Calibration test method, device, equipment and storage medium for fan blades

Abstract

The present invention discloses a calibration test method, device, equipment and storage medium for a wind turbine blade, the method comprising dividing the blade to be tested into a root section, a body section and a tip section along its length direction; performing multi-modal dynamic scanning on the blade to be tested, obtaining acoustic, vibration and deformation response data sets of the root section, body section and tip section under different rotation speeds and wind speeds; using a deep learning network to perform time series analysis on the response data sets, generating dynamic weight distribution diagrams and stress distribution diagrams of the root section, body section and tip section; performing multi-dimensional comparison of the dynamic weight distribution diagram and stress distribution diagram with a standard model in a pre-established blade database, selecting the most matching standard model, and performing parameterized adjustment according to actual differences; based on the adjusted model, generating a physical calibration guide for the root section, body section and tip section, and performing segmented weight adjustment on the blade to be tested. The technical solution of the present invention ensures that the test results have a high degree of authenticity and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of calibration and testing of fan blades, and in particular to a calibration and testing method, device, equipment and storage medium for fan blades. Background Art

[0002] Wind turbine blades are key components of wind turbines, and their performance directly affects the efficiency and stability of the entire wind power system. Since wind turbine blades are subjected to complex wind loads during operation, changes in their weight distribution and stress distribution will cause vibration, deformation, and fatigue accumulation of the blades, which in turn affects the service life and power generation efficiency of the blades. If the blades fail to undergo precise calibration tests, problems such as excessive local stress and uneven blade deformation may occur, and even affect the overall safety of the wind turbine. Therefore, precise calibration tests must be performed on wind turbine blades to ensure the structural integrity and operational stability of the blades under different operating conditions and to improve the overall performance of wind turbines.

[0003] Existing wind turbine blade calibration technology usually adopts static measurement or single parameter testing methods, that is, when the blade is stopped or running at low speed, the blade's geometric deformation, stress concentration area and other physical parameters are measured to infer its operating status. This method relies on limited working conditions in a laboratory environment and cannot truly simulate the dynamic response characteristics of wind turbine blades under variable wind speeds and rotation speeds in actual operation. Since these test methods fail to comprehensively consider the multimodal data of blades under different operating conditions (such as acoustics, vibration, deformation, etc.), the results can only reflect the performance of the blade under a certain specific condition, but cannot accurately capture the stress and weight distribution of the blade under actual complex working conditions. In summary, the existing calibration tests have the problem of insufficient accuracy in practical applications, and it is difficult to effectively measure the performance changes of blades under different working conditions. Summary of the invention

[0004] The main purpose of the present invention is to solve the problem of insufficient accuracy of existing fan blade calibration tests in practical applications, and the technical problem that it is difficult to effectively measure the performance changes of the blades under different working conditions.

[0005] A first aspect of the present invention provides a calibration test method for a fan blade, the calibration test method for the fan blade comprising:

[0006] Divide the blade to be tested into a blade root section, a blade body section and a blade tip section along its length direction, wherein the blade root section is a fixed end connected to the hub;

[0007] Performing multi-modal dynamic scanning on the blade to be tested to obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds;

[0008] Performing time series analysis on the response data set using a deep learning network to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, the blade body section, and the blade tip section;

[0009] Perform multi-dimensional comparison of the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database, select the most matching standard model, and perform parameter adjustment according to the actual differences;

[0010] Based on the adjusted model, physical calibration guides for the blade root section, the blade body section and the blade tip section are generated, and the section weight adjustment is performed on the blade to be tested.

[0011] Optionally, dividing the blade to be measured into a blade root section, a blade body section and a blade tip section along its length direction includes:

[0012] Measuring the total length L of the blade to be measured, and marking equally spaced measuring points on the surface of the blade to be measured along the length direction;

[0013] Perform strain testing at the measuring point to determine the stress distribution of the blade to be tested under different working conditions, define the distance from the maximum stress point to the blade root end as L1, and divide the area into a blade root section;

[0014] Perform an aerodynamic characteristic test at the measuring point to determine the critical position of aerodynamic characteristic change, define the distance from the point to the blade tip as L3, and divide the area into a blade tip section;

[0015] The area between the blade root section and the blade tip section is divided into a blade body section, the length of which is L2=L-L1-L3;

[0016] The ratio of L1, L2 and L3 is adjusted according to the specific model and working environment of the blade to be tested.

[0017] Optionally, performing multi-modal dynamic scanning on the blade to be tested to obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds includes:

[0018] A plurality of working conditions of rotation speed and wind speed combination are set in the wind tunnel, and the root section, the body section and the tip section of the blade to be tested are rotated to obtain the original response data of each section under different working conditions;

[0019] Using an acoustic sensor array, a blade root section, a blade body section, and a blade tip section of the blade to be tested are respectively subjected to an omnidirectional acoustic scan to obtain an acoustic response signal of each section, and a time-frequency analysis is performed on the acoustic response signal of each section to obtain an acoustic characteristic spectrum of the blade root section, the blade body section, and the blade tip section;

[0020] The fiber grating sensor network is arranged on the surface of the blade root section, blade body section and blade tip section of the blade to be tested, and the vibration and strain data of each section are collected respectively. The vibration and strain data of each section are subjected to modal analysis to obtain the vibration characteristic vectors of the blade root section, blade body section and blade tip section.

[0021] A high-speed camera system is used to capture the dynamic deformation process of the blade root section, blade body section and blade tip section under different working conditions. The profile change information of each blade section is extracted through image processing algorithm to obtain the deformation feature matrix of the blade root section, blade body section and blade tip section.

[0022] The acoustic characteristic spectra, vibration characteristic vectors and deformation characteristic matrices of the blade root section, blade body section and blade tip section are fused respectively to construct a comprehensive response data set of each section.

[0023] Optionally, the using a deep learning network to perform a time series analysis on the response data set to generate a dynamic weight distribution diagram and a stress distribution diagram of the blade root section, the blade body section, and the blade tip section includes:

[0024] Performing time series decomposition on the response data sets of the blade root section, the blade body section and the blade tip section respectively to obtain trend items, period items and random items of each section;

[0025] The trend items of the blade root section, the blade body section and the blade tip section are respectively input into a pre-trained three-dimensional convolutional neural network, the spatial features of each section are extracted, and a preliminary weight distribution feature map of each section is obtained;

[0026] Applying long short-term memory networks to the periodic terms and random terms of the blade root segment, the blade body segment and the blade tip segment respectively to capture the temporal dependency of each segment and obtain a time-varying feature vector of each segment;

[0027] The preliminary weight distribution feature maps of the blade root section, the blade body section and the blade tip section are fused with the corresponding time-varying feature vectors, and the three-dimensional spatial distribution of each section is reconstructed through a deconvolution network to obtain a dynamic weight distribution map of each section;

[0028] The dynamic weight distribution diagrams of the blade root section, blade body section and blade tip section are processed as a whole using a graph neural network, and the connection relationship between the blade sections is considered to generate a stress distribution diagram of the complete blade.

[0029] Optionally, the dynamic weight distribution diagram and the stress distribution diagram are compared with a standard model in a pre-established blade database in multiple dimensions, the most matching standard model is selected, and parameterized adjustments are made according to actual differences, including:

[0030] Extracting features from the dynamic weight distribution diagram and stress distribution diagram of the blade root section, the blade body section and the blade tip section to obtain a multi-dimensional feature vector of the blade to be tested;

[0031] Using a multi-objective optimization algorithm, similarity calculation is performed between the multi-dimensional feature vector and a standard model in a leaf database to obtain a candidate matching model set;

[0032] Applying the analytic hierarchy process to the candidate matching model set, taking into account weight distribution, stress distribution and geometric shape factors, to obtain the most matching standard model;

[0033] Constructing a parameter mapping relationship between the blade to be tested and the most matching standard model to generate a parameterized adjustment matrix;

[0034] The parameterized adjustment matrix is ​​optimized using a genetic algorithm to obtain an optimal parameter adjustment solution.

[0035] Optionally, generating physical calibration guides for the blade root section, the blade body section, and the blade tip section based on the adjusted model, and performing segmented weight adjustment on the blade to be tested, includes:

[0036] Performing finite element analysis on the adjusted model, calculating the ideal weight distribution of the blade root section, the blade body section and the blade tip section, and obtaining a target weight distribution diagram of each section;

[0037] Performing difference calculation between the target weight distribution diagram and the current weight distribution diagram of the blade to be measured, so as to generate weight adjustment amount distribution diagrams of the blade root section, the blade body section and the blade tip section;

[0038] According to the weight adjustment amount distribution map, a topology optimization algorithm is used to calculate the weight adjustment point locations and adjustment amounts in each segment, and a detailed physical calibration guide is generated;

[0039] The blade root section, blade airfoil section and blade tip section are injected or removed with targeted weight material according to the physical calibration guide.

[0040] Optionally, after injecting or removing the fixed-point weight material into or from the blade root section, the blade body section, and the blade tip section according to the physical calibration guide, the method further includes:

[0041] Perform a quick scan on the blade to be tested after weight adjustment to obtain the actual weight distribution after adjustment;

[0042] Comparing the adjusted actual weight distribution with the target weight distribution diagram to obtain a weight distribution deviation value;

[0043] Determine whether the weight distribution deviation value exceeds a preset threshold value. If so, return to the step of "calculating the weight adjustment point position and adjustment amount in each segment using a topology optimization algorithm according to the weight adjustment amount distribution map, and generating a detailed physical calibration guide" to regenerate a local fine-tuning physical calibration guide;

[0044] According to the physical calibration guide of the local fine-tuning, the weight of the blade to be tested is adjusted again.

[0045] A second aspect of the present invention provides a calibration test device for a fan blade, comprising:

[0046] A blade segmentation module is used to divide the blade to be tested into a blade root segment, a blade body segment and a blade tip segment along its length direction, wherein the blade root segment is a fixed end connected to the hub;

[0047] A multi-modal scanning module, used to perform multi-modal dynamic scanning on the blade to be tested, and obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds;

[0048] A deep learning analysis module, used for performing time series analysis on the response data set using a deep learning network to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, the blade body section, and the blade tip section;

[0049] A model comparison processing module is used to compare the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database in multiple dimensions, select the most matching standard model, and perform parameterized adjustments according to actual differences;

[0050] The calibration execution module is used to generate physical calibration guidelines for the blade root section, the blade body section and the blade tip section based on the adjusted model, and to perform segmented weight adjustment on the blade to be tested.

[0051] A third aspect of the present invention provides a calibration and testing device for wind blades, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the calibration and testing device for wind blades performs the steps of the above-mentioned calibration and testing method for wind blades.

[0052] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned wind turbine blade calibration test method.

[0053] The wind turbine blade calibration test method provided by the present invention can comprehensively capture the response data of the blade under various working conditions under different rotational speeds and wind speeds by dividing the blade into a root section, a body section and a tip section for multi-modal dynamic scanning. This multi-modal dynamic scanning combines the collection of acoustic, vibration and deformation data to ensure that all physical properties of the blade can be accurately measured and recorded under various operating conditions. Compared with the static or single parameter testing methods in the prior art, the method of the present invention obtains more real and comprehensive data through dynamic testing, solving the problem that the existing methods cannot effectively evaluate the dynamic weight distribution and stress distribution of the blade under complex working conditions. In this way, the working state of the blade can not only be verified in the laboratory, but also better simulate the complex operating conditions in the actual environment, ensuring that the test results are highly authentic and accurate.

[0054] In addition, the present invention uses a deep learning network to perform time series analysis on the response data set to generate dynamic weight distribution diagrams and stress distribution diagrams for the blade root section, blade body section, and blade tip section. This feature enables the test results to reflect the changing trend of the blade under different working conditions, rather than just the test results at a single moment. The deep learning network captures the dynamic changes in the stress and weight distribution of the blade under different conditions by extracting the time series characteristics of each section of the blade. This method effectively solves the key problem of how to accurately calibrate the blade under variable working conditions. Combined with the multi-dimensional comparison of the standard model, the present invention can also accurately adjust the calibration results of the blade, and ensure that the physical properties of the blade are more in line with the actual working conditions through parametric adjustment. This data-driven calibration method can effectively improve the working stability of the blade, extend its service life, and reduce the failure rate and maintenance costs during operation, ultimately ensuring the safe operation and efficient power generation of the wind turbine. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0056] Figure 1 A schematic diagram of an embodiment of a calibration test method for a fan blade according to an embodiment of the present invention;

[0057] Figure 2 A schematic diagram of an embodiment of a calibration test device for a fan blade according to an embodiment of the present invention;

[0058] Figure 3 It is a schematic diagram of an embodiment of a calibration and testing device for wind turbine blades in an embodiment of the present invention.

[0059] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0062] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0063] An embodiment of the present application provides a calibration test method for a wind turbine blade. Figure 1 A flow chart of a calibration test method for a fan blade provided in an embodiment of the present application. The method comprises:

[0064] Please refer to Figure 1 , dividing the blade to be tested into a blade root section, a blade body section and a blade tip section along its length direction, wherein the blade root section is a fixed end connected to the hub;

[0065] In one embodiment of the present application, the blade to be measured is divided into a root section, a body section and a tip section along its length direction, including: measuring the total length L of the blade to be measured, and marking equally spaced measurement points on the surface of the blade to be measured along the length direction; performing strain tests at the measurement points to determine the stress distribution of the blade to be measured under different working conditions, defining the distance from the maximum stress point to the root end as L1, and dividing the area into the root section; performing aerodynamic characteristic tests at the measurement points to determine the critical position of the change in aerodynamic characteristics, defining the distance from the point to the blade tip as L3, and dividing the area into the tip section; dividing the area between the root section and the tip section into a body section, whose length is L2=L-L1-L3; and adjusting the ratio of L1, L2 and L3 according to the specific model and working environment of the blade to be measured.

[0066] Specifically, it is first necessary to measure the total length L of the blade to be tested. This length L is the full length from the root end of the blade (i.e. the fixed end connected to the wind turbine hub) to the tip of the blade. By accurately measuring this length, a basic reference framework can be provided for subsequent blade segmentation and testing. After measuring the total length, it is necessary to mark multiple equally spaced measurement points on the surface of the blade along its length. These measurement points will serve as the specific locations for subsequent strain and aerodynamic characteristics tests. The purpose of equally spaced markings is to ensure that a uniform test area can be covered along the entire length of the blade, so that the data collection is comprehensive and representative.

[0067] Next, strain tests are performed around these measurement points. The strain tests are achieved through strain sensors arranged at the measurement points. These sensors can capture the strain of the blades in real time during the stress process under different working conditions. The working conditions include changes in different wind speeds, rotation speeds and other working conditions, which will directly affect the stress distribution of the blades. After the strain tests at each measurement point are completed, the stress distribution diagram of the blade along the length direction under different conditions can be obtained. By analyzing the distribution diagram, the maximum stress point of the blade can be found. The maximum stress point refers to the part of the blade that is subjected to the greatest stress under specific working conditions. This part is usually located near the root section of the blade because the blade is subjected to the greatest force at the root section. The distance from this maximum stress point to the root end of the blade is defined as L1, so that the range of the root section can be determined by this L1.

[0068] The definition and division of the blade root section is of great significance. The blade root section is the part where the blade is directly connected to the hub and is subjected to the strongest torque and vibration when the wind turbine is running. By dividing the maximum stress area of ​​the blade into the blade root section, the physical state of this area can be more effectively monitored and evaluated, so that more targeted calibration and adjustment can be performed. In addition, such a division also helps to distinguish the functional characteristics of different blade segments, which is convenient for segmented testing and calibration in subsequent steps.

[0069] After determining the root section, the next step is to test the aerodynamic characteristics of each measuring point. The aerodynamic characteristics test is mainly to evaluate the aerodynamic performance of the blade surface under the action of wind flow. Through the test, the flow state and changes of the airflow at different positions of the blade can be obtained, especially the aerodynamic characteristics of the blade at a certain critical position will change significantly. This critical position is usually located near the tip section, because the shape and aerodynamic design of the blade cause the aerodynamic characteristics at the tip to be significantly different from those of other parts of the blade. After finding the critical point of this aerodynamic characteristic change, the distance from this point to the tip is defined as L3, so that the range of the tip section can be determined by L3. The tip section is of great significance in aerodynamic performance, because the tip is the part of the fan blade that moves at high speed, and its aerodynamic performance directly affects the overall efficiency and stability of the fan. Therefore, dividing the tip section separately and conducting corresponding tests and calibrations can effectively improve the overall performance of the blade.

[0070] After the root section and the tip section are divided out, the remaining middle section of the blade is the blade body section, and its length L2 is calculated by the formula L2=L-L1-L3. In this formula, L represents the total length of the blade, L1 is the length of the root section, and L3 is the length of the tip section. Through this simple calculation, the length L2 of the middle section of the blade can be obtained, thus completing the segmentation of the blade. The blade body section is the main part of the blade, and is mainly responsible for the balance between aerodynamics and structural strength, so its calibration and adjustment are also very critical.

[0071] Finally, the ratio of L1, L2 and L3 is adjusted according to the specific model of the blade and the working environment. Different wind turbine blades differ in size, material and design, so in actual operation, corresponding adjustments need to be made according to the specific model of each blade. The working environment of the wind turbine blades (such as the wind speed and climatic conditions of the wind farm) will also affect the stress distribution and aerodynamic characteristics of the blades. Therefore, the ratio of L1, L2 and L3 cannot remain unchanged, but must be flexibly adjusted according to actual conditions. Such a flexible adjustment mechanism ensures that this method can be applied to different types of wind turbine blades and can achieve precise calibration under a variety of working conditions.

[0072] It is not difficult to understand that by clearly dividing the different functional sections of the blade, differentiated testing and calibration can be performed on different blade areas. This method not only improves the comprehensiveness of the test, but also can more accurately evaluate the stress and aerodynamic characteristics of the blade under various working conditions, thereby achieving dynamic and personalized blade calibration. Compared with the traditional overall calibration method, this segmented calibration method significantly improves the accuracy of blade calibration and effectively improves the operating stability and service life of the wind turbine blades.

[0073] Continue to refer Figure 1, performing multi-modal dynamic scanning on the blade to be tested to obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds;

[0074] In one embodiment of the present application, the multi-modal dynamic scanning of the blade to be tested to obtain the acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speed conditions includes: setting a plurality of groups of rotation speed and wind speed combination working conditions in a wind tunnel, and performing rotation tests on the blade root section, blade body section and blade tip section of the blade to be tested respectively, to obtain the original response data of each section under different working conditions; using an acoustic sensor array to perform a full-range acoustic scanning on the blade root section, blade body section and blade tip section of the blade to be tested respectively, to obtain the acoustic response signal of each section, and performing time-frequency analysis on the acoustic response signal of each section to obtain the acoustic response signal of the blade root section, blade body section and blade tip section. The method comprises the following steps: first, capturing the dynamic deformation process of the blade root section, blade body section and blade tip section under different working conditions, and extracting the contour change information of each blade section through image processing algorithm, and obtaining the deformation characteristic matrix of the blade root section, blade body section and blade tip section; second, fusion of the acoustic characteristic spectrum, vibration characteristic vector and deformation characteristic matrix of the blade root section, blade body section and blade tip section is performed, and the comprehensive response data set of each section is constructed.

[0075] Specifically, it is first necessary to simulate multiple groups of rotation speed and wind speed combinations through wind tunnel experiments. These combinations represent different wind environments of blades in actual operation. By setting different wind speeds and rotation speeds, the operation state of blades under different conditions such as high speed, medium speed and low speed can be simulated. The purpose of the wind tunnel experiment is to provide a controllable wind environment for the blades, so that the root section, body section and tip section of the blades are tested under various conditions to obtain their response data in the actual environment. In the wind tunnel, the wind speed is generated by the fan provided by the wind tunnel equipment, and the rotation speed is controlled by the electric motor connected to the blades. The adjustment of wind speed and rotation speed can be accurately set and controlled by a computer. For example, if a working condition is set to a wind speed of 10 m / s and a rotation speed of 15 rpm, the system will automatically adjust the wind tunnel equipment to generate the wind flow under this working condition, and the response data of the blades can be collected under this working condition.

[0076] Under these conditions, the root section, the body section and the tip section are rotated and tested respectively. The key to the rotation test is to accurately capture the physical response of each section of the blade under different conditions. To achieve this goal, a number of high-precision sensors, including strain sensors, vibration sensors and displacement sensors, are first arranged on the surface of the blade. These sensors will be arranged at key positions of the root section, the body section and the tip section to collect physical quantities such as force, vibration and deformation in real time during the rotation of the blade. For example, strain sensors can be arranged in the high stress area of ​​the root of the blade to measure the deformation caused by the force of the blade during rotation. Vibration sensors can be arranged in the middle section of the blade to capture changes in its vibration frequency and amplitude. These sensors are connected to the data acquisition system via wired or wireless means. The data acquisition system is usually composed of a high-precision data acquisition card (DAQ) that can collect sensor feedback data in real time under high-speed rotation.

[0077] Next, the acoustic sensor array is used to perform an all-round acoustic scan of the blade. The acoustic sensor array is a network composed of multiple acoustic sensors, which can be arranged in various corners of the wind tunnel to form an all-round coverage of the blade. During the rotation of the blade, the air flow in the wind tunnel and the vibration of the blade will generate sound waves, which carry the physical response information of the blade under the action of wind. The acoustic sensor will capture these sound waves and convert them into electrical signals. Acoustic sensors are usually highly sensitive and can detect tiny changes in sound pressure. In order to obtain a more comprehensive acoustic response signal, the acoustic sensor array usually includes multiple microphones, each located at a different angle to ensure that the sound waves emitted by the blade at different positions can be captured. By performing time-frequency analysis on the sound wave signal, the acoustic characteristic spectra of the root section, the body section and the tip section can be obtained. Time-frequency analysis is an analysis method that combines the frequency information of the signal with the time information. It can help identify the change of the vibration frequency of the blade over time and provide the spectrum distribution of each blade section at different time points. By calculating the acoustic characteristic spectrum, the vibration intensity and frequency characteristics of each blade section under different working conditions can be further analyzed.

[0078] At the same time, the fiber Bragg grating sensor network arranged on the surface of the blade is used to collect vibration and strain data of each section of the blade. The fiber Bragg grating sensor is a high-precision sensing device that can measure the strain and vibration of an object by modulating an optical signal. The fiber Bragg grating sensor network is usually composed of a series of optical fiber sensors, each of which is embedded in a key position of the blade and measures the strain and vibration data of the blade by monitoring the change of the light wavelength inside the optical fiber. The change of the light wavelength is proportional to the vibration amplitude and strain value of the blade. Therefore, by decoding these optical signals, the strain data and vibration data of each section of the blade can be obtained. In an environment where the blade rotates at high speed, the fiber Bragg grating sensor can provide real-time, high-resolution strain and vibration information. These data are then processed by modal analysis technology. Modal analysis is a method of analyzing structural vibration modes. It can identify the vibration modes of the blade at different frequencies and generate the vibration characteristic vector of the blade. The vibration characteristic vector is a set of numerical values ​​that describe the vibration state of the blade. It contains information such as the vibration frequency and vibration amplitude of each section of the blade, and can reflect the dynamic behavior of the blade under different working conditions.

[0079] In addition, capturing the dynamic deformation of the blade during rotation through a high-speed camera system is an important means of obtaining deformation data. The high-speed camera system can record the tiny deformation of the blade surface and generate a continuous sequence of images. The frame rate of the camera system is usually as high as thousands of frames per second, which can accurately record the deformation state of the blade at every moment during the rapid rotation process. Through the image processing algorithm, these images are analyzed to extract the contour change information of the blade. The image processing algorithm performs edge detection on each frame of the image and tracks the position changes of the feature points on the blade surface to generate the deformation feature matrix of the blade. The deformation feature matrix is ​​a set of data matrices that reflect the changes in the shape of the blade surface. It contains information such as the contour changes and bending degree of the blade under different working conditions, and can fully describe the dynamic deformation process of the blade.

[0080] Finally, these data are fused to construct a comprehensive response data set for the blade root section, blade body section, and blade tip section. The data fusion process combines the acoustic characteristic spectrum, vibration characteristic vector, and deformation characteristic matrix to form a comprehensive data set to more accurately evaluate the dynamic behavior of the blade. Data fusion technology ensures the accuracy and consistency of each data point by cross-validating data from different sources. For example, acoustic data can verify the accuracy of vibration data, while deformation data can further supplement the dynamic change information of vibration and strain. Through this data fusion technology, a highly consistent data set that fully reflects the dynamic response of the blade can be constructed. This data set provides a reliable basis for subsequent calibration and adjustment, ensuring that the performance and stability of the blade under different working conditions are fully guaranteed.

[0081] Continue to refer Figure 1, using a deep learning network to perform time series analysis on the response data set to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, blade body section, and blade tip section;

[0082] In one embodiment of the present application, the use of a deep learning network to perform a time series analysis on the response data set to generate a dynamic weight distribution diagram and a stress distribution diagram of the blade root section, the blade body section and the blade tip section includes: performing a time series decomposition on the response data sets of the blade root section, the blade body section and the blade tip section to obtain a trend item, a period item and a random item of each section; inputting the trend items of the blade root section, the blade body section and the blade tip section into a pre-trained three-dimensional convolutional neural network to extract the spatial features of each section to obtain a preliminary weight distribution feature diagram of each section; and performing a time series decomposition on the response data sets of the blade root section, the blade body section and the blade tip section to obtain a trend item, a period item and a random item of each section. The periodic terms and random terms of the blade body section and the blade tip section are respectively applied with long short-term memory networks to capture the temporal dependencies of each section and obtain the time-varying feature vectors of each section; the preliminary weight distribution feature maps of the blade root section, blade body section and blade tip section are feature fused with the corresponding time-varying feature vectors, and the three-dimensional spatial distribution of each section is reconstructed through a deconvolution network to obtain the dynamic weight distribution map of each section; the dynamic weight distribution maps of the blade root section, blade body section and blade tip section are processed as a whole using a graph neural network, and the connection relationship between the blade sections is considered to generate a stress distribution map of the complete blade.

[0083] Specifically, it is first necessary to perform time series decomposition on the response data sets of the blade root section, blade body section, and blade tip section. The purpose of this process is to decompose the complex raw data into different components that are easier to understand and process. The response data set usually contains the dynamic response data of each section of the blade collected by multimodal sensors (such as strain sensors, vibration sensors, and acoustic sensors). These data are recorded over time and include physical parameters such as stress and vibration of the blade under different rotation speeds and wind speeds. In order to analyze these complex time series data, they first need to be decomposed.

[0084] Time series decomposition is to decompose the response data into trend terms, periodic terms, and random terms through algorithms such as empirical mode decomposition (EMD) or wavelet transform. For example, assume that the stress of the root section of the blade shows a certain growth trend over time, accompanied by periodic vibration fluctuations and some noise effects. Through decomposition, the trend term will capture the overall growth trend of the stress, such as the gradual increase in stress under long-term high wind speeds; the periodic term captures the periodic changes in stress, reflecting the vibration frequency and amplitude of the blade; the random term captures those random fluctuations that cannot be predicted in the short term, such as sudden stress changes caused by microscopic changes in materials or instantaneous load fluctuations. These decomposed data are easier to process because they reflect the dynamic characteristics of different levels of the blade.

[0085] Next, the trend items of the root, body, and tip segments are input into a pre-trained three-dimensional convolutional neural network (3D CNN). A 3D convolutional neural network is a deep learning model that can extract features from spatial data. To achieve this operation, the decomposed trend items need to be organized into three-dimensional input data, which corresponds to the stress or weight distribution at different locations of the blade. The convolution operation generates a feature map by sliding a fixed-size window (called a convolution kernel) on the spatial data of the blade and calculating the data features within the window at each slide. For example, assuming that the blade shows a large stress concentration in the root segment and a small stress in the tip segment in the trend item, the 3D convolutional neural network can capture the stress concentration in different areas of the blade through continuous convolution operations and generate a preliminary weight distribution feature map for each segment of the blade. This feature map reflects the weight distribution of various parts of the blade, especially in areas where the blade is subjected to greater stress. The convolution operation will enhance these features, making the features of these areas more prominent in the feature map.

[0086] For the periodic terms and random terms, the long short-term memory network (LSTM) is used for processing respectively. LSTM is a special recurrent neural network that can process data with time dependencies. The data of the periodic terms and random terms reflect the dynamic changes of the blades at different time points. In particular, the periodic terms reflect the periodic vibration behavior of the blades when the wind speed or rotation speed changes. For example, the data in the periodic terms may show that the root section of the blade vibrates at a certain frequency under high wind speed, while the vibration frequency of the tip section is higher. LSTM can capture these temporal dependencies through its internal memory units and generate time-varying feature vectors. These time-varying feature vectors reflect the periodic behavior of each section of the blade over time, such as how the vibration frequency of a certain section of the blade changes over time. The data of the random terms may reflect some short-term non-periodic fluctuations. After being processed by LSTM, some irrelevant noise information can be filtered out, and only important time series features are retained.

[0087] After obtaining the preliminary weight distribution feature map and the time-varying feature vector, the next step is to perform feature fusion on the two. The purpose of feature fusion is to combine spatial features and temporal features so as to fully reflect the dynamic changes of the blade at different time and spatial positions. Feature fusion can be accomplished through weighted averaging or other complex fusion algorithms. For example, suppose that the blade root segment appears as a weight concentration area in the preliminary weight distribution map, and its time-varying feature vector shows that the area has a stress concentration problem under high-frequency vibration. Through feature fusion, this information can be combined to provide a more comprehensive perspective, showing the overall dynamic behavior of the blade at this position.

[0088] After feature fusion, the next step is to reconstruct the three-dimensional spatial distribution of each blade segment through a deconvolution network. The deconvolution network is the opposite of the convolution network and aims to reconstruct a high-resolution spatial distribution from a lower-resolution feature map. Through the deconvolution operation, the preliminary weight distribution feature map can be enlarged and restored to the three-dimensional spatial distribution of the blade. For example, through layer-by-layer upsampling, the deconvolution network can gradually transform the low-dimensional preliminary weight distribution feature map into a high-dimensional three-dimensional distribution map, which represents the weight distribution of each small area of ​​the blade under different working conditions. This step is very important because it provides a complete three-dimensional model that can accurately describe the dynamic weight distribution of the blade at different positions.

[0089] Finally, the dynamic weight distribution map of the blade is processed as a whole using a graph neural network (GNN). Graph neural networks are very suitable for processing systems with complex connections. The blade segments are not isolated, but are closely connected through the overall structure of the blade. Graph neural networks can model these interconnected relationships, taking into account the mechanical transfer effects between the blade segments. For example, the weight change of the root segment may affect the stress distribution of the blade body segment and the tip segment. The graph neural network can comprehensively consider the interactions between these areas through the structure of nodes and edges. Through this process, the graph neural network can generate a stress distribution map for the entire blade. This stress distribution map reflects the stress concentration areas of the blade under different working conditions and predicts which areas may have structural problems due to excessive stress.

[0090] The whole process can accurately reconstruct the dynamic weight and stress distribution of wind turbine blades under different working conditions through multi-step processing of time series decomposition, convolutional neural network, LSTM, deconvolution network and graph neural network. Each step is carefully designed to ensure that the spatial and temporal characteristics of the blades can be fully captured and processed. This method can not only obtain the dynamic weight distribution of the blades, but also provide a basis for the physical calibration of the blades, ensuring the performance stability and safety of the blades during operation.

[0091] Continue to refer Figure 1 , compare the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database in multiple dimensions, select the best matching standard model, and make parameterized adjustments according to the actual differences;

[0092] In one embodiment of the present application, the dynamic weight distribution diagram and the stress distribution diagram are compared with the standard models in a pre-established blade database in multiple dimensions, the best matching standard model is selected, and parameterized adjustment is performed according to actual differences, including: feature extraction of the dynamic weight distribution diagram and the stress distribution diagram of the blade root section, the blade body section and the blade tip section to obtain a multi-dimensional feature vector of the blade to be tested; using a multi-objective optimization algorithm to calculate the similarity between the multi-dimensional feature vector and the standard model in the blade database to obtain a set of candidate matching models; applying the hierarchical analysis method to the candidate matching model set to comprehensively consider weight distribution, stress distribution and geometric shape factors to obtain the best matching standard model; constructing a parameter mapping relationship between the blade to be tested and the best matching standard model to generate a parameterized adjustment matrix; and optimizing the parameterized adjustment matrix using a genetic algorithm to obtain an optimal parameter adjustment scheme.

[0093] Specifically, it is first necessary to extract features from the dynamic weight distribution map and stress distribution map of the blade root section, blade body section and blade tip section. These distribution maps are generated by multimodal sensor data fusion, reflecting the stress concentration area and weight distribution of the blade under different working conditions. The process of feature extraction is to convert these complex distribution map information into mathematical representations that are easy to calculate and analyze, namely multidimensional feature vectors. Specifically, feature extraction is to process the weight distribution map and stress distribution map of the blade to extract key features that can describe the state of the blade. For example, a common method of feature extraction is to partition the distribution map, divide the blade surface into multiple small areas, and then calculate the average stress and weight values ​​of each area to form a feature vector. Each element of this feature vector represents the stress or weight distribution information of a certain area of ​​the blade, thereby converting the two-dimensional distribution map into a high-dimensional mathematical representation.

[0094] After obtaining the multi-dimensional feature vectors of the blade root section, blade body section and blade tip section, the next step is to perform similarity calculation. The purpose of similarity calculation is to compare the multi-dimensional feature vector of the blade to be tested with the standard model in the blade database to find the closest matching model. To this end, a multi-objective optimization algorithm is used to achieve accurate calculation of similarity. The core of the multi-objective optimization algorithm is to consider multiple different objectives at the same time for optimization. In the application here, multiple objectives such as weight distribution, stress distribution and geometric shape need to be considered at the same time. The objective function of multi-objective optimization can be set as the weighted sum of each objective. For example, a similarity function can be defined:

[0095] .

[0096] in, is the multi-dimensional feature vector of the blade to be tested, is the multidimensional feature vector of the standard model, Function represents the similarity calculation function, and They represent the weight distribution characteristics of the blades, and They represent the stress distribution characteristics, and Represent the geometric shape characteristics, , , is the weight coefficient of each target.

[0097] The function of this formula is to combine the similarities of each target by weighted summation to obtain the final similarity value. In this way, the candidate matching model set closest to the blade to be tested can be found.

[0098] After obtaining the set of candidate matching models, the next step is to apply the analytic hierarchy process to further screen the candidate matching models. The analytic hierarchy process is a decision support tool that can compare multiple evaluation criteria in layers and make a comprehensive evaluation based on the importance of each criterion. For the calibration of wind turbine blades, factors such as weight distribution, stress distribution and geometric shape need to be considered comprehensively, and the weights of these factors can be adjusted according to actual conditions. For example, if the aerodynamic characteristics of the blades in a certain wind farm have a greater impact on the performance of the blades, the geometric shape can be given a higher weight, which can be determined by expert scoring or data analysis. The application of the analytic hierarchy process can ensure that the most matching standard model that best meets the actual working conditions is selected from multiple candidate models. Finally, the analytic hierarchy process will select the optimal standard model based on the comprehensive score of each candidate model.

[0099] After determining the best matching standard model, the next step is to construct a parameter mapping relationship between the blade to be tested and the best matching standard model, and generate a parameterized adjustment matrix. The construction of the parameter mapping relationship is based on the differences in weight distribution and stress distribution between the blade to be tested and the standard model. In order to achieve this process, it is first necessary to compare the various features of the blade to be tested and the standard model, and calculate the difference values ​​between the two in various areas. For example, the weight difference between the blade to be tested and the standard model in a certain area, as well as the size difference of the stress concentration area, can be calculated. These difference values ​​are used to generate a parameterized adjustment matrix. Each element in the adjustment matrix represents the calibration adjustment required for a certain part of the blade. For example, in a certain stress concentration area, it may be necessary to increase or decrease material to adjust the weight distribution. The role of the parameterized adjustment matrix is ​​to guide the subsequent calibration process to ensure that the adjustment can accurately correspond to the specific differences in various areas of the blade.

[0100] After the parameterized adjustment matrix is ​​generated, the adjustment scheme needs to be optimized using a genetic algorithm. Genetic algorithm is an optimization algorithm based on biological evolution theory. It mainly searches for the optimal solution through selection, crossover, mutation and other operations. For the calibration problem of wind turbine blades, the application of genetic algorithms can effectively explore the optimal solution for parameter adjustment. First, each set of possible adjustment schemes of the parameterized adjustment matrix is ​​used as the initial population of the genetic algorithm, and then the adjustment scheme with the highest fitness is selected through iterative optimization of the genetic algorithm. The fitness function can be defined as the similarity between the adjusted blade and the standard model in weight distribution and stress distribution. The optimization goal is to make the adjusted blade as similar as possible to the standard model in all aspects. Through multiple iterations and crossover mutation operations, the genetic algorithm can eventually find a globally optimal parameter adjustment scheme to ensure that the calibration result of the blade achieves the best effect.

[0101] The whole process can accurately solve the calibration problem of wind turbine blades under different working conditions through the combination of feature extraction, multi-objective optimization, hierarchical analysis, parameter adjustment and genetic algorithm. Each step is precisely designed and optimized to ensure that the final parameter adjustment solution can meet the operating requirements of the blades and effectively improve the operating stability and life of the wind turbine blades.

[0102] Continue to refer Figure 1 Based on the adjusted model, physical calibration guides for the blade root section, the blade body section and the blade tip section are generated, and the section weight adjustment is performed on the blade to be tested.

[0103] In one embodiment of the present application, the physical calibration guide for the blade root section, blade body section and blade tip section is generated based on the adjusted model, and the segmented weight adjustment of the blade to be tested is performed, including: performing finite element analysis on the adjusted model, calculating the ideal weight distribution of the blade root section, blade body section and blade tip section, and obtaining a target weight distribution diagram of each section; performing difference calculation between the target weight distribution diagram and the current weight distribution diagram of the blade to be tested, and generating a weight adjustment amount distribution diagram of the blade root section, blade body section and blade tip section; according to the weight adjustment amount distribution diagram, using a topology optimization algorithm to calculate the weight adjustment point position and adjustment amount in each section, and generating a detailed physical calibration guide; and injecting or removing fixed-point weight material from the blade root section, blade body section and blade tip section according to the physical calibration guide.

[0104] Specifically, the adjusted model first needs to use finite element analysis to calculate the ideal weight distribution of each section of the blade. By dividing the blade into multiple small finite elements, the mechanical behavior of each unit under the action of external forces is analyzed, and then the force and weight distribution of the entire blade are calculated. In order to perform this analysis, the geometric model and material parameters of the blade need to be input into the finite element analysis software. The geometric model includes the three-dimensional shape of the root section, the body section and the tip section, while the material parameters include the density and elastic modulus of the composite material used in the blade. These input data can be obtained through previous experimental measurements or manufacturing data.

[0105] In finite element analysis, the blade is divided into multiple small unit grids, and each unit is regarded as an independent calculation unit. For each unit, the finite element analysis will calculate its stress and deformation according to the operating conditions of the blade (such as wind speed, rotation speed, etc.). By summarizing the mechanical behavior of all units, the overall stress distribution and weight distribution of the blade can be calculated. This process can obtain the ideal weight distribution diagram of the root section, blade body section and blade tip section, that is, the optimal weight distribution that each section of the blade should have under specific working conditions. The ideal weight distribution diagram is usually expressed as a three-dimensional surface graph, and each point represents the ideal weight value at a certain position on the blade.

[0106] Next, the ideal weight distribution diagram obtained by finite element analysis is calculated by difference with the current weight distribution diagram of the blade to be tested. The current weight distribution diagram of the blade to be tested is obtained through sensor measurement or previous experimental data, which reflects the weight distribution of the blade in actual operation. The purpose of the difference calculation is to determine the deviation between the current state of the blade and the ideal state. The formula can be expressed as:

[0107] .

[0108] in, Indicates the weight adjustment of a certain position of the blade, is the value of the ideal weight distribution diagram, is the value of the current weight distribution graph.

[0109] This formula can be used to generate a weight adjustment distribution map for the root section, blade body section and blade tip section. This distribution map shows the amount of weight that needs to be added or subtracted at each location to make the actual weight distribution of the blade as close to the ideal state as possible.

[0110] After obtaining the weight adjustment distribution diagram, the next step is to use the topology optimization algorithm to calculate the weight adjustment point position and adjustment amount in each section. The topology optimization algorithm is an optimization design method that finds the optimal solution for material distribution by modifying the topology of the structure. In the present invention, the goal of topology optimization is to find the position and corresponding adjustment amount where weight adjustment should be performed in each section of the blade. The specific optimization objective function can be set as:

[0111] .

[0112] in, The optimization goal is to make the adjusted weight distribution as close to the ideal weight distribution as possible.

[0113] The calculation process of topology optimization first needs to determine some candidate adjustment points, which can be selected based on the extreme points in the weight adjustment distribution diagram. For example, if the weight adjustment at a certain position is positive (indicating that the weight needs to be increased), then material can be added at that position; conversely, if the adjustment is negative (indicating that the weight needs to be reduced), then material removal is required at that position.

[0114] A detailed physical calibration guide is generated through the calculation of the topology optimization algorithm. The physical calibration guide describes in detail the location of the adjustment points within each blade segment and the corresponding adjustment amount. For example, the guide may indicate that 5 grams of composite material needs to be added to a certain location in the root segment, while 3 grams of material needs to be removed from a certain location in the tip segment. The calibration guide may also include recommendations on how to implement these adjustments, such as using specific material injection tools or material removal equipment. The physical calibration guide ensures that subsequent calibration operations can be carried out strictly in accordance with the calculated results, thereby ensuring that the weight distribution of the blade is ideal.

[0115] According to the physical calibration guidelines, the next step is to perform targeted weight material injection or removal on the root section, blade body section and blade tip section. In actual operation, material injection and removal can be achieved in many ways. For the parts that need to increase weight, high-precision material injection equipment can be used to inject a specific amount of composite material into the structure of the blade. Material injection usually uses composite materials that are the same or similar to the raw materials of the blade to ensure that the overall mechanical properties of the blade are not affected. For example, the injection equipment can be controlled by a computer to accurately inject 5 grams of material into a specific position of the blade. For the parts that need to reduce weight, this can be achieved through material removal equipment. The material removal equipment can be a laser cutting machine or a mechanical grinding equipment, which is used to accurately remove excess material from the blade surface. The entire material adjustment process needs to be carried out under high-precision control to ensure that the adjustment amount at each position meets the requirements of the physical calibration guidelines.

[0116] After the material injection and removal is complete, the adjusted blade can also be rechecked. The purpose of the recheck is to ensure that all adjustments have been made according to the calibration guidelines and that the actual weight distribution of the blade is close to the ideal state. The recheck can be done by measuring the weight distribution of the blade again with sensors or by recalculating the performance of the adjusted blade using finite element analysis to verify that the calibration effect is as expected.

[0117] Through the above steps, accurate calibration of wind turbine blades can be effectively achieved. Each step in this solution has been rigorously calculated and optimized to ensure that the weight distribution and stress distribution of the blades are ideal. Finite element analysis provides accurate mechanical analysis results, differential calculation helps identify the specific needs of weight adjustment, topology optimization provides the optimal material adjustment solution, and the physical calibration guide and actual material adjustment operations ensure the feasibility and efficiency of the calibration work. The entire calibration process ensures the stability and service life of the wind turbine blades during operation, and is especially suitable for precise adjustment of blades under complex working conditions.

[0118] In one embodiment of the present application, after injecting or removing fixed-point weight material into or from the blade root section, blade body section and blade tip section according to the physical calibration guide, it also includes: quickly scanning the blade to be tested after weight adjustment to obtain the actual weight distribution after adjustment; comparing the actual weight distribution after adjustment with the target weight distribution map to obtain the weight distribution deviation value; judging whether the weight distribution deviation value exceeds a preset threshold value, and if so, returning to the step of "calculating the weight adjustment point position and adjustment amount in each section according to the weight adjustment amount distribution map using a topology optimization algorithm to generate a detailed physical calibration guide" to regenerate a locally fine-tuned physical calibration guide; and adjusting the weight of the blade to be tested again according to the locally fine-tuned physical calibration guide.

[0119] Specifically, the blade to be tested after weight adjustment first needs to be quickly scanned to obtain the actual weight distribution after adjustment. The purpose of the quick scan is to accurately capture the weight distribution of the blade after calibration and adjustment in a short time through non-contact measurement technology. Commonly used scanning equipment includes laser scanners or high-resolution optical scanning systems, which can obtain weight distribution information at various positions on the blade surface through precise optical signal reflection. During the specific operation process, the laser scanner calculates the weight distribution of each area of ​​the blade by scanning the laser beam reflected by the blade surface and using the time difference and intensity change of the echo. The scanning results are stored in the form of a three-dimensional point cloud. Each point in the point cloud corresponds to the weight value of a specific position on the blade surface, forming an actual weight distribution map after adjustment.

[0120] After obtaining the adjusted actual weight distribution diagram, the next step is to compare it with the pre-calculated target weight distribution diagram. The target weight distribution diagram is the ideal weight distribution state obtained in the early stage through finite element analysis and other means, while the adjusted actual weight distribution diagram reflects the current state of the blade after the material addition and reduction operation. In order to evaluate the difference between the two, it is necessary to calculate the weight distribution deviation value. The specific calculation formula is:

[0121] .

[0122] in, Indicates the actual weight value after adjustment at a certain position of the blade With target weight value The difference between.

[0123] This deviation value can be accumulated by calculating all the measurement points to form an overall weight deviation distribution diagram, showing the calibration effect of each area of ​​the blade after the initial adjustment. Usually, the calculation of the deviation value is not limited to single point comparison, but also combined with the structural characteristics of the entire blade to perform a global error analysis.

[0124] After obtaining the deviation value distribution map, the next step is to determine whether these weight distribution deviation values ​​exceed the preset calibration threshold. The calibration threshold is a pre-set value used to determine whether the calibration effect reaches an acceptable range. For example, the threshold can be defined as a certain percentage. If the absolute value of the deviation value of a certain area exceeds this percentage, it means that the calibration of this area is still not accurate enough and further adjustments are required. During the judgment process, if the deviation values ​​of all areas are within the calibration threshold range, it can be considered that the current calibration has met the requirements; conversely, if the deviation values ​​of some areas exceed the threshold, it means that the weight adjustment of these areas is not ideal enough and it is necessary to return to the previous step and make local adjustments again.

[0125] If the weight distribution deviation exceeds the preset threshold, local fine-tuning needs to be performed again according to the deviation. At this point, the system will return to the previous step and recalculate the weight adjustment point position and adjustment amount of each section according to the weight adjustment amount distribution map. In this process, the topology optimization algorithm is still used. This algorithm can redefine the optimization target based on the latest deviation value distribution map and determine the material increase and decrease operations in areas that require further calibration. For example, in a blade area where the deviation exceeds the standard, the topology optimization algorithm will re-evaluate the mechanical properties of the area and calculate the best material adjustment plan. This local adjustment can more accurately control the weight distribution of the blade, thereby further reducing the deviation.

[0126] After the topology optimization calculation is completed, a new physical calibration guide is generated. The physical calibration guide is a detailed set of steps to guide the operator on how to make further weight adjustments to the blade. The specific steps include determining where to add or reduce material, how much to adjust, and what tools and techniques to use to make the adjustments. For example, the guide may indicate adding 2 grams of material to a certain point in the blade body section, or removing 3 grams of material from a certain point in the blade tip section. The detailed nature of this guide ensures that the operator can make adjustments strictly in accordance with the optimization results.

[0127] According to the physical calibration guide, the operator will once again make a fixed-point weight adjustment on the blade. This process is similar to the first weight adjustment, and the weight distribution of the blade is still changed by injecting or removing materials. For areas where weight needs to be increased, material injection can use high-precision injection equipment to achieve weight adjustment by injecting a small amount of composite materials. For areas where weight needs to be reduced, material removal can be done by laser cutting or mechanical grinding. This operation ensures that the weight adjustment of each area can be implemented accurately, and ultimately achieves an effect consistent with the ideal state.

[0128] After completing local fine-tuning, further re-inspection is required to verify whether all adjustments have met expectations. The re-inspection method is also to use rapid scanning technology to obtain the actual weight distribution map of the blade again and compare it with the target weight distribution map. At this time, if the weight deviation values ​​of all areas are within the preset threshold range, it means that the calibration of the blade has been completed; if there are still some areas that exceed the threshold, repeat the above fine-tuning steps until all deviation values ​​meet the requirements.

[0129] This solution is highly repeatable and accurate. Through multiple iterations of rapid scanning technology, weight distribution deviation calculation, topology optimization algorithm and physical calibration guide, the ideal weight distribution of wind turbine blades can be ensured. The advantage of this method is that it can make local adjustments based on the specific conditions of the blades, avoiding major global modifications and improving the efficiency and accuracy of calibration. At the same time, the combination of advanced algorithms and precise material addition and reduction methods ensures the operational stability and long-term reliability of the blades. This complete process design provides strong technical support for the precise calibration of wind turbine blades.

[0130] The above describes the calibration test method of the fan blade in the embodiment of the present invention. The following describes the calibration test device of the fan blade in the embodiment of the present invention. Figure 2 , an embodiment of a calibration test device for a fan blade in an embodiment of the present invention comprises:

[0131] The blade segmentation module 201 is used to divide the blade to be tested into a blade root section, a blade body section and a blade tip section along its length direction, wherein the blade root section is a fixed end connected to the hub;

[0132] A multi-modal scanning module 202 is used to perform a multi-modal dynamic scanning on the blade to be tested, and obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds;

[0133] A deep learning analysis module 203 is used to perform time series analysis on the response data set using a deep learning network to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, the blade body section, and the blade tip section;

[0134] The model comparison processing module 204 is used to compare the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database in multiple dimensions, select the most matching standard model, and perform parameter adjustment according to the actual differences;

[0135] The calibration execution module 205 is used to generate physical calibration guidelines for the blade root section, the blade body section and the blade tip section based on the adjusted model, and perform segment weight adjustment on the blade to be tested.

[0136] above Figure 2 The calibration and testing device for wind turbine blades in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The calibration and testing device for wind turbine blades in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0137] Figure 3 1 is a schematic diagram of the structure of a calibration and testing device for a wind turbine blade provided by an embodiment of the present invention. The calibration and testing device 300 for a wind turbine blade may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 may be temporary storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the calibration and testing device 300 for a wind turbine blade. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the calibration and testing device 300 for a wind turbine blade to implement the steps of the calibration and testing method for a wind turbine blade described above.

[0138] The wind turbine blade calibration test device 300 may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The structure of the calibration and testing device for wind blades shown does not constitute a limitation on the calibration and testing device for wind blades provided by the present invention, and may include more or less components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0139] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the wind turbine blade calibration test method.

[0140] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0141] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0142] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A calibration test method for a fan blade, characterized in that: include: Divide the blade to be tested into a blade root section, a blade body section and a blade tip section along its length direction, wherein the blade root section is a fixed end connected to the hub; Performing multi-modal dynamic scanning on the blade to be tested to obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds; Performing time series analysis on the response data set using a deep learning network to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, the blade body section, and the blade tip section; Perform multi-dimensional comparison of the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database, select the most matching standard model, and perform parameter adjustment according to the actual differences; Based on the adjusted model, physical calibration guides for the blade root section, the blade body section and the blade tip section are generated, and the section weight adjustment is performed on the blade to be tested.

2. The calibration test method for wind turbine blades according to claim 1, characterized in that: The method of dividing the blade to be measured into a blade root section, a blade body section and a blade tip section along its length direction comprises: Measuring the total length L of the blade to be measured, and marking equally spaced measuring points on the surface of the blade to be measured along the length direction; Performing strain testing at the measuring point to determine the stress distribution of the blade to be tested under different working conditions, defining the distance from the maximum stress point to the blade root end as L1, and dividing the area corresponding to L1 into the blade root section; Perform an aerodynamic characteristic test at the measuring point to determine a critical position of aerodynamic characteristic change, define the distance from the critical point of aerodynamic characteristic change to the blade tip as L3, and divide the area corresponding to L3 into a blade tip section; The area between the blade root section and the blade tip section is divided into a blade body section, the length of which is L2=L-L1-L3; The ratio of L1, L2 and L3 is adjusted according to the specific model and working environment of the blade to be tested.

3. The calibration test method for wind turbine blades according to claim 1, characterized in that: The multi-modal dynamic scanning of the blade to be tested is performed to obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speed conditions, including: A plurality of working conditions of rotation speed and wind speed combination are set in the wind tunnel, and the root section, the body section and the tip section of the blade to be tested are rotated to obtain the original response data of each section under different working conditions; Using an acoustic sensor array, a blade root section, a blade body section, and a blade tip section of the blade to be tested are respectively subjected to an omnidirectional acoustic scan to obtain an acoustic response signal of each section, and a time-frequency analysis is performed on the acoustic response signal of each section to obtain an acoustic characteristic spectrum of the blade root section, the blade body section, and the blade tip section; The fiber grating sensor network is arranged on the surface of the blade root section, blade body section and blade tip section of the blade to be tested, and the vibration and strain data of each section are collected respectively. The vibration and strain data of each section are subjected to modal analysis to obtain the vibration characteristic vectors of the blade root section, blade body section and blade tip section. A high-speed camera system is used to capture the dynamic deformation process of the blade root section, blade body section and blade tip section under different working conditions. The profile change information of each blade section is extracted through image processing algorithm to obtain the deformation feature matrix of the blade root section, blade body section and blade tip section. The acoustic characteristic spectra, vibration characteristic vectors and deformation characteristic matrices of the blade root section, blade body section and blade tip section are fused respectively to construct a comprehensive response data set of each section.

4. The calibration test method for a fan blade according to claim 1, characterized in that: The using of a deep learning network to perform a time series analysis on the response data set to generate a dynamic weight distribution diagram and a stress distribution diagram of the blade root section, the blade body section and the blade tip section includes: Performing time series decomposition on the response data sets of the blade root section, the blade body section and the blade tip section respectively to obtain trend items, period items and random items of each section; The trend items of the blade root section, the blade body section and the blade tip section are respectively input into a pre-trained three-dimensional convolutional neural network, the spatial features of each section are extracted, and a preliminary weight distribution feature map of each section is obtained; Applying long short-term memory networks to the periodic terms and random terms of the blade root segment, the blade body segment and the blade tip segment respectively to capture the temporal dependency of each segment and obtain a time-varying feature vector of each segment; The preliminary weight distribution feature maps of the blade root section, the blade body section and the blade tip section are fused with the corresponding time-varying feature vectors, and the three-dimensional spatial distribution of each section is reconstructed through a deconvolution network to obtain a dynamic weight distribution map of each section; The dynamic weight distribution diagrams of the blade root section, blade body section and blade tip section are processed as a whole using a graph neural network, and the connection relationship between the blade sections is considered to generate a stress distribution diagram of the complete blade.

5. The calibration test method for a wind turbine blade according to claim 1, characterized in that: The dynamic weight distribution diagram and the stress distribution diagram are compared with the standard model in the pre-established blade database in multiple dimensions, the most matching standard model is selected, and parameterized adjustment is performed according to the actual differences, including: Extracting features from the dynamic weight distribution diagram and stress distribution diagram of the blade root section, the blade body section and the blade tip section to obtain a multi-dimensional feature vector of the blade to be tested; Using a multi-objective optimization algorithm, similarity calculation is performed between the multi-dimensional feature vector and a standard model in a leaf database to obtain a candidate matching model set; Applying the analytic hierarchy process to the candidate matching model set, taking into account weight distribution, stress distribution and geometric shape factors, to obtain the most matching standard model; Constructing a parameter mapping relationship between the blade to be tested and the most matching standard model to generate a parameterized adjustment matrix; The parameterized adjustment matrix is ​​optimized using a genetic algorithm to obtain an optimal parameter adjustment solution.

6. The calibration test method for wind turbine blades according to claim 1, characterized in that: The method generates physical calibration guides for the blade root section, the blade body section, and the blade tip section based on the adjusted model, and performs segmented weight adjustment on the blade to be tested, including: Performing finite element analysis on the adjusted model, calculating the ideal weight distribution of the blade root section, the blade body section and the blade tip section, and obtaining a target weight distribution diagram of each section; Performing difference calculation between the target weight distribution diagram and the current weight distribution diagram of the blade to be measured, so as to generate weight adjustment amount distribution diagrams of the blade root section, the blade body section and the blade tip section; According to the weight adjustment amount distribution map, a topology optimization algorithm is used to calculate the weight adjustment point locations and adjustment amounts in each segment, and a detailed physical calibration guide is generated; The blade root section, blade airfoil section and blade tip section are injected or removed with targeted weight material according to the physical calibration guide.

7. The calibration test method for wind turbine blades according to claim 6, characterized in that: After injecting or removing the fixed-point weight material into or from the blade root section, the blade body section, and the blade tip section according to the physical calibration guide, the method further includes: Perform a quick scan on the blade to be tested after weight adjustment to obtain the actual weight distribution after adjustment; Comparing the adjusted actual weight distribution with the target weight distribution diagram to obtain a weight distribution deviation value; Determine whether the weight distribution deviation value exceeds a preset threshold value. If so, return to step "calculate the weight adjustment point position and adjustment amount in each segment using a topology optimization algorithm according to the weight adjustment amount distribution map, and generate a detailed physical calibration guide" to regenerate a local fine-tuning physical calibration guide; According to the physical calibration guide of the local fine-tuning, the weight of the blade to be tested is adjusted again.

8. A calibration test device for fan blades, characterized in that: The calibration test device for the fan blades comprises: A blade segmentation module is used to divide the blade to be tested into a blade root segment, a blade body segment and a blade tip segment along its length direction, wherein the blade root segment is a fixed end connected to the hub; A multi-modal scanning module, used to perform multi-modal dynamic scanning on the blade to be tested, and obtain acoustic, vibration and deformation response data sets of the blade root section, blade body section and blade tip section under different rotation speeds and wind speeds; A deep learning analysis module, used for performing time series analysis on the response data set using a deep learning network to generate dynamic weight distribution diagrams and stress distribution diagrams of the blade root section, the blade body section, and the blade tip section; A model comparison processing module is used to compare the dynamic weight distribution diagram and the stress distribution diagram with the standard model in the pre-established blade database in multiple dimensions, select the most matching standard model, and perform parameterized adjustments according to actual differences; The calibration execution module is used to generate physical calibration guidelines for the blade root section, the blade body section and the blade tip section based on the adjusted model, and to perform segmented weight adjustment on the blade to be tested.

9. A calibration test device for a fan blade, characterized in that: The calibration and testing device for wind turbine blades comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the calibration and testing device for the wind blade to perform the steps of the calibration and testing method for the wind blade according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the wind turbine blade calibration test method as described in any one of claims 1-7 are implemented.

Images