Multi-parameter optical fiber sensing network optimization method and system

Abstract

This invention discloses a multi-parameter fiber optic sensor network optimization method and system, relating to the field of fiber optic sensor network optimization. The method includes: establishing a three-dimensional finite element model of a wind turbine blade to obtain the blade's dynamic load distribution, mode shapes, and temperature field; identifying key areas of the blade and dividing the blade structure into five functional monitoring zones; deploying a hybrid fiber optic sensor array based on these functional monitoring zones; constructing a layered guiding optical cable network using a dual-channel helical winding structure; dynamically compensating for the original strain-vibration-temperature parameters acquired by the hybrid sensor array; and constructing a physical model-driven long short-term memory lifetime prediction network. The advantages of this invention are: through precise modeling, differentiated sensor deployment, error compensation, and intelligent lifetime prediction, it significantly improves the accuracy, reliability, and operation and maintenance efficiency of wind turbine blade monitoring.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensor network optimization, and in particular to a method and system for optimizing multi-parameter fiber optic sensor networks. Background Technology

[0002] Fiber optic sensing technology possesses unique advantages in real-time monitoring of physical quantities such as temperature, pressure, and strain due to its resistance to electromagnetic interference, long-distance transmission, high sensitivity, and multi-parameter measurement capabilities. However, with the increasing complexity of monitoring scenarios, traditional fiber optic sensor networks face problems such as low data processing efficiency, uneven distribution of sensor nodes, signal interference, and high system costs, which limit their application in large-scale, complex environments.

[0003] Most current fiber optic sensor network optimization methods on the market employ a single physical field or a single type of sensor, making it difficult to achieve comprehensive monitoring of multiple parameters and physical fields of wind turbine blades, resulting in incomplete monitoring information. Secondly, the sensor deployment methods are relatively simple, failing to take into account the differentiated characteristics of blade structure and load distribution, easily leading to monitoring blind spots or resource waste. Thirdly, the fiber optic network structure is simple, lacking hierarchical guidance and redundant backup design, resulting in low system reliability and data transmission efficiency. Fourthly, data processing mostly uses traditional filtering or simple compensation algorithms, failing to fully consider the coupling errors between multiple physical fields, affecting the accuracy of monitoring data. Fifthly, lifetime prediction mostly relies on empirical models or a single data source, lacking physical model constraints and intelligent algorithm support, resulting in inaccurate prediction results. Summary of the Invention

[0004] To improve existing methods and systems, this paper presents a multi-parameter fiber optic sensor network optimization method and system. This method significantly improves the accuracy, reliability, and operation and maintenance efficiency of wind turbine blade monitoring through precise modeling, differentiated sensor deployment, error compensation, and intelligent lifetime prediction.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for optimizing multi-parameter fiber optic sensor networks includes: A three-dimensional finite element model of a wind turbine blade was established, and the dynamic load distribution, mode shape and temperature field of the blade under different working conditions were calculated by combining aerodynamic load simulation. Based on simulation results, key areas of the blade were identified, including high stress concentration area, low-order mode vibration mode node area, and high temperature gradient area. The blade structure was divided into five functional monitoring areas, including root connection area, leading edge impact area, trailing edge vibration area, blade tip turbulence area, and web force transmission area. Based on the structural deformation characteristics and temperature sensitivity coefficients of the five functional monitoring areas, a hybrid fiber optic sensor array is deployed in a differentiated manner. The raw data acquired by the hybrid fiber optic sensor array is preprocessed, including noise filtering and feature extraction, to generate preprocessed sensor data. A dual-channel spiral winding structure is adopted to construct a layered optical cable network; Real-time monitoring of sensor performance and quality control to verify the reliability and stability of sensor data; Based on the Kalman filter algorithm and the multiphysics coupling error compensation model, dynamic compensation is performed on the original strain-vibration-temperature parameters acquired by the hybrid sensor array. By integrating multi-source sensor data with finite element models, a physical model-driven long short-term memory lifetime prediction network is constructed.

[0006] Preferably, the step of establishing a three-dimensional finite element model of the wind turbine blade and combining aerodynamic load simulation to calculate the dynamic load distribution, mode shapes, and temperature field of the blade under different operating conditions specifically includes: Obtain design data for wind turbine blades, including blade length, airfoil distribution, twist angle, thickness distribution, and root connection method, and create an accurate three-dimensional geometric model of the blades; Obtain the mechanical property parameters of the blade material, including elastic modulus, Poisson's ratio, density, thermal conductivity, and specific heat capacity; Aerodynamic load simulation is performed using computational fluid dynamics software to obtain the pressure and shear force distribution at various points on the blade surface, and output load data along the blade spanwise and chordwise directions. Calculate the stress, strain, and displacement distribution of the blade under various operating conditions to obtain the dynamic load distribution; Select the modal order, run modal analysis, and obtain the natural frequencies and mode shapes of the blade; The temperature field inside and on the surface of the blade is calculated by running steady-state and transient heat conduction analyses.

[0007] Preferably, the identification of key blade regions based on simulation results includes high stress concentration areas, low-order mode shape node areas, and high-temperature gradient areas. The blade structure is divided into five functional monitoring areas, including root connection area, leading-edge impact area, trailing-edge vibration area, tip turbulence area, and web force transmission area. Specifically, these include: A stress distribution cloud map is constructed based on dynamic load distribution data. According to the stress value distribution, areas with stress values ​​higher than the average value are marked as high stress concentration areas. Based on the modal data, the first 6-10 modes are extracted and the positions of the mode nodes are marked. By comparing the natural frequencies of the low-order modes with the excitation frequencies of the operating conditions, it is determined whether the node region is a vibration-sensitive point, and the node region of the low-order mode is obtained. Temperature distribution and gradient cloud map are generated based on temperature field data, and areas with temperature gradients greater than a preset threshold are marked as high temperature gradient areas. Based on the above analysis results, wind turbine blades are divided into five functional monitoring zones, including the root connection zone, leading edge impact zone, trailing edge vibration zone, blade tip turbulence zone, and web force transmission zone.

[0008] Preferably, the differentiated deployment of the hybrid fiber optic sensor array for the structural deformation characteristics and temperature sensitivity coefficients of the five functional monitoring areas specifically includes: A triaxial strained fiber grating array is deployed in the root connection region, with the directions being axial, circumferential and 45° shear direction; Distributed fiber optic sensors are arranged along the blade span in the leading-edge impact zone to cover the entire leading edge and monitor the strain distribution caused by aerodynamic loads. FBG sensors are added for local high-stress points. Fiber optic vibration sensors are arranged at mode nodes and in high-vibration areas at the trailing edge to monitor dynamic displacement and alternating stress, while FBG sensors measure static strain. Deploy micro-bend fiber optic accelerometers in the tip turbulence region to monitor dynamic strain caused by complex turbulence; FBG sensors are arranged along the connection between the web and the main beam in the web force transmission zone to monitor shear strain and the overall force transmission path.

[0009] Preferably, the preprocessing of the raw data acquired by the hybrid fiber optic sensor array, including noise filtering and feature extraction, to generate preprocessed sensor data specifically includes: Wavelet transform is used to filter the original sensor signal to remove high-frequency noise and low-frequency drift interference. Feature parameters are extracted through time-domain and frequency-domain analysis; The extracted feature parameters are fused with the original data to generate a preprocessed sensor data sequence.

[0010] Preferably, the construction of a layered guiding optical cable network using a dual-channel spiral winding structure specifically includes: The backbone layer uses a main optical cable spirally wound from the blade root to the blade tip and embedded in the blade main beam as the main data channel, responsible for large-scale data transmission and main area monitoring; Branch optical cables are drawn out at specific nodes at the leading edge, trailing edge, and leaf tip of the main optical cable. The branch optical cables are also spirally wound to cover local high-stress or high-temperature sensitive areas, forming a local monitoring network. Each monitoring area should have at least two independent channels, which should serve as backups for each other; Layered connectors are installed at the intersection of the trunk and branches to guide signals in layers and aggregate data, and to plan the transmission path of optical signals to avoid mutual interference between trunk and branch signals.

[0011] Preferably, the real-time monitoring of sensor performance and quality control, and the verification of the reliability and stability of sensor data, specifically include: The wavelength drift and signal attenuation of the FBG sensor are monitored by the fiber grating reflection spectrum to determine the sensor's health status. Deploy redundant sensor nodes in the branch optical cable network, compare the data consistency of adjacent nodes, and identify abnormal data and faulty sensors. An automated mechanism is implemented based on anomaly detection results, with manual intervention as needed.

[0012] Preferably, the dynamic compensation of the original strain-vibration-temperature parameters acquired by the hybrid sensor array based on the Kalman filter algorithm and the multiphysics coupling error compensation model specifically includes: Based on the data types acquired by various fiber optic sensors, a coupled analysis is performed, including the influence of temperature and vibration on strain, the influence of temperature and strain on vibration, and the influence of environmental and structural heat conduction on temperature. Based on the three-dimensional finite element model of wind turbine blades and the results of coupling analysis, a multi-physics coupling relationship matrix is ​​obtained, and a multi-physics coupling error model is constructed by introducing error terms. Using the Kalman filter algorithm, a state transition equation is constructed based on the state variable variation law of the wind turbine blade. Coupled with a multi-physics coupling error model, coupling error compensation is performed to correct the cross-influence of temperature on strain and vibration. Differentiated compensation is performed for different regions of the wind turbine blade. All collected data are subjected to Kalman filtering, and the compensated strain, vibration and temperature data are output in real time.

[0013] Preferably, the method of fusing multi-source sensor data and finite element models to construct a physical model-driven long short-term memory lifetime prediction network specifically includes: Based on the analysis of physical field coupling relationship between multi-source sensor data and finite element model, a multi-physics coupling model is constructed. The lifespan of wind turbine blades is predicted by training an LSTM neural network model and embedding physical constraints. The effectiveness of physical constraints is verified by comparing the life predicted by LSTM with the actual fatigue test results. The real-time sensor data and the physical features updated by the finite element model are input into the LSTM neural network, and the output is the lifetime prediction result represented in the form of a probability distribution.

[0014] Furthermore, a multi-parameter fiber optic sensor network optimization system is proposed, comprising: Finite element modeling and simulation module: This module is responsible for establishing a three-dimensional finite element model of the wind turbine blade, combining aerodynamic load simulation, and outputting dynamic load, mode shape and temperature field distribution data under different working conditions; Region identification and functional area division module: Based on simulation results, the module automatically identifies high stress concentration areas, low-order mode vibration mode node areas, and high temperature gradient areas, and divides the blade structure into five types of functional monitoring areas; Fiber optic sensor deployment module: The module is designed and deployed with multiple types of fiber optic sensor arrays to achieve multi-parameter monitoring, taking into account the structural deformation and temperature sensitivity characteristics of different functional monitoring areas. Layered optical fiber network construction module: The module adopts a dual-channel spiral winding structure to construct a hierarchical optical fiber network with backbone and branch guidance, so as to achieve efficient data transmission and local redundancy backup; Error compensation module: The module is based on the Kalman filter algorithm and the multiphysics coupling error model to dynamically compensate the original strain, vibration and temperature data, thereby improving the accuracy of the data; Lifetime prediction module: The module integrates multi-source sensor data and finite element model to perform multi-physics coupling relationship analysis and predicts the lifetime of wind turbine blades through LSTM neural network; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0015] Compared with the prior art, the advantages of the present invention are: Through finite element simulation and aerodynamic load analysis, accurate modeling of multi-physics fields and identification of key areas of wind turbine blades were achieved, improving the scientific rigor and relevance of monitoring. Secondly, a hybrid fiber optic sensor array was deployed differently for different functional areas, combined with a hierarchical guiding optical cable network design, ensuring both comprehensive and efficient data acquisition while enhancing system redundancy and reliability. Kalman filtering and multi-physics coupling error compensation were employed to effectively eliminate cross-interference between multiple parameters such as strain, vibration, and temperature, improving the accuracy of monitoring data. Finally, by integrating multi-source sensor data with the finite element model and utilizing a physically constrained LSTM neural network, intelligent prediction of wind turbine blade life was achieved, providing a scientific basis for blade health management and operation and maintenance decisions. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method proposed in this invention; Figure 2 This is a schematic diagram illustrating the calculation of wind turbine blade data proposed in this invention; Figure 3 This is a schematic diagram illustrating the division of key regions proposed in this invention; Figure 4 This is a schematic diagram of the deployment of the hybrid fiber optic sensor array proposed in this invention; Figure 5 This is a schematic diagram of the data preprocessing proposed in this invention. Figure 6This is a schematic diagram of the hierarchical optical cable network proposed in this invention; Figure 7 This is a schematic diagram of the sensor performance quality control proposed in this invention. Figure 8 This is a schematic diagram of the dynamic compensation of the original parameters proposed in this invention; Figure 9 This is a schematic diagram of the lifetime prediction network proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] A multi-parameter fiber optic sensor network optimization system includes: Finite element modeling and simulation module: This module is responsible for establishing a three-dimensional finite element model of the wind turbine blade, combining aerodynamic load simulation, and outputting dynamic load, mode shape and temperature field distribution data under different working conditions; Region identification and functional area division module: Based on simulation results, the module automatically identifies high stress concentration areas, low-order mode vibration mode node areas, and high temperature gradient areas, and divides the blade structure into five types of functional monitoring areas; Fiber optic sensor deployment module: The module is designed and deployed with multiple types of fiber optic sensor arrays to achieve multi-parameter monitoring, taking into account the structural deformation and temperature sensitivity characteristics of different functional monitoring areas. Layered optical fiber network construction module: The module adopts a dual-channel spiral winding structure to construct a hierarchical optical fiber network with backbone and branch guidance, so as to achieve efficient data transmission and local redundancy backup; Error compensation module: The module is based on the Kalman filter algorithm and the multiphysics coupling error model to dynamically compensate the original strain, vibration and temperature data, thereby improving the accuracy of the data; Lifetime prediction module: The module integrates multi-source sensor data and finite element model to perform multi-physics coupling relationship analysis and predicts the lifetime of wind turbine blades through LSTM neural network; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0019] See Figure 1 As shown, a method for optimizing a multi-parameter fiber optic sensor network includes: Step 1: Establish a three-dimensional finite element model of the wind turbine blade, and combine aerodynamic load simulation to calculate the dynamic load distribution, mode shape and temperature field of the blade under different working conditions; Step 2: Based on the simulation results, identify the key areas of the blade, including the high stress concentration area, the low-order mode vibration mode node area, and the high temperature gradient area. Divide the blade structure into five functional monitoring areas, including the root connection area, the leading edge impact area, the trailing edge vibration area, the tip turbulence area, and the web force transmission area. Step 3: Based on the structural deformation characteristics and temperature sensitivity coefficients of the five functional monitoring areas, deploy hybrid fiber optic sensor arrays in a differentiated manner; Step 4: Preprocess the raw data acquired by the hybrid fiber optic sensor array, including noise filtering and feature extraction, to generate preprocessed sensor data; Step 5: Construct a layered optical cable network using a dual-channel spiral winding structure; Step Six: Monitor sensor performance in real time and perform quality control to verify the reliability and stability of sensor data; Step 7: Based on the Kalman filter algorithm and the multiphysics coupling error compensation model, dynamically compensate the original strain-vibration-temperature parameters acquired by the hybrid sensor array; Step 8: Integrate multi-source sensor data with the finite element model to construct a physical model-driven long short-term memory lifetime prediction network.

[0020] See Figure 2 As shown, a three-dimensional finite element model of the wind turbine blade is established, and the dynamic load distribution, mode shapes, and temperature field of the blade under different operating conditions are calculated by combining aerodynamic load simulation. Specifically, this includes: Obtain design data for wind turbine blades, including blade length, airfoil distribution, twist angle, thickness distribution, and root connection method, and create an accurate three-dimensional geometric model of the blades; Obtain the mechanical property parameters of the blade material, including elastic modulus, Poisson's ratio, density, thermal conductivity, and specific heat capacity; Aerodynamic load simulation is performed using computational fluid dynamics software to obtain the pressure and shear force distribution at various points on the blade surface, and output load data along the blade spanwise and chordwise directions. Calculate the stress, strain, and displacement distribution of the blade under various operating conditions to obtain the dynamic load distribution; Select the modal order, run modal analysis, and obtain the natural frequencies and mode shapes of the blade; The temperature field inside and on the surface of the blade is calculated by running steady-state and transient heat conduction analyses.

[0021] Specifically, in aerodynamic load simulation, the aerodynamic force per unit area is:

[0022] in, For the aerodynamic forces on the infinitesimal area, For pressure, For the surface normal vector, The shear stress vector The area of ​​the infinitesimal element; Static analysis is performed using pressure and shear force distribution as boundary conditions to calculate steady-state stress, strain, and displacement. Simultaneously, transient dynamic analysis is performed to obtain the stress, strain, and displacement distribution under dynamic loads. In the finite element software, set up the modal analysis module, specify the required modal order, evaluate the dynamic stability of the blade, and avoid the excitation frequency of the operating speed to prevent resonance. The first few modes usually include bending, torsion and coupled modes. Steady-state analysis is suitable for long-term stable environments, while transient analysis is suitable for rapidly changing thermal loads. The low thermal conductivity of composite materials may lead to local high temperatures, so special attention should be paid to the effect of temperature gradients on material properties.

[0023] See Figure 3 As shown, based on simulation results, key areas of the blade are identified, including high stress concentration areas, low-order mode shape node areas, and high-temperature gradient areas. The blade structure is divided into five functional monitoring areas, including the root connection area, leading-edge impact area, trailing-edge vibration area, tip turbulence area, and web force transmission area. Specifically, these include: A stress distribution cloud map is constructed based on dynamic load distribution data. According to the stress value distribution, areas with stress values ​​higher than the average value are marked as high stress concentration areas. Based on the modal data, the first 6-10 modes are extracted and the positions of the mode nodes are marked. By comparing the natural frequencies of the low-order modes with the excitation frequencies of the operating conditions, it is determined whether the node region is a vibration-sensitive point, and the node region of the low-order mode is obtained. Temperature distribution and gradient cloud map are generated based on temperature field data, and areas with temperature gradients greater than a preset threshold are marked as high temperature gradient areas. Based on the above analysis results, wind turbine blades are divided into five functional monitoring zones, including the root connection zone, leading edge impact zone, trailing edge vibration zone, blade tip turbulence zone, and web force transmission zone.

[0024] Specifically, dynamic load distribution data is imported and used as boundary conditions for the finite element model to perform transient structural analysis. The stress distribution of the blade at each time and under each working condition is calculated. The maximum, minimum, and average stress values ​​are obtained for each node or element. The stress distribution is statistically analyzed and a stress cloud diagram is plotted. Modal analysis is used to obtain the first 6-10 mode shape data of the blade, including natural frequencies and mode shape functions. For each mode shape, mode shape nodes are identified, node position coordinates are recorded, excitation frequencies under wind turbine blade operating conditions are collected, and the natural frequencies of lower-order modes are compared with excitation frequencies. If the natural frequencies are less than a preset frequency threshold, the node region of that mode is a vibration sensitive point. The node regions of vibration sensitive points in the 6-10 modes are summarized to form a set of vibration sensitive regions. Acquire temperature distribution data of wind turbine blades under operating or environmental conditions, calculate temperature gradient based on temperature field data, set temperature gradient threshold according to material tolerance and design requirements, compare temperature gradient at each point with the threshold, and mark high temperature gradient areas. The root connection zone, located near the blade root, is affected by high stress concentration and low-order modal vibration, requiring monitoring of stress and fatigue damage. The leading-edge impact zone, at the blade's leading edge, is affected by aerodynamic loads and wind and sand impacts, exhibiting high stress concentration and localized high-temperature gradients. The trailing-edge vibration zone, at the blade's trailing edge, is affected by airflow disturbances and is prone to vibration, making it a vibration-sensitive point. The blade tip turbulence zone, at the blade tip, is affected by complex airflow, exhibiting high-temperature gradients and being a vibration-sensitive point. The web force transmission zone, located within the blade's internal web region, is responsible for transmitting loads and exhibits high stress concentration.

[0025] See Figure 4 As shown, the differentiated deployment of hybrid fiber optic sensor arrays, based on the structural deformation characteristics and temperature sensitivity coefficients of the five functional monitoring areas, specifically includes: A triaxial strained fiber grating array is deployed in the root connection region, with the directions being axial, circumferential and 45° shear direction; Distributed fiber optic sensors are arranged along the blade span in the leading-edge impact zone to cover the entire leading edge and monitor the strain distribution caused by aerodynamic loads. FBG sensors are added for local high-stress points. Fiber optic vibration sensors are arranged at mode nodes and in high-vibration areas at the trailing edge to monitor dynamic displacement and alternating stress, while FBG sensors measure static strain. Deploy micro-bend fiber optic accelerometers in the tip turbulence region to monitor dynamic strain caused by complex turbulence; FBG sensors are arranged along the connection between the web and the main beam in the web force transmission zone to monitor shear strain and the overall force transmission path.

[0026] See Figure 5 As shown, the preprocessing of the raw data acquired by the hybrid fiber optic sensor array, including noise filtering and feature extraction, to generate preprocessed sensor data specifically includes: Wavelet transform is used to filter the original sensor signal to remove high-frequency noise and low-frequency drift interference. Feature parameters are extracted through time-domain and frequency-domain analysis; The extracted feature parameters are fused with the original data to generate a preprocessed sensor data sequence.

[0027] See Figure 6 As shown, the hierarchical optical cable network constructed using a dual-channel spiral winding structure specifically includes: The backbone layer uses a main optical cable spirally wound from the blade root to the blade tip and embedded in the blade main beam as the main data channel, responsible for large-scale data transmission and main area monitoring; Branch optical cables are drawn out at specific nodes at the leading edge, trailing edge, and leaf tip of the main optical cable. The branch optical cables are also spirally wound to cover local high-stress or high-temperature sensitive areas, forming a local monitoring network. Each monitoring area should have at least two independent channels, which should serve as backups for each other; Layered connectors are installed at the intersection of the trunk and branches to guide signals in layers and aggregate data, and to plan the transmission path of optical signals to avoid mutual interference between trunk and branch signals.

[0028] Specifically, the blade main beam is selected as the embedding carrier of the main optical cable. A main optical cable is arranged along the center line of the main beam from the blade root to the blade tip. The spiral winding method is used to enhance the coverage and structural adaptability. The main optical cable is embedded in the main beam or on the surface to ensure a firm bond with the blade material and avoid slippage or damage. The optical cable serves as the main data channel and is responsible for transmitting strain, vibration and temperature signals of the five major functional monitoring areas. Based on the analysis of the five functional monitoring zones (high stress zone, vibration sensitive point, and high temperature gradient zone), branch nodes are selected on the main optical cable, located in the high stress or high temperature sensitive areas of the leading edge, trailing edge, and blade tip, respectively. Branch optical cables are led out from the nodes of the main optical cable and covered in a spiral winding manner along local areas of the monitoring zone (such as the entire length of the leading edge, the trailing edge vibration mode node, and the blade tip turbulence zone). The branch optical cables connect the distributed optical fibers, FBG sensors, or fiber optic accelerometers in the area to form a local high-resolution monitoring network. At least two independent fiber optic channels are arranged in each functional monitoring area, one as the main channel and the other as a backup channel. The main channel and the backup channel are arranged along different paths along the blade, such as the main channel along the upper surface of the main beam and the backup channel along the lower surface, to avoid the impact of single point failure. By periodically switching the main and backup channels, the signal integrity of the backup channel is checked to ensure the reliability of data transmission.

[0029] See Figure 7 As shown, the real-time monitoring of sensor performance and quality control, and the verification of the reliability and stability of sensor data, specifically include: The wavelength drift and signal attenuation of the FBG sensor are monitored by the fiber grating reflection spectrum to determine the sensor's health status. Deploy redundant sensor nodes in the branch optical cable network, compare the data consistency of adjacent nodes, and identify abnormal data and faulty sensors. An automated mechanism is implemented based on anomaly detection results, with manual intervention as needed.

[0030] See Figure 8 As shown, the dynamic compensation of the original strain-vibration-temperature parameters acquired by the hybrid sensor array based on the Kalman filter algorithm and the multiphysics coupling error compensation model specifically includes: Based on the data types acquired by various fiber optic sensors, a coupled analysis is performed, including the influence of temperature and vibration on strain, the influence of temperature and strain on vibration, and the influence of environmental and structural heat conduction on temperature. Based on the three-dimensional finite element model of wind turbine blades and the results of coupling analysis, a multi-physics coupling relationship matrix is ​​obtained, and a multi-physics coupling error model is constructed by introducing error terms. Using the Kalman filter algorithm, a state transition equation is constructed based on the state variable variation law of the wind turbine blade. Coupled with a multi-physics coupling error model, coupling error compensation is performed to correct the cross-influence of temperature on strain and vibration. Differentiated compensation is performed for different regions of the wind turbine blade. All collected data are subjected to Kalman filtering, and the compensated strain, vibration and temperature data are output in real time.

[0031] Specifically, the effects of temperature on strain include thermal expansion caused by temperature changes, leading to strain deviation; the effects of vibration on strain include the superposition of dynamic strain caused by vibration on static strain; the effects of temperature on vibration include changes in material stiffness caused by temperature changes, affecting natural frequencies; the effects of strain on vibration include the potential for local stiffness changes in high-stress areas, affecting mode shapes; environmental influences include changes in blade surface temperature caused by external temperature, wind speed, solar radiation, etc.; and structural heat conduction includes temperature gradients caused by internal heat conduction within the blade. By comprehensively analyzing the interactions between various physical fields, coupling relationship functions are formed, including the strain, vibration, and temperature coupling relationship functions of each monitoring area, which describe the multi-physics field interaction effects. Thermal structure coupling and vibration structure coupling modules are introduced into the finite element model to calculate the mutual influence between various physical fields. Finite element simulation is performed using actual operating data to verify the accuracy, adjust matrix parameters, multi-physics coupling relationship matrix and error model for subsequent error compensation. Define state variables, construct a state transition model based on blade physics (such as material stiffness and heat conduction laws) and operating conditions, define the fiber optic sensor measurement values ​​as observation vectors, and integrate the multi-physics coupling error model into the noise covariance matrix of Kalman filtering to correct the cross-influence of temperature on strain and vibration. Through Kalman filtering iteration, the corrected state vector is output, eliminating the cross-influence of temperature on strain and vibration, and outputting the state transition equation, noise covariance matrix and compensated state vector of the Kalman filter.

[0032] See Figure 9 As shown, the physical model-driven long short-term memory lifetime prediction network, which integrates multi-source sensor data and finite element models, specifically includes: Based on the analysis of physical field coupling relationship between multi-source sensor data and finite element model, a multi-physics coupling model is constructed. The lifespan of wind turbine blades is predicted by training an LSTM neural network model and embedding physical constraints. The effectiveness of physical constraints is verified by comparing the life predicted by LSTM with the actual fatigue test results. The real-time sensor data and the physical features updated by the finite element model are input into the LSTM neural network, and the output is the lifetime prediction result represented in the form of a probability distribution.

[0033] Specifically, the multi-source sensor data and operating conditions are standardized according to time series to generate an input sequence. The input layer receives the time series, the hidden layer contains multiple LSTM units to capture long-term dependencies in the time series, the output layer predicts damage accumulation or remaining lifetime, and outputs a trained LSTM model that embeds multi-physics coupling and fatigue damage physical constraints, outputting damage accumulation and remaining lifetime. Fatigue tests are conducted on the blades, and the actual fatigue life and strain, vibration, and temperature data under the corresponding working conditions are recorded. The fatigue test conditions are input into the LSTM model to predict damage accumulation and remaining life. The predictions are compared with the actual results, and error analysis reports and prediction results are obtained, along with the remaining life probability distribution updated in real time.

[0034] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0035] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0036] 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 principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing multi-parameter fiber optic sensor networks, characterized in that, include: A three-dimensional finite element model of a wind turbine blade was established, and the dynamic load distribution, mode shape and temperature field of the blade under different working conditions were calculated by combining aerodynamic load simulation. Based on simulation results, key areas of the blade were identified, including high stress concentration area, low-order mode vibration mode node area, and high temperature gradient area. The blade structure was divided into five functional monitoring areas, including root connection area, leading edge impact area, trailing edge vibration area, blade tip turbulence area, and web force transmission area. Based on the structural deformation characteristics and temperature sensitivity coefficients of the five functional monitoring areas, a hybrid fiber optic sensor array is deployed in a differentiated manner. The raw data acquired by the hybrid fiber optic sensor array is preprocessed, including noise filtering and feature extraction, to generate preprocessed sensor data. A dual-channel spiral winding structure is adopted to construct a layered optical cable network; Real-time monitoring of sensor performance and quality control to verify the reliability and stability of sensor data; Based on the Kalman filter algorithm and the multiphysics coupling error compensation model, dynamic compensation is performed on the original strain-vibration-temperature parameters acquired by the hybrid sensor array. By integrating multi-source sensor data with finite element models, a physical model-driven long short-term memory lifetime prediction network is constructed.

2. The method for optimizing a multi-parameter fiber optic sensor network according to claim 1, characterized in that, The establishment of a three-dimensional finite element model of the wind turbine blade, combined with aerodynamic load simulation calculations of the blade's dynamic load distribution, mode shapes, and temperature field under different operating conditions, specifically includes: Obtain design data for wind turbine blades, including blade length, airfoil distribution, twist angle, thickness distribution, and root connection method, and create an accurate three-dimensional geometric model of the blades; Obtain the mechanical property parameters of the blade material, including elastic modulus, Poisson's ratio, density, thermal conductivity, and specific heat capacity; Aerodynamic load simulation is performed using computational fluid dynamics software to obtain the pressure and shear force distribution at various points on the blade surface, and output load data along the blade spanwise and chordwise directions. Calculate the stress, strain, and displacement distribution of the blade under various operating conditions to obtain the dynamic load distribution; Select the modal order, run modal analysis, and obtain the natural frequencies and mode shapes of the blade; The temperature field inside and on the surface of the blade is calculated by running steady-state and transient heat conduction analyses.

3. A multi-parameter fiber optic sensor network optimization system, used to implement the multi-parameter fiber optic sensor network optimization method as described in any one of claims 1-2, characterized in that, include: Finite element modeling and simulation module: This module is responsible for establishing a three-dimensional finite element model of the wind turbine blade, combining aerodynamic load simulation, and outputting dynamic load, mode shape and temperature field distribution data under different working conditions; Region identification and functional area division module: Based on simulation results, the module automatically identifies high stress concentration areas, low-order mode vibration mode node areas, and high temperature gradient areas, and divides the blade structure into five types of functional monitoring areas; Fiber optic sensor deployment module: The module is designed and deployed with multiple types of fiber optic sensor arrays to achieve multi-parameter monitoring, taking into account the structural deformation and temperature sensitivity characteristics of different functional monitoring areas. Layered optical fiber network construction module: The module adopts a dual-channel spiral winding structure to construct a hierarchical optical fiber network with backbone and branch guidance, so as to achieve efficient data transmission and local redundancy backup; Error compensation module: The module is based on the Kalman filter algorithm and the multiphysics coupling error model to dynamically compensate the original strain, vibration and temperature data, thereby improving the accuracy of the data; Lifetime prediction module: The module integrates multi-source sensor data and finite element model to perform multi-physics coupling relationship analysis and predicts the lifetime of wind turbine blades through LSTM neural network; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images