implementation
[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0048] Example 1: Basic Method
[0049] Please see Figure 1 The diagram shows a flowchart of the method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC in the first embodiment of the present invention.
[0050] The overall technical route of the present invention is as follows: Figure 2-1 As shown, its core idea is: taking the initial displacement field and working condition parameters of DIC as input, it sequentially passes through a physical information embedding mechanism, a working condition adaptive dual-branch coupled network fusion, and physical credibility quantification and multi-fidelity verification, finally outputting a physically consistent displacement field, strain field, and credibility evaluation spectrum. The method includes steps S1 to S5:
[0051] S1, obtain the initial displacement field and working parameters of DIC.
[0052] A sequence of speckle images of the wind turbine blade surface before and after deformation was collected, and the total displacement field was initially calculated using the traditional DIC method. Get the current operating condition parameter vector at the time , representing wind speed, propeller pitch angle, rotational speed, and turbulence intensity, respectively.
[0053] S2, constructing a physical information embedding mechanism guided by universal laws of mechanics.
[0054] Step S2 includes steps S21 to S23:
[0055] S21, Strong constraint construction of deformation compatibility equations.
[0056] For the displacement fields u(x,y) and v(x,y) predicted by the neural network, the strain field is calculated according to the Cauchy-Green strain relation:
[0057]
[0058] Substituting the strain field into the Saint-Venant compatibility equation based on the small deformation assumption:
[0059]
[0060] Calculate the residual R of this equation. compat :
[0061]
[0062] in , These are the normal strain components in the x and y directions, respectively. For shear strain components.
[0063] This constraint ensures that the displacement and strain fields output by the network are geometrically self-consistent, fundamentally avoiding non-physical deformation modes.
[0064] S22, the regularization constraint of the equilibrium law is introduced.
[0065] Based on the linear elastic constitutive relation of the blade material, the stress field is calculated from the predicted strain field:
[0066]
[0067] Where σ is the stress field, C is the material stiffness matrix, and ε is the material elastic modulus.
[0068] Neglecting body forces, calculate the residuals of the internal equilibrium equations:
[0069]
[0070] S23, Construct the physical information loss function.
[0071] The residuals of the coordination equations and the residuals of the equilibrium equations together constitute the physical information loss:
[0072]
[0073] By jointly optimizing data loss L data (DIC matching loss, i.e., the mean square error between the predicted displacement and the initial DIC displacement) and physical loss, the total loss is This enables physical correction of the initial solution of DIC.
[0074] The physical information embedding mechanism guided by the aforementioned universal laws of mechanics, such as Figure 2-2As shown, its core lies in the following: the network predicts the displacement field and derives the strain field through the Cauchy-Green relation. The strain field is substituted into the Saint-Venant compatibility equation to calculate the compatibility residual. On the other hand, the stress field is obtained through the linear elastic constitutive relation and then substituted into the equilibrium equation to calculate the equilibrium residual. The two residuals together constitute the physical information loss, and the network is physically corrected by combining the DIC data loss.
[0075] S3 is a network fusion architecture that adapts to and perceives operating conditions.
[0076] Step S3 includes steps S31 to S32:
[0077] S31, Design of a dual-branch coupled network structure.
[0078] The observation branch employs an encoder-decoder structure, using the full-field displacement data U_DIC calculated by DIC as input to extract high-dimensional deformation features F. obs The encoder consists of four downsampled convolutional blocks, with the number of channels increasing from 64 to 128 to 256 to 512.
[0079] The physical sensing branch, acting as a parameterized PINN, takes the current operating condition parameters p (wind speed, pitch angle, and rotational speed) as input and dynamically generates physical constraint parameters θ adapted to the current operating condition through an operating condition coding network consisting of three fully connected layers. phy (including equivalent material parameter E) eff and loss weight λ phy This enables the linkage between the physical model and the operating conditions.
[0080] S32, implemented by the gating feature fusion module.
[0081] Design a gated fusion module to observe the deformation features F extracted from the branches. obs The working condition sensing parameter θ output by the physical sensing branch phy As input, the data is adaptively weighted and fused using a gating mechanism. The gating weights are calculated as follows:
[0082]
[0083] Where σ is the Sigmoid activation function, W g This is a learnable weight matrix. The fused features are:
[0084]
[0085] Where F phy The physical prior features are output by the physical perception branch. The fused features are fed into the decoder, ultimately outputting a physically consistent displacement field U. pred and strain field ε predThis mechanism allows the network to rely on observation data in areas with high signal-to-noise ratios, while in areas with sparse data or high noise, it tends to follow the guidance of physical laws.
[0086] The above-mentioned adaptive sensing network fusion architecture is as follows: Figure 2-3 As shown in the figure, the overall structure of the dual-branch coupled network is illustrated: the observation branch takes the full-field displacement of the DIC as input and extracts high-dimensional deformation features through the encoder; the physical perception branch takes the working condition parameter p as input and dynamically generates adaptive physical constraint parameters through the working condition coding network, thereby extracting physical prior features; the two features are adaptively weighted by the gated fusion module and then sent to the decoder, finally outputting a physically consistent displacement field and strain field.
[0087] S4, construct a physical credibility quantification and fusion verification system.
[0088] Step S4 includes steps S41 to S42:
[0089] S41, Generation of physical credibility assessment map.
[0090] Define the norm of the physical residuals across the entire field:
[0091]
[0092] Calculate the local residual for each spatial location point (x, y):
[0093]
[0094] in For the total physical residual, To adapt to and coordinate residuals, This represents the residuals of the equilibrium equation.
[0095] By normalizing the local residuals By mapping the data with visualizations (such as heatmaps), a "physical reliability assessment map" is generated that corresponds to the measurement results. This map can visually identify which areas in the measurement results highly conform to physical laws (low residuals) and which areas may have uncertainties or errors (high residuals).
[0096] S42, Multi-fidelity data fusion verification strategy.
[0097] sparse high-fidelity sensor data (such as strain gauge measurements ε) sensor This is incorporated as a hard constraint during network training. A high-fidelity data loss term is added.
[0098]
[0099] in For the model in Predicted strain value, Let be the strain value measured by the j-th sensor.
[0100] The total loss function is now:
[0101]
[0102] in For the total loss function, For data loss, For physical constraint loss, Due to sensor data loss, , These are the weighting coefficients for each loss.
[0103] The network is forced to strictly match the predicted values at the strain gauge installation locations with the actual measured values. Through systematic comparative analysis of the full-field strain field reconstruction accuracy and physical consistency under three conditions—using only DIC data, introducing physical constraints, and fusing high-fidelity data—the engineering advantages of this method are verified.
[0104] The above-mentioned physical credibility quantification and fusion verification system is as follows: Figure 2-4 As shown, its construction process is as follows: First, the full-field physical residual norm and spatial local residual are calculated based on the displacement field and strain field output by the network; then, the local residual is normalized and visualized to generate a physical confidence assessment map, which intuitively identifies the measurement confidence level of each region; at the same time, the high-fidelity measured data of sparse strain gauge sensors are incorporated as hard constraints into the total loss of network training, ultimately realizing the unification of multi-source information from geometric observation, mechanical priors and physical sensing.
[0105] S5 outputs a physically consistent deformation field.
[0106] The blade DIC data and operating parameters to be measured are input into a trained bi-branch coupled network. The network outputs physically consistent displacement and strain fields, and generates corresponding physical reliability assessment maps for structural health monitoring and analysis.
[0107] Example 2: Variational Form of Physical Information Embedding Mechanism
[0108] This embodiment presents another implementation method for the physical constraints in step S2—the variational form. The deformation compatibility equation and the equilibrium equation are incorporated into the loss function in the form of variational residuals. Derivatives of each order are calculated through automatic differentiation, avoiding numerical errors caused by finite differences. Specifically, for the equilibrium equation, a weak form is used: , where v is an arbitrary test function. This form is more robust to low-quality grids or sparse data.
[0109] Example 3: Specific Structure of the Working Condition Coding Network
[0110] This embodiment provides a detailed structure of the working condition coding network in step S31. The network consists of three fully connected layers: the first layer maps the 4-dimensional working condition parameters to 64 dimensions and uses ReLU activation; the second layer maps the 64 dimensions to 128 dimensions and uses ReLU activation; the third layer maps the 128 dimensions to 32 dimensions and outputs the working condition coding vector z. cond The encoded vector is then fed into two branches: one branch predicts the equivalent material parameter E. eff (For constitutive relations), another branch predicts the loss weight λ. phy (Used to balance the contributions of data and physics). The output of the working condition coding network is... .
[0111] Example 4: Improvement of Gated Fusion Module – Multi-head Attention Mechanism
[0112] This embodiment improves the gating fusion module in step S32 by using a multi-head attention mechanism instead of simple gating weighting. Specifically, the observed features F obs As a query, the physical prior features F phy As the key and value, compute cross-modal attention:
[0113]
[0114] Where d k This represents the dimension of the Key vector.
[0115] Using four attention heads, the dimensionality of the output fused features remains unchanged. This mechanism can more flexibly capture the complex correlation between observed data and physical priors.
[0116] Example 5: Post-processing application of physical reliability maps
[0117] This embodiment provides a method for applying the physical reliability assessment map in step S41. The generated physical reliability map is used as the basis for subsequent structural health monitoring decisions: for high residual regions (low reliability), it is recommended to add local sensors or re-collect data; for low residual regions, the network prediction results can be directly trusted. Simultaneously, the residual threshold τ can be used as a criterion for judging the reliability of the model output. Warnings are issued in a timely manner.
[0118] Example 6: Measurement System
[0119] like Figure 3 As shown, based on the same idea, this embodiment of the invention also provides a physically consistent wind blade deformation field measurement system based on the fusion of PINN and DIC, the system comprising:
[0120] The physical information embedding module is used to embed the deformation compatibility equation as a strong constraint into the network output and to introduce the equilibrium law residual as a regularization term into the loss function.
[0121] The dual-branch coupled network module includes an observation branch and a physical sensing branch, which are used to realize adaptive sensing of working conditions and feature fusion.
[0122] The credibility quantification and evaluation module is used to generate a physical credibility evaluation map and perform multi-fidelity data fusion verification.
[0123] The deformation field output module is used to output physically consistent displacement and strain fields.
[0124] The dual-branch coupled network module includes: an observation branch unit (encoder-decoder structure), a physical sensing branch unit (operational condition coding network + parameterized PINN), and a gated fusion unit (gating mechanism or multi-head attention). The credibility quantification and evaluation module includes: a residual calculation unit, a normalization mapping unit, and a sensor hard constraint unit.
[0125] The functions or operation steps implemented by the above modules and units are largely the same as those in the above method embodiments, and will not be repeated here.
[0126] Example 7: Electronic Equipment
[0127] Based on the same idea, the seventh embodiment of the present invention also provides an electronic device, including a processor, a memory, a communication interface and a bus, wherein the processor implements the above method when executing a computer program stored in the memory.
[0128] Example 8: Storage Medium
[0129] Based on the same idea, the present invention also provides a storage medium on which a computer program is stored, which implements the above-described method when executed by a processor.
[0130] The basic principles of the present invention have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in the present invention are merely examples and not limitations, and should not be considered as essential features of each embodiment of the present invention. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the present invention to the necessity of employing the aforementioned specific details.
[0131] The block diagrams of devices, apparatuses, devices, and systems involved in this invention are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.
[0132] It should also be noted that in the apparatus, device, and method of the present invention, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of the present invention.
[0133] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the invention. Therefore, the invention is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.
[0134] It should be understood that the qualifying terms "first", "second", "third", "fourth", "fifth" and "sixth" used in the description of the embodiments of the present invention are only used to more clearly illustrate the technical solutions and are not intended to limit the scope of protection of the present invention.
[0135] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the invention to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.
Claims
1. A method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC, characterized in that, Includes the following steps: S1: Input the initial displacement field and current operating parameters of the wind turbine blade DIC; S2: Based on the input DIC initial displacement field and working parameters, a physical information embedding mechanism guided by the universal laws of mechanics is constructed. The deformation compatibility equation is embedded into the network output structure as a strong constraint, and the equilibrium law residual is introduced into the loss function as a regularization term to perform physical correction on the DIC initial displacement field. S3: Construct a dual-branch coupled network fusion architecture for adaptive working condition perception. The observation branch extracts the deformation features of the initial displacement field of DIC, and the physical perception branch dynamically generates physical constraint parameters that adapt to the current working condition with the working condition parameters as input. The observed deformation features and the prior physical features of the working condition are weighted and fused through the gated fusion module to obtain the fused features. S4: Based on the fusion features, a physical credibility quantification and fusion verification system is constructed. The full-field physical residual norm is calculated to generate a physical credibility evaluation map. Sparse high-fidelity sensor data is used as a hard constraint and integrated into the network training. S5: Outputs physically consistent wind turbine blade displacement field, strain field, and corresponding physical reliability assessment graph.
2. The method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC as described in claim 1, characterized in that, Step S2, which constructs a physical information embedding mechanism guided by universal laws of mechanics, specifically includes: The strain field is calculated based on the displacement field predicted by the network. The strain field is then substituted into the Saint-Venant compatibility equation to solve for the compatibility equation residuals. The stress field is calculated from the strain field based on the linear elastic constitutive relation to solve for the internal equilibrium equation residuals. The physical information loss function is constructed by combining the compatibility equation residuals and the equilibrium equation residuals. The total loss is obtained by combining the DIC matching data loss to complete the physical correction.
3. The method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC as described in claim 1, characterized in that, Step S3, constructing the network fusion architecture for adaptive working condition perception, specifically includes: The observation branch adopts an encoder-decoder structure, inputting full-field displacement data from the DIC to extract high-dimensional deformation features; the physical sensing branch is a parameterized PINN, taking working condition parameters as input, and dynamically generating adaptive physical constraint parameters through the working condition coding network; the gating weights are calculated to adaptively fuse deformation features and physical prior features, and the fused features are sent to the decoder to output the predicted displacement field and strain field.
4. The method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC as described in claim 1, characterized in that, Step S4, constructing a physical trust quantification and fusion verification system, specifically includes: The physical residual norm of the entire field and the local residual of a single point in space are calculated. The local residuals are normalized and visualized to generate a physical credibility assessment map. A high-fidelity sensor data loss term is added, and the data loss, physical loss and sensor loss are combined as the total loss for network training.
5. A physically consistent wind blade deformation field measurement system based on the fusion of PINN and DIC, characterized in that, For performing the measurement method according to any one of claims 1 to 4, comprising: The physical information embedding module is used to embed the deformation compatibility equation as a strong constraint into the network output and to introduce the equilibrium law residual as a regularization term into the loss function. The dual-branch coupled network module includes an observation branch and a physical sensing branch, which are used to realize adaptive sensing of working conditions and feature fusion. The credibility quantification and evaluation module is used to generate a physical credibility evaluation map and perform multi-fidelity data fusion verification. The deformation field output module is used to output physically consistent displacement and strain fields.
6. The physically consistent wind blade deformation field measurement system based on PINN and DIC fusion according to claim 5, characterized in that, The dual-branch coupled network module includes: The observation branch unit adopts an encoder-decoder structure, inputting the DIC displacement field and outputting deformation characteristics; The physical sensing branch unit includes a working condition coding network and a parameterized PINN, which takes working condition parameters as input and outputs adaptive physical constraint parameters. The gating fusion unit is used to calculate gating weights and fuse observed features with physical prior features.
7. The physically consistent wind blade deformation field measurement system based on PINN and DIC fusion according to claim 5, characterized in that, The credibility quantification and evaluation module includes: The residual calculation unit is used to calculate the residuals of the compatibility equations and the equilibrium equations. Normalized mapping units are used to generate physical credibility assessment maps; Sensor hard constraint units are used to incorporate sparse high-fidelity sensor data into the network loss function.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC as described in any one of claims 1 to 4.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for measuring the deformation field of a physically consistent wind turbine blade based on the fusion of PINN and DIC as described in any one of claims 1 to 4.