Wind power gear box test system

Abstract

The invention discloses a wind power gear box test system, and relates to the technical field of wind power test, and the system comprises the following components: a digital twinborn model building unit, a virtual working condition simulation pre-test unit, a test scheme intelligent optimization unit and a real-time feedback and dynamic adjustment unit. The comprehensive optimization of the wind power gear box test process is realized by integrating the digital twin model building unit, the virtual working condition simulation pretest unit, the test scheme intelligent optimization unit and the real-time feedback and dynamic adjustment unit, and the digital twin model can accurately simulate the operation states of the gear box and the test system thereof, so that the test efficiency of the wind power gear box is improved. The virtual working condition simulation and prediction unit can simulate and predict the operation condition of the gearbox under different working conditions, so that potential problems are found and coping strategies are formulated before actual testing, and the intelligent testing scheme optimization unit optimizes an actual testing scheme according to a virtual working condition simulation result.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power testing, and particularly to a wind power gearbox testing system. Background Art

[0002] With the continuous growth of the global demand for renewable energy, wind power generation, as a clean and renewable energy form, has become increasingly important. The wind power gearbox, as a key component of the wind power generation system, undertakes the important task of converting wind energy into mechanical energy and transmitting it to the generator. However, due to the harsh working environment and complex forces of the wind power gearbox, its failure rate is relatively high, seriously affecting the stability and reliability of the wind power generation system.

[0003] Traditional wind power gearbox testing technologies have deficiencies. Firstly, when constructing traditional digital twin models, the complexity and diversity of each component in the gearbox and its testing system are often ignored, resulting in insufficient model accuracy and inability to accurately reflect the operating state of the actual system. Secondly, traditional virtual pre-testing methods lack sufficient flexibility and adaptability when dealing with the operating conditions of the gearbox under different working conditions, and it is difficult to comprehensively simulate the performance of the gearbox under various complex working conditions. In addition, the optimization process of traditional testing schemes often relies on manual experience and the trial-and-error method, with low efficiency and poor optimization effects. Finally, traditional real-time feedback and dynamic adjustment mechanisms are easily affected by noise interference and model errors when processing real-time data, resulting in inaccurate adjustment strategies and affecting the accuracy and stability of testing.

[0004] In summary, traditional wind power gearbox testing methods have certain limitations in practical applications and cannot meet the requirements of the modern wind power generation industry for efficient and accurate testing. Therefore, it is particularly important to develop a wind power gearbox testing system. Summary of the Invention

[0005] The purpose of the present invention is to make up for the deficiencies of the existing technology and provide a wind power gearbox testing system. It can comprehensively optimize the testing process of the wind power gearbox by integrating multiple units such as a digital twin model, virtual working condition simulation pre-testing, intelligent optimization of the testing scheme, and real-time feedback and dynamic adjustment. Through the application of this system, the performance of the wind power gearbox under various working conditions can be accurately predicted and evaluated, providing a more reliable basis for actual testing and maintenance, helping to improve the performance prediction ability of the gearbox, extend its service life and reduce maintenance costs.

[0006] To solve the above technical problems, the present invention provides the following technical solution: A wind power gearbox testing system, which includes the following components: a digital twin model building unit, a virtual working condition simulation pre-testing unit, a testing scheme intelligent optimization unit, and a real-time feedback and dynamic adjustment unit;

[0007] The digital twin model construction unit: Using multi-physical field coupling modeling technology, it models the gears, shafts, bearing components of the wind turbine gearbox and the loading devices and measuring instruments in the test system. By combining finite element analysis and machine learning algorithms, an accurate digital twin model is constructed. The specific algorithm is as follows:

[0008]

[0009] Among them, M(X) represents the output result of the digital twin model under the input parameter X. X includes the geometric parameters, material properties, and operating condition information of the components, and F i (X) is the analysis result of the i-th physical field obtained based on finite element analysis under the input X. n is the number of physical fields, and w i is the weight of the i-th physical field. At the same time, the temperature, pressure, and vibration data of the actual physical system are collected in real time through sensors to correct the model in real time;

[0010] The virtual working condition simulation and pre-test unit: Using the established digital twin model, it simulates the operation of the wind turbine gearbox under different working conditions. It adopts a dynamic simulation algorithm based on the state space method. The specific formula is:

[0011]

[0012] Among them, X(t) is the state vector, U(t) is the input vector, Y(t) is the output vector, and A(t), B(t), C(t), D(t) are state space matrices. Through this algorithm, the operation of the gearbox under different working conditions is simulated, the operation data of each component of the gearbox are analyzed, the problems that occur, and the performance indicators are predicted;

[0013] The test plan intelligent optimization unit: According to the results of the virtual working condition simulation and pre-test unit, it optimizes the actual test plan. It adopts an optimization algorithm that combines the analytic hierarchy process and the genetic algorithm. First, a hierarchical structure model is constructed, and the optimization objectives, criterion layers, and plan layers of the test plan are hierarchically divided. Then, through expert scoring and pairwise comparison, the relative importance between the elements of each layer is determined, a judgment matrix is constructed, the eigenvector and the largest eigenvalue of the judgment matrix are calculated, and the weights of the elements of each criterion layer and plan layer are obtained. The genetic algorithm is used to optimize the plan layer with the fitness function as the guide, where S is the test plan, and g k (S) is the function corresponding to the k-th optimization objective, and α k is the weight of the k-th optimization objective. Through selection, crossover, and mutation genetic operations, the optimal actual test plan is obtained. For the problems found in the pre-test, improvement measures for component design and adjustment of test parameters are formulated, and the process and steps of the actual test are scientifically planned;

[0014] The real-time feedback and dynamic adjustment unit: During the actual testing process, continuously collect the real-time data of the wind power gearbox and its testing system, and feed these data back to the digital twin model. An adaptive filtering algorithm is used to process the real-time data, and the formula is:

[0015]

[0016] where X(n) is the input real-time data sequence, Y(n) is the filtered output data, w i (n) is the filter coefficient at the nth moment, N is the order of the filter, and the filter coefficient w i (n) is updated by the least mean square error algorithm, that is, w i (n + 1) = w i (n) + 2μ·e(n)·X(n - i), where μ is the step factor, e(n) = D(n) - Y(n) is the error between the expected output D(n) and the actual output Y(n). The digital twin model compares and analyzes the processed data with the virtual pre-test results and the expected performance indicators. If it is found that there is a deviation between the actual operation situation and the pre-test results, or a new abnormal situation occurs, a fuzzy control algorithm is used to determine the adjustment strategy. The input of the fuzzy control algorithm is the deviation e and the deviation change rate Δe, and the control quantity u is obtained through fuzzyfication, fuzzy inference, and defuzzification, which is used to adjust the actual test.

[0017] Furthermore, in the digital twin model building unit, when modeling the gear, further consider the factors of tooth profile error, tooth direction error, and manufacturing error of the gear. By measuring the errors of a large number of gears in actual production, an error database is established. During the modeling process, the error is introduced into the finite element model as a random variable, and the Monte Carlo simulation method is used to simulate each gear model multiple times to obtain the mechanical property distribution of the gear under different error combinations. The specific steps are as follows:

[0018] Randomly extract the tooth profile error δ f and the tooth direction error δ b from the error database, substitute them into the gear finite element model, and calculate the stress distribution σ(δ f , δ b ), deformation Δl(δ f , δ b ) and other mechanical property parameters of the gear under this error combination. Repeat the above process M times to obtain the probability distribution of the mechanical property parameters. When determining the output result M gear (X) of the gear part in the digital twin model, comprehensively consider these probability distributions to make the model more accurately reflect the performance of the actual gear, thereby improving the accuracy of the entire digital twin model.

[0019] Furthermore, in the virtual working condition simulation pre-test unit, when simulating different working conditions, the factor of wind direction change is also introduced, a mathematical model of the relationship between wind direction and the force on the gearbox is established, and through the analysis of the actual operation data of the wind turbine and in combination with the principle of aerodynamics, the functional relationships between the wind direction angle θ and the force components F x (θ), F y (θ), F z (θ) are obtained. In the dynamic simulation algorithm based on the state space method, the wind direction angle is used as a component of the input vector , that is, U(t) = [U 1 (t), U 2 (t), …, θ(t), …], where U 1 (t), U 2 (t) are other working condition parameters, and the elements of the state space matrices A(t), B(t), C(t), D(t) are adjusted accordingly according to the relationship between wind direction and the force on the gearbox, so as to more comprehensively simulate the operation of the gearbox under different wind direction working conditions, and thus more accurately predict the problems and performance changes of the gearbox under complex wind conditions.

[0020] Furthermore, in the test scheme intelligent optimization unit, when using the analytic hierarchy process to determine the weights of the elements in each criterion layer and scheme layer, the entropy weight method is introduced for correction. The entropy weight method measures the degree of dispersion of the indicators by calculating the information entropy of each indicator, so as to determine the objective weights of each indicator. First, for the k-th indicator, calculate its information entropy where X kj is the value of the j-th scheme under the k-th indicator, m is the number of schemes, and then calculate the entropy weight of this indicator The objective weight obtained by the entropy weight method and the subjective weight obtained by the analytic hierarchy process are linearly combined to obtain the comprehensive weight where β is the adjustment coefficient, and its optimal value is determined through multiple tests to make the determination of the weights more objective and reasonable, thereby improving the accuracy and reliability of the test scheme optimization.

[0021] Furthermore, in the real-time feedback and dynamic adjustment unit, when determining the adjustment strategy using the fuzzy control algorithm, the fuzzy inference rules are refined. According to actual test experience and expert knowledge, a more detailed fuzzy rule base is constructed. For the deviation e and the deviation change rate Δe, they are respectively divided into multiple fuzzy subsets, such as negative big (NB), negative medium (NM), negative small (NS), zero (Z), positive small (PS), positive medium (PM), and positive big (PB). For different combinations of fuzzy subsets, specific fuzzy inference rules are formulated. For example, when e is positive small (PS) and Δe is positive small (PS), the control quantity u obtained by fuzzy inference is to appropriately increase the fine-tuning amount of the test parameters; when e is positive big (PB) and Δe is negative big (NB), the control quantity u obtained by fuzzy inference is an instruction to greatly reduce the test parameters and conduct equipment inspection. Through the refined fuzzy inference rules, the fuzzy control algorithm can more accurately determine a reasonable adjustment strategy based on the deviation between the actual operation situation and the pre-test results, ensuring the stability and accuracy of the actual test.

[0022] Furthermore, in the digital twin model building unit, to improve the real-time performance of the model, parallel computing technology is adopted. The digital twin model is divided according to components or physical fields and allocated to multiple computing nodes for simultaneous calculation. The modeling and calculation of components such as gears, shafts, and bearings are respectively allocated to different computing nodes. Each computing node independently conducts finite element analysis and data processing. Through high-speed network communication technology, data interaction and synchronization between computing nodes are achieved. During the calculation process, the message passing interface protocol is used for data communication to ensure that model parameters and calculation results can be accurately and timely shared between nodes. At the same time, a task scheduling algorithm is adopted to dynamically allocate calculation tasks according to the computing power and load conditions of each computing node, avoiding the situation of idle or overloaded computing resources on a certain node, thereby effectively improving the building and real-time correction efficiency of the digital twin model and meeting the requirements for the real-time performance of the model during the actual test process.

[0023] Furthermore, in the virtual working condition simulation and pre-test unit, to more accurately predict the fatigue life of gearbox components, a fatigue life prediction model based on damage mechanics is introduced. This model is based on the damage evolution law of materials and considers the influence of stress amplitude, mean stress, and loading frequency factors on fatigue damage. The specific formula is:

[0024]

[0025] where D is the damage variable, N is the number of loading cycles, is the damage rate, which is obtained by fitting experimental data with the stress amplitude σ a and the mean stress σ m、The functional relationship of the load frequency f parameter. During the virtual working condition simulation process, based on the stress and strain data of each component obtained from the simulation, the stress amplitude and mean stress parameters are calculated and substituted into the fatigue life prediction model to calculate the fatigue damage degree of the component in real time and predict the fatigue life of the component. In this way, the fatigue failure conditions of the components of the gearbox under different working conditions can be predicted more accurately, providing a more reliable basis for actual testing and maintenance.

[0026] Furthermore, in the test scheme intelligent optimization unit, when using the genetic algorithm for optimization, the genetic operations are improved. In the selection operation, the tournament selection method is adopted. A certain number of individuals are randomly selected from the population to form a tournament group, and the individual with the highest fitness in the group is selected to enter the next generation population. This selection method can avoid the unreasonable situation in the roulette wheel selection method where individuals with high fitness are selected multiple times and individuals with low fitness also have a certain probability of being selected, improving the efficiency and accuracy of the selection operation. In the crossover operation, the multi-point crossover method is adopted, randomly selecting multiple crossover points to exchange the genes of the parental individuals, increasing the diversity of the population and avoiding premature convergence of the algorithm. In the mutation operation, the mutation probability is dynamically adjusted according to the fitness value of the individual. For individuals with lower fitness, the mutation probability is increased to give them more opportunities to explore new solution spaces. For individuals with higher fitness, the mutation probability is decreased to maintain the stability of excellent individuals. Through these improved genetic operations, the optimization performance of the genetic algorithm is improved, and thus the optimal actual test scheme can be found more effectively.

[0027] Furthermore, in the real-time feedback and dynamic adjustment unit, to better handle the noise interference in real-time data, a method combining wavelet transform and adaptive filtering is adopted. First, the collected real-time data X(n) is subjected to wavelet transform, and the data is decomposed into different frequency sub-bands. By analyzing the energy distribution of each sub-band, the sub-band where the noise is mainly concentrated is determined. Then, threshold processing is performed on the noise sub-band to remove the noise component. After that, the processed sub-band data is subjected to inverse wavelet transform to obtain the preliminarily denoised data. Finally, the preliminarily denoised data is input into the adaptive filter for further filtering. This method can make full use of the advantages of wavelet transform in time-frequency analysis, effectively remove the noise in real-time data, improve the quality of the data, and provide more reliable data support for the accurate comparative analysis of the digital twin model and the dynamic adjustment of actual testing.

[0028] Compared with the prior art, this wind power gearbox test system has the following beneficial effects:

[0029] 1. The present invention realizes the comprehensive optimization of the test process of the wind turbine gearbox by integrating a digital twin model building unit, a virtual working condition simulation pre-test unit, a test plan intelligent optimization unit, and a real-time feedback and dynamic adjustment unit. Among them, the digital twin model can accurately simulate the operating state of the gearbox and its test system. The virtual working condition simulation pre-test unit can simulate and predict the operating conditions of the gearbox under different working conditions, so as to discover potential problems and formulate countermeasures before actual testing. The test plan intelligent optimization unit optimizes the actual test plan based on the results of the virtual working condition simulation to ensure the scientificity and effectiveness of the test. The real-time feedback and dynamic adjustment unit continuously collects data during the actual test process, compares and analyzes it with the digital twin model, and adjusts the test strategy in a timely manner, thereby improving the efficiency and accuracy of the test.

[0030] 2. The present invention introduces a series of advanced technical means, such as a fatigue life prediction model based on damage mechanics and a data processing method combining wavelet transform and adaptive filtering. The application of these technical means enables the system to more accurately predict the fatigue life and performance changes of gearbox components. At the same time, the real-time feedback and dynamic adjustment unit can timely discover the deviation between the actual operating conditions and the pre-test results and adopt reasonable adjustment strategies to ensure the stable operation of the gearbox. These functions not only improve the performance prediction ability of the gearbox but also provide a more reliable basis for subsequent maintenance and servicing, helping to extend the service life of the gearbox and reduce maintenance costs.

[0031] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a flow operation diagram of a wind turbine gearbox test system;

[0034] Figure 2 It is a flow operation diagram of the digital twin model building unit and the virtual working condition simulation pre-test unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features and their effects of the present invention as follows.

[0036] Embodiment 1

[0037] This embodiment describes that during the R & D process of a wind power gearbox, the R & D team is committed to ensuring that the newly designed gearbox has excellent performance and high reliability. For this purpose, a comprehensive testing work is carried out, involving multiple key links, and each link is closely coordinated to achieve accurate evaluation and optimization of the gearbox performance.

[0038] Using multi - physical - field coupling modeling technology, comprehensively considering the complex structure and physical properties of the gearbox, detailed modeling is carried out on gear, shaft, bearing components, as well as loading devices and measuring instruments. Especially when modeling the gear, great attention is paid to factors such as tooth profile error, tooth direction error, and manufacturing error. By measuring the errors of a large number of gears in actual production, a rich error database is constructed. During the modeling process, the error is introduced as a random variable into the finite - element model. With the help of the Monte Carlo simulation method, the tooth profile error δ f and the tooth direction error δ b are repeatedly and randomly extracted from the error database and substituted into the gear finite - element model for calculation. Each calculation can obtain the stress distribution σ(δ f ,δ b ), deformation Δl(δ f ,δ b ) and other key mechanical property parameters of the gear under this error combination. Repeating the above process M times can obtain the probability distributions of these mechanical property parameters. By synthesizing these probability distributions, the output result M gear (X) of the gear part in the digital twin model is determined, so that the model can more accurately reflect the performance of the actual gear.

[0039] Combined with the formula to construct an accurate digital twin model, where M(X) represents the output result of the digital twin model under the input parameter X, X covers key elements such as geometric parameters, material properties, and operating condition information of the component, F i (X) is the analysis result of the i - th physical field under the input X obtained based on finite - element analysis, n is the number of physical fields, w i is the weight of the i - th physical field. At the same time, sensors are used to collect real - time temperature, pressure, and vibration data of the actual physical system, and the model is dynamically corrected based on these real - time data to ensure the accuracy and timeliness of the model.

[0040] Using the established digital twin model, fully considering the complex working conditions in actual operation, introducing the important factor of wind direction change, through in-depth analysis of the actual operation data of wind turbines, and combining with the principles of aerodynamics, an accurate mathematical model of the relationship between wind direction and the force on the gearbox is established, and then the functional relationships between the wind direction angle θ and the force components F x (θ), F y (θ), F z (θ) are obtained. The wind direction angle is taken as a component of the input vector U(t), that is, U(t) = [U 1 (t), U 2 (t), …, θ(t), …], where U 1 (t), U 2 (t) are other working condition parameters. Based on the dynamic simulation algorithm of the state space method simulate the operation of the gearbox under different working conditions. Through this simulation process, analyze the operation data of each component of the gearbox in detail, predict the problems that may occur in advance, such as component wear and fatigue cracks, and accurately evaluate the performance indicators of the gearbox, such as transmission efficiency and stability.

[0041] According to the results of virtual working condition simulation and prediction, optimize the actual test plan. First, construct a hierarchical structure model, clearly divide the optimization objectives, criterion layer, and plan layer of the test plan into different levels. By means of expert scoring and pairwise comparison, determine the relative importance between elements at each level, and then construct a judgment matrix, calculate the eigenvector and the largest eigenvalue of the judgment matrix, so as to obtain the weights of elements in each criterion layer and plan layer. To improve the accuracy of weight determination, the entropy weight method is introduced for correction. The entropy weight method measures the dispersion degree of indicators by calculating the information entropy of each indicator, so as to determine the objective weight of each indicator. The specific calculation process is as follows. For the kth indicator, first calculate its information entropy where X kj is the value of the jth plan under the kth indicator, m is the number of plans, and then calculate the entropy weight of this indicator Finally, linearly combine the objective weight obtained by the entropy weight method and the subjective weight obtained by the analytic hierarchy process to get the comprehensive weight

[0042] Use the genetic algorithm to optimize the plan layer, guided by the fitness function , where S is the test plan, g k (S) is the function corresponding to the kth optimization objective, α k$w_k$ is the weight of the $k$-th optimization objective. During the genetic operation, the tournament selection method is adopted. A certain number of individuals are randomly selected from the population to form a tournament group, and the individual with the highest fitness in the group is selected to enter the next generation population. The crossover operation adopts the multi-point crossover method, randomly selects multiple crossover points, and exchanges the genes of the parental individuals, effectively increasing the diversity of the population and avoiding premature convergence of the algorithm. The mutation operation dynamically adjusts the mutation probability according to the fitness value of the individual. For individuals with lower fitness, the mutation probability is appropriately increased to give them more opportunities to explore new solution spaces. For individuals with higher fitness, the mutation probability is reduced to maintain the stability of excellent individuals. After a series of selection, crossover, and mutation genetic operations, the optimal actual test plan is finally obtained. For the problems found in the pre-test, such as stress concentration in some components and potential faults caused by too high temperature, targeted countermeasures are formulated, such as improving component design, adjusting test parameters, and scientifically planning the process and steps of the actual test.

[0043] During the actual test process, sensors are used to continuously collect real-time data of the wind power gearbox and its test system, including key parameters such as temperature, pressure, vibration, and rotational speed. The collected data is processed by combining wavelet transform and adaptive filtering. First, the collected real-time data $X(n)$ is subjected to wavelet transform, and the data is decomposed into different frequency sub-bands. By analyzing the energy distribution of each sub-band, the sub-band where the noise mainly concentrates is accurately determined. Then, threshold processing is performed on the noise sub-band to remove the noise components. Next, the processed sub-band data is subjected to inverse wavelet transform to obtain the preliminarily denoised data. Finally, the preliminarily denoised data is input into the adaptive filter for further filtering. The formula is where $X(n)$ is the input real-time data sequence, $Y(n)$ is the filtered output data, $w$ i $(n)$ is the filter coefficient at the $n$-th moment, $N$ is the order of the filter, and the filter coefficient $w$ i $(n + 1)=w$ i $(n)+2\mu\cdot e(n)\cdot X(n - i)$, where $\mu$ is the step size factor, and $e(n)=D(n)-Y(n)$ is the error between the desired output $D(n)$ and the actual output $Y(n)$.

[0044] The processed data is comprehensively compared and analyzed with the virtual pre-test results and the expected performance indicators. If it is found that the actual operation situation deviates from the pre-test results, or new abnormal conditions occur, such as a sudden increase in vibration amplitude or an abnormal increase in temperature, the fuzzy control algorithm is used to determine the adjustment strategy. The input of the fuzzy control algorithm is the deviation e and the deviation change rate Δe. The control quantity u is obtained through fuzzification, fuzzy reasoning and defuzzification. For example, when the deviation e is large and the deviation change rate Δe is also large, the fuzzy control algorithm outputs a larger control quantity u, which is used to adjust the actual test, such as adjusting the test loading speed and changing the test environment temperature, to ensure that the test process can more accurately reflect the actual performance of the gearbox, while promptly discovering potential problems and taking corresponding measures to optimize them.

[0045] Embodiment 2

[0046] This embodiment describes an offshore wind farm located in a sea area 50 kilometers away from the coastline. The area is affected by monsoons and ocean cyclones all year round. The wind speed varies widely, with an average wind speed of 8-15 m / s and a maximum wind speed of 30 m / s. The wind direction is uncertain, and the temperature difference between day and night can reach more than 10°C. At the same time, the high-humidity sea breeze environment contains a large amount of salt, which is highly corrosive to wind power equipment. The wind farm has 50 existing wind turbines, and plans to conduct comprehensive tests on 20 newly purchased wind turbine gearboxes of the same model to ensure that they can operate stably for 20 years to meet the needs of long-term and efficient power generation in the wind farm.

[0047] Using multi-physics field coupling modeling technology, we conducted an in-depth analysis of the impact of high humidity and strong corrosion environment at sea on the material properties of gears, shafts, and bearing components. The study found that after being exposed to such an environment for a long time, the hardness of the gear material would decrease by 5-10%, and the elastic modulus would decrease by 8-12%. Therefore, the material parameters were adjusted accordingly during modeling. At the same time, the corrosive effect of salt spray on measuring instruments and equipment was taken into account, and the impact of salt spray concentration and corrosion time on instrument accuracy was simulated to correct the measurement data.

[0048] By combining finite element analysis with machine learning algorithms, an accurate digital twin model is constructed. Finite element analysis is used to simulate the stress and strain distribution of gears under complex working conditions. Machine learning algorithms are used to process large amounts of historical operating data, explore potential patterns in the data, and improve the accuracy of the model.

[0049] Sensors are installed on each test wind turbine in the wind farm to collect temperature, pressure, and vibration data in real time. The data is transmitted to the control center every 10 minutes. Using parallel computing technology, the digital twin model is divided into multiple computing nodes according to components or physical fields for simultaneous calculations. The model is comprehensively revised every day based on the collected data to ensure that the model is closely aligned with the actual situation.

[0050] The constructed digital twin model is used to simulate the operation of offshore wind turbine gearboxes under different wind speeds, wind directions, and wave turbulence conditions. The wind speed is set to 11 different levels of 5-25 m / s, and the wind direction is simulated every 15°. The wave turbulence is divided into three levels: mild, moderate, and severe according to the sea conditions, which are combined to form a variety of complex operating conditions.

[0051] By introducing the factor of wind direction change, through long-term monitoring and analysis of the actual operation data of wind turbines and combining it with the principles of aerodynamics, an accurate mathematical model of the wind direction angle θ and the force components of each gearbox component was established. The study found that for every 30° change in wind direction, the force on a key part of the gearbox increases by 10-15%.

[0052] Combined with the fatigue life prediction model based on damage mechanics, during the virtual working condition simulation, the stress amplitude and average stress parameters are calculated in real time according to the stress and strain data of each component obtained by simulation, and substituted into the fatigue life prediction model. The prediction results show that after running for one year under harsh working conditions, the fatigue damage degree of some key components of the gearbox will reach 10-15%.

[0053] According to the pre-test results of virtual working condition simulation, an optimization algorithm combining hierarchical analysis method and genetic algorithm is adopted, and the entropy weight method is introduced to correct the weights. The optimization goals of the test plan are determined to improve test accuracy, shorten test time and reduce test cost.

[0054] In view of the limited space on the offshore test platform and the difficulty in transporting equipment, we formulated a response strategy and gave priority to the testing of key components when planning the test process. We also arranged the test sequence in a reasonable manner according to the complexity of different test items and the required equipment to ensure that comprehensive testing can be completed within a limited test cycle.

[0055] When the gearbox is actually tested on the offshore test platform, data is collected every 5 minutes, including the vibration, temperature, speed, and torque parameters of the gearbox, and the data is fed back to the digital twin model in real time through wireless transmission technology.

[0056] The data is processed by combining wavelet transform with adaptive filtering, and the fuzzy control algorithm is used to refine the inference rules. The test parameters are adjusted in time according to the deviation between the actual operation and the pre-test results. When it is found that the temperature of a certain part of the gearbox rises abnormally, the fuzzy control algorithm automatically adjusts the test load and speed according to the temperature deviation and the deviation change rate, and starts the backup cooling system at the same time to ensure the safety of the test.

[0057] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments with equivalent changes within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A wind turbine gearbox testing system, characterized in that: The system includes the following components: digital twin model building unit, virtual working condition simulation pre-test unit, test scheme intelligent optimization unit and real-time feedback and dynamic adjustment unit; The digital twin model building unit uses multi-physics field coupling modeling technology to model the gears, shafts, bearing components of the wind turbine gearbox, and the loading devices and measuring instruments in the test system, and builds an accurate digital twin model by combining finite element analysis with machine learning algorithms. The specific algorithm is as follows: Where M(X) represents the output result of the digital twin model under the input parameter X, X includes the geometric parameters, material properties, and operating condition information of the component, and F i (X) is the analysis result of the i-th physical field under input X based on finite element analysis, n is the number of physical fields, w i is the weight of the i-th physical field. At the same time, the temperature, pressure, and vibration data of the actual physical system are collected in real time through sensors to correct the model in real time. The virtual working condition simulation pre-test unit: uses the built digital twin model to simulate the operation of the wind turbine gearbox under different working conditions, and adopts a dynamic simulation algorithm based on the state space method. The specific formula is: Among them, X(t) is the state vector, U(t) is the input vector, Y(t) is the output vector, A(t), B(t), C(t), and D(t) are state space matrices. The algorithm is used to simulate the operation of the gearbox under different working conditions, analyze the operating data of each component of the gearbox, predict the problems that may occur, and predict the performance indicators; The test scheme intelligent optimization unit: optimizes the actual test scheme according to the results of the virtual test simulation unit, and adopts the optimization algorithm combining the hierarchical analysis method and the genetic algorithm. First, a hierarchical structure model is constructed, and the optimization target, criterion layer and scheme layer of the test scheme are hierarchically divided. Then, the relative importance of the elements of each layer is determined by expert scoring and pairwise comparison, and a judgment matrix is ​​constructed. The eigenvector and the maximum eigenvalue of the judgment matrix are calculated, and the weights of the elements of each criterion layer and scheme layer are obtained. The genetic algorithm is used to optimize the scheme layer, and the fitness function is used to determine the relative importance of the elements of each layer. As a guide, S is the test plan, g k (S) is the function corresponding to the kth optimization objective, α k The weight of the kth optimization goal is obtained through selection, crossover, and mutation genetic operations to obtain the optimal actual test plan. For the problems found in the pre-test, strategies for improving component design and adjusting test parameters are formulated to scientifically plan the process and steps of the actual test; The real-time feedback and dynamic adjustment unit: in the actual test process, continuously collects real-time data of the wind turbine gearbox and its test system, and feeds these data back to the digital twin model, and uses an adaptive filtering algorithm to process the real-time data. The formula is: Among them, X(n) is the input real-time data sequence, Y(n) is the output data after filtering, and w i (n is the filter coefficient at the nth moment, N is the order of the filter, and the filter coefficient w is updated by the minimum mean square error algorithm i (n), that is, w i (n+1)=w i (n)+2μ·e(n)·X(ni), where μ is the step factor, e(n)=D(n)-Y(n) is the error between the expected output D(n) and the actual output Y(n). The digital twin model compares and analyzes the processed data with the virtual pre-test results and the expected performance indicators. If it is found that the actual operation situation deviates from the pre-test results, or a new abnormal situation occurs, the fuzzy control algorithm is used to determine the adjustment strategy. The input of the fuzzy control algorithm is the deviation e and the deviation change rate Δe. The control quantity u is obtained through fuzzification, fuzzy reasoning and defuzzification, which is used to adjust the actual test.

2. A wind turbine gearbox testing system according to claim 1, characterized in that: In the digital twin model building unit, when modeling the gear, the gear tooth profile error and tooth direction error manufacturing error factors are further considered. By measuring the errors of a large number of gears actually produced, an error database is established. In the modeling process, the errors are introduced into the finite element model as random variables. The Monte Carlo simulation method is used to simulate each gear model multiple times to obtain the mechanical property distribution of the gear under different error combinations. The specific steps are as follows: Randomly extract the tooth profile error δ from the error database f and tooth direction error δ b , and substitute it into the gear finite element model to calculate the stress distribution σ(δ f ,δ b ), deformation Δl(δ f ,δ b ) and other mechanical performance parameters, repeat the above process M times to obtain the probability distribution of the mechanical performance parameters, and then determine the output result M of the gear part in the digital twin model. gear (X), these probability distributions are comprehensively considered to make the model reflect the performance of the actual gear more accurately.

3. A wind turbine gearbox testing system according to claim 1, characterized in that: In the virtual working condition simulation pre-test unit, when simulating different working conditions, the factor of wind direction change is also introduced to establish a mathematical model of the relationship between wind direction and gearbox force. By analyzing the actual operation data of the wind turbine, combined with the principle of aerodynamics, the wind direction angle θ and the force component F of each gearbox component are obtained. x (θ), F y (θ), F z (θ), in the dynamic simulation algorithm based on the state space method, the wind direction angle is used as the input vector A component of , that is, U(t) = [U1(t), U2(t), …, θ(t), …], where U1(t) and U2(t) are other operating parameters, and the elements of the state space matrices A(t), B(t), C(t), and D(t) are adjusted accordingly according to the relationship between wind direction and gearbox force.

4. A wind turbine gearbox testing system according to claim 1, characterized in that: In the test scheme intelligent optimization unit, when the weights of the elements of each criterion layer and scheme layer are determined by using the hierarchical analysis method, the entropy weight method is introduced for correction. The entropy weight method measures the discrete degree of the indicator by calculating the information entropy of each indicator, thereby determining the objective weight of each indicator. First, for the kth indicator, its information entropy is calculated. in X kj is the value of the jth solution under the kth indicator, m is the number of solutions, and then the entropy weight of the indicator is calculated The objective weight obtained by the entropy weight method The subjective weight obtained by analytic hierarchy process Perform linear combination to obtain comprehensive weight Where β is the adjustment coefficient.

5. A wind turbine gearbox testing system according to claim 1, characterized in that: In the real-time feedback and dynamic adjustment unit, when the fuzzy control algorithm is used to determine the adjustment strategy, the fuzzy reasoning rules are refined, and a more detailed fuzzy rule base is constructed based on actual test experience and expert knowledge. Through the refined fuzzy reasoning rules, the fuzzy control algorithm can more accurately determine a reasonable adjustment strategy based on the deviation between the actual operating conditions and the pre-test results.

6. A wind turbine gearbox testing system according to claim 1, characterized in that: In the digital twin model building unit, parallel computing technology is used to divide the digital twin model according to components or physical fields, and distribute it to multiple computing nodes for simultaneous calculation. The modeling and calculation of gears, shafts, and bearing components are respectively distributed to different computing nodes. Each computing node independently performs finite element analysis and data processing. During the calculation process, the message passing interface protocol is used for data communication. At the same time, a task scheduling algorithm is used to dynamically allocate computing tasks according to the computing power and load of each computing node to avoid idle or overloaded computing resources at a certain node.

7. A wind turbine gearbox testing system according to claim 1, characterized in that: In the virtual working condition simulation pre-test unit, a fatigue life prediction model based on damage mechanics is introduced. The model is based on the damage evolution law of the material and considers the influence of stress amplitude, average stress and loading frequency factors on fatigue damage. The specific formula is: Where D is the damage variable, N is the number of loading cycles, is the damage rate, which is obtained by fitting the experimental data and the stress amplitude σ a , mean stress σ m , and the functional relationship between the loading frequency f parameter. During the virtual working condition simulation, the stress amplitude and average stress parameters are calculated according to the stress-strain data of each component obtained by simulation, and substituted into the fatigue life prediction model to calculate the fatigue damage degree of the component in real time and predict the fatigue life of the component.

8. A wind turbine gearbox testing system according to claim 1, characterized in that: In the test scheme intelligent optimization unit, when optimizing by using genetic algorithm, the genetic operation is improved. In the selection operation, the tournament selection method is adopted to randomly select a certain number of individuals from the population to form a tournament group, and the individual with the highest fitness is selected from the group to enter the next generation population. In the crossover operation, a multi-point crossover method is adopted to randomly select multiple crossover points to exchange the genes of the parent individuals to increase the diversity of the population and avoid premature convergence of the algorithm. In the mutation operation, the mutation probability is dynamically adjusted according to the fitness value of the individual. For individuals with lower fitness, their mutation probability is increased so that they have more opportunities to explore new solution spaces. For individuals with higher fitness, their mutation probability is reduced to maintain the stability of excellent individuals.

9. A wind turbine gearbox testing system according to claim 1, characterized in that: In the real-time feedback and dynamic adjustment unit, a method combining wavelet transform and adaptive filtering is adopted. First, the collected real-time data X(n) is subjected to wavelet transform to decompose the data into different frequency sub-bands. By analyzing the energy distribution of each sub-band, the sub-band where the noise is mainly concentrated is determined. Then, the noise sub-band is subjected to threshold processing to remove the noise component. Then, the processed sub-band data is subjected to inverse wavelet transform to obtain preliminary denoised data. Finally, the preliminary denoised data is input into an adaptive filter for further filtering.

Images