Fan tower drum bolt fault diagnosis method and system based on multi-source data fusion

Abstract

The application belongs to the technical field of fault diagnosis, and discloses a fan tower drum bolt fault diagnosis method and system based on multi-source data fusion. The method comprises the following steps: constructing a fan tower drum bolt fault diagnosis architecture comprising a multi-source data data acquisition layer, a multi-source data fusion layer and a fan tower drum bolt fault diagnosis layer connected in sequence; using the data acquisition layer to acquire multi-source data of the fan tower drum bolt, and performing pretreatment to obtain multi-source bolt axial stress data; using the multi-source data fusion layer to perform data processing and fusion on the multi-source bolt axial stress data to obtain multi-source sensitive modal components; and using the fan tower drum bolt fault diagnosis layer to perform bolt fault diagnosis on the multi-source sensitive signal components to obtain a fan tower drum bolt fault diagnosis result. The application solves the problems of single data source, high cost, low efficiency and insufficient information utilization in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology, specifically relating to a method and system for fault diagnosis of wind turbine tower bolts based on multi-source data fusion. Background Technology

[0002] With the rapid development of the wind power industry, the operational reliability and safety of wind turbine towers, as key supporting structures, are of paramount importance. Under complex loads such as bolt preload and wind loads, towers are prone to various faults, which seriously affect the stable operation and service life of wind power systems. Therefore, accurate fault diagnosis is crucial for ensuring the safe operation of towers and is of great significance for improving wind power generation efficiency and reducing maintenance costs.

[0003] Existing technologies still have many shortcomings, including:

[0004] 1) Single data source: Most existing methods rely excessively on data from a single vibration sensor. However, vibration signals are easily affected by environmental noise (such as wind noise and electromagnetic interference) and changes in operating conditions (such as wind speed and power output), resulting in low signal-to-noise ratios and difficulties in feature extraction. Vibration characteristics caused by bolt faults are often weak and easily masked, making it difficult to accurately identify faults in the early stages using only vibration signals.

[0005] 2) High cost and low efficiency: Existing technologies often rely on manual inspections for regular inspections, which are costly, inefficient, and pose safety risks, making it difficult to meet the needs of large wind farms for real-time monitoring of equipment status.

[0006] 3) Insufficient information utilization: Some existing multi-source data fusion methods often only perform simple splicing, weighted averaging, or principal component analysis (PCA) on data from different sensors (such as vibration, stress, and temperature). This superficial fusion fails to deeply explore the potential coupling relationships and complementary information between different physical quantities. Summary of the Invention

[0007] To address the problems of single data source, high cost, low efficiency, and insufficient information utilization in existing technologies, the present invention aims to provide a method and system for diagnosing wind turbine tower bolt faults based on multi-source data fusion.

[0008] The technical solution adopted in this invention is as follows:

[0009] A method for diagnosing wind turbine tower bolt faults based on multi-source data fusion includes the following steps:

[0010] A wind turbine tower bolt fault diagnosis architecture is constructed, consisting of a multi-source data acquisition layer, a multi-source data fusion layer, and a wind turbine tower bolt fault diagnosis layer connected in sequence.

[0011] The data acquisition layer of the wind turbine tower bolt fault diagnosis architecture is used to collect multi-source data of wind turbine tower bolts and preprocess it to obtain multi-source bolt axial stress data.

[0012] Using the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture, the axial stress data of multi-source bolts is processed and fused to obtain multi-source sensitive modal components.

[0013] The wind turbine tower bolt fault diagnosis layer, which uses a wind turbine tower bolt fault diagnosis architecture, performs bolt fault diagnosis on multi-source sensitive signal components to obtain the bolt fault diagnosis results of the wind turbine tower.

[0014] Furthermore, a wind turbine tower bolt fault diagnosis architecture is constructed, comprising a multi-source data acquisition layer, a multi-source data fusion layer, and a wind turbine tower bolt fault diagnosis layer connected in sequence, including the following steps:

[0015] A simulation model of the tower bolts was constructed, and bolt displacement data acquisition equipment and bolt stress data acquisition equipment were installed at the wind turbine tower.

[0016] A multi-source data acquisition layer is constructed and connected to the tower bolt simulation model, bolt displacement data acquisition equipment, and bolt stress data acquisition equipment;

[0017] A multi-source data fusion layer is constructed, and modules for wind load normalization, mode decomposition, mode component clustering, and sensitive component screening are set up.

[0018] Construct a fault diagnosis layer for wind turbine tower bolts, and set up a correlation comparison module, a fault judgment module, and a bolt fault diagnosis module;

[0019] The multi-source data acquisition layer, the multi-source data fusion layer, and the wind turbine tower bolt fault diagnosis layer are connected sequentially to obtain the wind turbine tower bolt fault diagnosis architecture.

[0020] Furthermore, the multi-source data includes simulated bolt axial stress data, measured bolt displacement data, and the first measured bolt axial stress data.

[0021] Furthermore, using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, multi-source data of the wind turbine tower bolts is collected and preprocessed to obtain multi-source bolt axial stress data, including the following steps:

[0022] Under wind load and preload, the measured bolt displacement data of the wind turbine tower is collected using bolt displacement data acquisition equipment, and the corresponding first measured bolt axial stress data is collected using bolt stress data acquisition equipment.

[0023] Under wind load and preload, a tower bolt simulation model was used to perform simulation and obtain the axial stress data of the simulated bolts.

[0024] Using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, simulated bolt axial stress data, measured bolt displacement data, and first measured bolt axial stress data are collected to obtain multi-source data of wind turbine tower bolts.

[0025] The measured bolt displacement data were sequentially denoised and converted to obtain the corresponding second measured bolt axial stress data.

[0026] By integrating the simulated bolt axial stress data corresponding to wind load and preload, the first measured bolt axial stress data, and the second measured bolt axial stress data, multi-source bolt axial stress data is obtained.

[0027] Furthermore, using the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture, the axial stress data of multi-source bolts are processed and fused to obtain multi-source sensitive modal components, including the following steps:

[0028] Input multi-source bolt axial stress data into the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture;

[0029] Using the wind load normalization module, the axial stress data of multi-source bolts is normalized by wind load to obtain the axial stress data of multi-source bolts after wind load normalization.

[0030] The modal decomposition module was used to perform modal decomposition on the wind load normalized multi-source bolt axial stress data to obtain multi-source modal components.

[0031] The modal component clustering module is used to perform modal component clustering on multi-source modal components to obtain a multi-source scatter plot;

[0032] The sensitive component filtering module is used to filter the multi-source modal components in the multi-source scatter plot to obtain the multi-source sensitive modal components.

[0033] Furthermore, the multi-source sensitive modal components include the simulated sensitive modal components corresponding to the simulated bolt axial stress data, the first measured sensitive modal components corresponding to the first measured bolt axial stress data, and the second measured sensitive modal components corresponding to the measured bolt displacement data.

[0034] Furthermore, using the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture, bolt fault diagnosis is performed on multi-source sensitive signal components to obtain the bolt fault diagnosis results of the wind turbine tower, including the following steps:

[0035] The multi-source sensitive mode components, including the simulated sensitive mode component, the first measured sensitive mode component, and the second measured sensitive mode component, are input into the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture.

[0036] The correlation comparison module is used to obtain the correlation coefficients between the simulated sensitive modal components, the first measured sensitive modal component, and the second measured sensitive modal component.

[0037] The fault diagnosis module is used to determine whether the correlation coefficient is less than the correlation coefficient threshold. If it is less, the process proceeds to the next step and triggers the bolt fault diagnosis process. Otherwise, data collection continues.

[0038] Based on the simulated sensitive mode components, the first measured sensitive mode component, and the second measured sensitive mode component, the bolt fault diagnosis module is used to perform bolt fault diagnosis and obtain the bolt fault diagnosis results of the wind turbine tower.

[0039] Furthermore, the bolt fault diagnosis module is equipped with a bolt fault diagnosis model, which is constructed based on the LSTM-Attention-MLP algorithm. The bolt fault diagnosis model includes a feature extractor constructed based on the LSTM algorithm, a weighted fusion unit constructed based on the Attention mechanism, and a bolt fault diagnosticer constructed based on the MLP algorithm, which are connected in sequence.

[0040] Furthermore, based on the simulated sensitive mode components, the first measured sensitive mode component, and the second measured sensitive mode component, a bolt fault diagnosis module is used to perform bolt fault diagnosis to obtain the bolt fault diagnosis results for the wind turbine tower, including the following steps:

[0041] The simulated sensitive modal components, the first measured sensitive modal components, and the second measured sensitive modal components are input into the bolt fault diagnosis model of the bolt fault diagnosis module.

[0042] Using a feature extractor, the simulation data features of the simulation sensitive modal component, the first measured data features of the first measured sensitive modal component, and the second measured data features of the second measured sensitive modal component are extracted respectively.

[0043] Based on the dynamic attention weight values, a weighted fusion builder is used to weight and fuse the simulation data features, the first measured data features, and the second measured data features to obtain the weighted fused features.

[0044] Based on the weighted fusion characteristics, a bolt fault diagnostic tool is used to perform bolt fault diagnosis and obtain the bolt fault diagnosis results for the wind turbine tower.

[0045] A wind turbine tower bolt fault diagnosis system based on multi-source data fusion is used to realize the fault diagnosis method of wind turbine tower bolts. The system includes an architecture construction unit, a data acquisition unit, a data processing and fusion unit, and a bolt fault diagnosis unit connected in sequence.

[0046] The beneficial effects of this invention are as follows:

[0047] This invention provides a method and system for fault diagnosis of wind turbine tower bolts based on multi-source data fusion. By collecting and fusing multi-source data (such as simulation data and measured stress data), it overcomes the limitations of traditional methods that rely on only a single vibration signal or a few measurement points. It can capture the structural state information of the tower from a broader perspective, making the diagnostic foundation more solid and comprehensive. Based on the wind turbine tower bolt fault diagnosis architecture, it performs automated data acquisition, processing, and fault diagnosis, avoiding manual inspection, reducing labor costs, minimizing human intervention, and improving fault diagnosis efficiency. Through steps such as wind load normalization, modal decomposition, clustering, and sensitive component screening, it performs in-depth processing of multi-source stress data, extracting modal components that are more sensitive to faults, removing redundant information and noise interference, and providing high-quality input for subsequent fault diagnosis. It is more sensitive to weak signal changes caused by early faults, enabling earlier identification of potential fault signs, achieving early warning, and gaining valuable time for maintenance decisions. More comprehensive information and a more intelligent fusion mechanism help distinguish real fault signals from background noise or normal fluctuations caused by changes in operating conditions, thereby reducing false alarms and missed alarms.

[0048] Other beneficial effects of the present invention will be further explained in the specific embodiments. Attached Figure Description

[0049] Figure 1 This is a flowchart of the wind turbine tower bolt fault diagnosis method based on multi-source data fusion in this invention.

[0050] Figure 2 This is a structural block diagram of the wind turbine tower bolt fault diagnosis system based on multi-source data fusion in this invention. Detailed Implementation

[0051] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0052] Example 1:

[0053] like Figure 1 As shown in the figure, this embodiment provides a method for diagnosing wind turbine tower bolt faults based on multi-source data fusion, including the following steps:

[0054] S1: Construct a wind turbine tower bolt fault diagnosis architecture comprising a multi-source data acquisition layer, a multi-source data fusion layer, and a wind turbine tower bolt fault diagnosis layer connected in sequence, including the following steps:

[0055] S1-1: Construct a simulation model of the tower bolts, and install bolt displacement data acquisition equipment and bolt stress data acquisition equipment at the wind turbine tower, including the following steps:

[0056] S1-1-1: The tower bolt connection is simplified into a spring-mass-damping system, and the corresponding wind load parameters, preload parameters, material properties and geometric parameters are set to obtain a dynamic model including dynamic equations;

[0057] S1-1-2: Import the dynamic model, including the dynamic equations, into the Computer-Aided Engineering (CAE) tool, and use the CAE tool to construct an accurate three-dimensional geometric model of the tower bolts based on the geometric parameters of the dynamic model.

[0058] S1-1-3: Based on the material properties of the dynamic model, define the material property conditions of the precise three-dimensional geometric model, and perform mesh generation on the precise three-dimensional geometric model to obtain the initial tower bolt simulation model;

[0059] S1-1-4: Based on the wind load parameters and preload parameters of the dynamic model, define the wind load conditions, boundary conditions, and preload conditions of the initial tower bolt simulation model to obtain the final tower bolt simulation model.

[0060] S1-1-5: Install bolt displacement data acquisition equipment and bolt stress data acquisition equipment at the wind turbine tower;

[0061] The bolt displacement data acquisition device is an ultrasonic sensor, and the bolt stress data acquisition device is a strain gauge.

[0062] S1-2: Construct a multi-source data acquisition layer and connect it to the tower bolt simulation model, bolt displacement data acquisition equipment, and bolt stress data acquisition equipment;

[0063] S1-3: Construct a multi-source data fusion layer and set up a wind load normalization module, a mode decomposition module, a mode component clustering module, and a sensitive component screening module;

[0064] S1-4: Construct a fault diagnosis layer for wind turbine tower bolts, and set up a correlation comparison module, a fault judgment module, and a bolt fault diagnosis module;

[0065] S1-5: Connect the multi-source data acquisition layer, the multi-source data fusion layer, and the wind turbine tower bolt fault diagnosis layer in sequence to obtain the wind turbine tower bolt fault diagnosis architecture.

[0066] S2: Using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, multi-source data of wind turbine tower bolts is collected and pre-processed to obtain multi-source bolt axial stress data;

[0067] The multi-source data includes simulated bolt axial stress data, measured bolt displacement data, and the first measured bolt axial stress data;

[0068] The data acquisition layer of the wind turbine tower bolt fault diagnosis architecture is used to collect multi-source data on wind turbine tower bolts and perform preprocessing to obtain multi-source bolt axial stress data, including the following steps:

[0069] S2-1: Under wind load and preload, use bolt displacement data acquisition equipment to collect measured bolt displacement data of the wind turbine tower, and use bolt stress data acquisition equipment to collect the corresponding first measured bolt axial stress data.

[0070] S2-2: Under wind load and preload, a tower bolt simulation model is used to perform simulation and obtain the axial stress data of the simulated bolts;

[0071] S2-3: Using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, simulated bolt axial stress data, measured bolt displacement data, and first measured bolt axial stress data are collected to obtain multi-source data of wind turbine tower bolts;

[0072] S2-4: Perform noise reduction and data conversion on the measured bolt displacement data in sequence to obtain the corresponding second measured bolt axial stress data;

[0073] S2-5: Integrate the simulated bolt axial stress data corresponding to wind load and preload, the first measured bolt axial stress data, and the second measured bolt axial stress data to obtain multi-source bolt axial stress data;

[0074] S3: Using the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture, multi-source bolt axial stress data is processed and fused to obtain multi-source sensitive modal components, including the following steps:

[0075] S3-1: Input multi-source bolt axial stress data into the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture;

[0076] S3-2: Using the wind load normalization module, the axial stress data of multi-source bolts is normalized to obtain the wind load normalized axial stress data of multi-source bolts, including the following steps:

[0077] S3-2-1: Using the wind load normalization module, the average wind load stress is calculated by statistically analyzing the simulated bolt axial stress data, the first measured bolt axial stress data, and the second measured bolt axial stress data within a certain time window.

[0078] The formula is:

[0079]

[0080]

[0081] In the formula, This represents the average wind load stress. This is the original stress signal; For wind load-related stress; Fault-related stress; For noise; This represents the number of sampling points within the time window. This is an indicator of the sampling time. Sampling time Wind load-related stress;

[0082] S3-2-2: Normalize the simulated bolt axial stress data / the first measured bolt axial stress data / the second measured bolt axial stress data based on the average wind load stress.

[0083] The formula is:

[0084]

[0085] In the formula, The normalized stress signal is used to make the wind load-related stress close to 1 after normalization, thereby highlighting other components such as fault-related stress. After normalization, the background stress caused by wind load is eliminated, and fault stress becomes the dominant feature.

[0086] In the original stress, wind load-related stress accounts for 60% - 80%, which is the main background component. Wind load normalization eliminates the background stress fluctuations caused by wind load through data processing, making fault-related stress (accounting for <20%) the dominant feature, which facilitates subsequent accurate identification of faults such as bolt loosening, crack initiation, bolt corrosion and bolt breakage.

[0087] S3-3: Using the modal decomposition module, perform symplectic geometric modal decomposition on the wind load-normalized multi-source bolt axial stress data to obtain multi-source modal components, including the following steps:

[0088] S3-3-1: Using the modal decomposition module, the one-dimensional signal corresponding to the axial stress data of multi-source bolts after wind load normalization is processed. Phase space reconstruction is performed, and a multidimensional state vector is constructed using the delayed embedding method to obtain the trajectory matrix of the phase space vector;

[0089] The formula is:

[0090]

[0091] In the formula, Let be the trajectory matrix of the phase space vectors; For delay time; For the embedding dimension; These are phase space elements; extending the signal from the one-dimensional time domain to a multi-dimensional phase space facilitates the capture of the signal's dynamic characteristics.

[0092] S3-3-2: Construct the Hamiltonian matrix based on the trajectory matrix;

[0093] The formula is:

[0094]

[0095] In the formula, It is a Hamiltonian matrix; It is a coefficient matrix; It is the transpose symbol;

[0096] S3-3-3: Based on the symmetric matrix, perform a similarity transformation on the Hamiltonian matrix to obtain the corresponding eigenvalues;

[0097] The formula is:

[0098]

[0099]

[0100]

[0101] In the formula, To satisfy specific conditions for a symplectic matrix; It is a symplectic matrix; It is the identity matrix; It is the eigenvalue matrix; For eigenvalues;

[0102] S3-3-4: Reconstruct the geometric components of each symplectic based on the eigenvectors corresponding to the eigenvalues. Each symplectic geometric component A corresponding mode of the signal reflects specific frequency and amplitude characteristics, enabling adaptive decomposition of the signal to obtain multi-source mode components. , where i is the component indicator;

[0103] S3-4: Use the modal component clustering module to perform Bayesian clustering on multi-source modal components to obtain a multi-source scatter plot;

[0104] The formula is:

[0105]

[0106] In the formula, Let be the maximum a posteriori probability of the clustering features based on the i-th symplectic geometric component; iid is the symbol for independent isomorphic distribution, representing each The iid condition must be met; Based on Having local parameters and global parameters Kernel function; For periodically induced kernel functions; For local parameters, representing each The local parameter is set to the fault frequency; For global parameters, representing the entire The core size can be set to 1;

[0107]

[0108] In the formula, SES is the squared envelope spectral function; max[SES( [ ] is used to extract potential fault frequencies; Execution for each The characteristic distribution; the pivotal role is The construction of this structure allows for more precise clustering of categories;

[0109]

[0110] In the formula, Weighted average ; For each The weights are set as binary sequences; The contraction coefficient of Bayesian clustering. The range is [0,1]. The higher the clustering level, the more accurate the clustering. Lower values ​​can cause category confusion. Theoretically, it should be set to 1, but due to the random fluctuations in the operating state of wind turbines, the failure frequency will fluctuate. For accurate clustering, it can be... The value is set to around 0.8, which means that the fluctuation of about 20% of the optimal amplitude after the fault frequency is mapped is regarded as a potential fault-related component;

[0111] S3-5: Use the sensitive component filtering module to perform sensitive component filtering on the multi-source modal components in the multi-source scatter plot to obtain the multi-source sensitive modal components.

[0112] Using the sensitive component filtering module, based on the scatter plot, it is easy to filter out high-amplitude sensitive points, i.e., multi-source modal components containing fault information. A sorting algorithm is used to sort the data from highest to lowest amplitude. Sort the data and select the first few items after sorting. The highest amplitude (Set an amplitude threshold, for example, only amplitudes greater than 0.05). The threshold is set very small because the amplitudes of sensitive components and insensitive components differ greatly after clustering; only high amplitudes need to be extracted. Thus, multi-source sensitive mode components are obtained;

[0113] The multi-source sensitive modal components include the simulated sensitive modal components corresponding to the simulated bolt axial stress data, the first measured sensitive modal components corresponding to the first measured bolt axial stress data, and the second measured sensitive modal components corresponding to the measured bolt displacement data.

[0114] S4: The wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture performs bolt fault diagnosis on multi-source sensitive signal components to obtain the bolt fault diagnosis results of the wind turbine tower, including the following steps:

[0115] S4-1: Input the multi-source sensitive mode component, including the simulated sensitive mode component, the first measured sensitive mode component, and the second measured sensitive mode component, into the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture.

[0116] S4-2: Using the correlation comparison module, obtain the Pearson correlation coefficients between the simulated sensitive modal components, the first measured sensitive modal component, and the second measured sensitive modal component, including the following steps:

[0117] S4-2-1: Using the correlation comparison module, the waveforms corresponding to the simulated sensitive mode component, the first measured sensitive mode component, and the second measured sensitive mode component are preprocessed to remove the mean and eliminate the influence of DC offset on the correlation, ensuring that the analysis is of waveform shape rather than absolute amplitude, and obtaining the preprocessed waveforms corresponding to the simulated sensitive mode component, the first measured sensitive mode component, and the second measured sensitive mode component.

[0118] The formula is:

[0119]

[0120] In the formula, These represent the preprocessed data points, the original data points, and the mean of the data points in the first waveform. These represent the preprocessed data points, the original data points, and the mean of the data points in the second waveform. For data point indication;

[0121]

[0122] In the formula, This represents the total number of data points in the waveform.

[0123] S4-2-2: If large waveform amplitude differences are to be eliminated, the preprocessed waveform can be standardized to eliminate waveform amplitude differences and further standardize the data into Z-scores.

[0124] The formula is:

[0125]

[0126] In the formula, These are the standardized data points in the first waveform; These are the standardized data points in the second waveform; This is the reference value for the first waveform; This is the reference value for the second waveform;

[0127] S4-2-3: Calculate the sum of the products of deviations from the mean, reflecting the degree of correlation between the first waveform X and the second waveform Y in deviating from the mean; if it is positive: X and Y are both higher or lower than the mean (positive correlation); if it is negative: when X is higher than the mean, Y is lower than the mean, and vice versa (negative correlation).

[0128] The formula is:

[0129]

[0130] In the formula, It is the sum of the products of deviations from the mean;

[0131] S4-2-4: Calculate the square root of the sum of squares of the deviations from the mean based on the product of the preprocessed waveform and the deviations from the mean;

[0132] The formula is:

[0133]

[0134] In the formula, It is the square root of the sum of squares of the deviations from the mean of the first and second waveforms;

[0135] S4-2-4: The sum of square roots of the sum of squares of deviations from the mean is used to calculate the Pearson correlation coefficient based on the sum of the products of deviations from the mean.

[0136]

[0137] In the formula, The Pearson correlation coefficient;

[0138] S4-3: Use the fault diagnosis module to determine whether the correlation coefficient is less than the correlation coefficient threshold (e.g., set the correlation coefficient threshold to 0.85). If the Pearson correlation coefficient is less than the threshold, then... If the value is less than 0.85, it is considered a fault, and the process proceeds to the next step, triggering the bolt fault diagnosis process; otherwise, data acquisition continues.

[0139] S4-4: Based on the simulated sensitive mode components, the first measured sensitive mode components, and the second measured sensitive mode components, the bolt fault diagnosis module is used to perform bolt fault diagnosis and obtain the bolt fault diagnosis results of the wind turbine tower.

[0140] The bolt fault diagnosis module is equipped with a bolt fault diagnosis model, which is built based on the Long Short-Term Memory (LSTM) network-Attention-Multilayer Perceptron (MLP) algorithm. The bolt fault diagnosis model includes a feature extractor built based on the LSTM algorithm, a weighted fusion unit built based on the Attention mechanism, and a bolt fault diagnosticer built based on the MLP algorithm, which are connected in sequence.

[0141] Based on the simulated sensitive mode components, the first measured sensitive mode component, and the second measured sensitive mode component, a bolt fault diagnosis module is used to perform bolt fault diagnosis to obtain the bolt fault diagnosis results for the wind turbine tower. The steps include:

[0142] S4-4-1: Input the simulated sensitive modal component, the first measured sensitive modal component, and the second measured sensitive modal component into the bolt fault diagnosis model of the bolt fault diagnosis module;

[0143] S4-4-2: Using a feature extractor, extract the simulation data features of the simulation sensitive modal component, the first measured data features of the first measured sensitive modal component, and the second measured data features of the second measured sensitive modal component, respectively;

[0144] S4-4-3: Based on the dynamic attention weight values, a weighted fusion unit is used to weight and fuse the simulation data features, the first measured data features, and the second measured data features to obtain the weighted fused features;

[0145] S4-4-4: Based on the weighted fusion characteristics, a bolt fault diagnostic tool is used to perform bolt fault diagnosis and obtain the bolt fault diagnosis results of the wind turbine tower.

[0146] The bolt fault diagnosis results include loose bolts, bolt cracks, bolt corrosion, and bolt breakage of wind turbine tower bolts.

[0147] Example 2:

[0148] like Figure 2 As shown in the figure, this embodiment provides a wind turbine tower bolt fault diagnosis system based on multi-source data fusion, which is used to realize the wind turbine tower bolt fault diagnosis method. The system includes an architecture construction unit, a data acquisition unit, a data processing and fusion unit, and a bolt fault diagnosis unit connected in sequence.

[0149] An architecture building unit is used to construct a wind turbine tower bolt fault diagnosis architecture that includes a multi-source data acquisition layer, a multi-source data fusion layer, and a wind turbine tower bolt fault diagnosis layer connected in sequence.

[0150] The data acquisition unit is used to collect multi-source data of wind turbine tower bolts using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, and to preprocess the data to obtain multi-source bolt axial stress data.

[0151] The data processing and fusion unit is used to process and fuse multi-source bolt axial stress data using the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture to obtain multi-source sensitive modal components.

[0152] The bolt fault diagnosis unit is used in the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture to perform bolt fault diagnosis on multi-source sensitive signal components and obtain the bolt fault diagnosis results of the wind turbine tower.

[0153] This invention provides a method and system for fault diagnosis of wind turbine tower bolts based on multi-source data fusion. By collecting and fusing multi-source data (such as simulation data and measured stress data), it overcomes the limitations of traditional methods that rely on only a single vibration signal or a few measurement points. It can capture the structural state information of the tower from a broader perspective, making the diagnostic foundation more solid and comprehensive. Based on the wind turbine tower bolt fault diagnosis architecture, it performs automated data acquisition, processing, and fault diagnosis, avoiding manual inspection, reducing labor costs, minimizing human intervention, and improving fault diagnosis efficiency. Through steps such as wind load normalization, modal decomposition, clustering, and sensitive component screening, it performs in-depth processing of multi-source stress data, extracting modal components that are more sensitive to faults, removing redundant information and noise interference, and providing high-quality input for subsequent fault diagnosis. It is more sensitive to weak signal changes caused by early faults, enabling earlier identification of potential fault signs, achieving early warning, and gaining valuable time for maintenance decisions. More comprehensive information and a more intelligent fusion mechanism help distinguish real fault signals from background noise or normal fluctuations caused by changes in operating conditions, thereby reducing false alarms and missed alarms.

[0154] This invention is not limited to the optional embodiments described above, and anyone can derive other various forms of products based on the inspiration of this invention. The specific embodiments described above should not be construed as limiting the scope of protection of this invention; the scope of protection of this invention should be determined by the claims, and the specification can be used to interpret the claims.

Claims

1. A fan tower drum bolt fault diagnosis method based on multi-source data fusion, characterized by: The method comprises the following steps: Constructing a wind turbine tower bolt fault diagnosis architecture comprising a multi-source data data acquisition layer, a multi-source data fusion layer and a wind turbine tower bolt fault diagnosis layer connected in sequence; Using the data acquisition layer of the wind turbine tower bolt fault diagnosis architecture, collecting multi-source data of the wind turbine tower bolt, and preprocessing to obtain multi-source bolt axial stress data; Using the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture, processing and fusing the multi-source bolt axial stress data to obtain multi-source sensitive modal components; The multi-source sensitive modal components comprise simulation sensitive modal components corresponding to simulation bolt axial stress data, first measured sensitive modal components corresponding to first measured bolt axial stress data, and second measured sensitive modal components corresponding to measured bolt displacement data; Using the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture, performing bolt fault diagnosis on the multi-source sensitive signal components to obtain a wind turbine tower bolt fault diagnosis result, comprising the following steps: Inputting the multi-source sensitive modal components comprising the simulation sensitive modal components, the first measured sensitive modal components and the second measured sensitive modal components into the wind turbine tower bolt fault diagnosis layer of the wind turbine tower bolt fault diagnosis architecture; Using a correlation comparison module, obtaining correlation coefficients between the simulation sensitive modal components, the first measured sensitive modal components and the second measured sensitive modal components; Using a fault judgment module, judging whether the correlation coefficients are less than a correlation coefficient threshold value, if yes, entering the next step, triggering a bolt fault diagnosis process, otherwise, continuing data acquisition; According to the simulation sensitive modal components, the first measured sensitive modal components and the second measured sensitive modal components, using a bolt fault diagnosis module to perform bolt fault diagnosis to obtain a wind turbine tower bolt fault diagnosis result, comprising the following steps: Inputting the simulation sensitive modal components, the first measured sensitive modal components and the second measured sensitive modal components into a bolt fault diagnosis model of the bolt fault diagnosis module; Using a feature extractor, extracting simulation data features of the simulation sensitive modal components, first measured data features of the first measured sensitive modal components and second measured data features of the second measured sensitive modal components respectively; According to dynamic attention weight values, using a weighted fusion ware to perform weighted fusion on the simulation data features, the first measured data features and the second measured data features to obtain weighted fusion features; According to the weighted fusion features, using a bolt fault diagnosis ware to perform bolt fault diagnosis to obtain a wind turbine tower bolt fault diagnosis result.

2. The fan tower drum bolt fault diagnosis method based on multi-source data fusion according to claim 1, characterized in that: The method comprises the following steps: Constructing a tower bolt simulation model, and setting a bolt displacement data acquisition device and a bolt stress data acquisition device at the wind turbine tower; Constructing a multi-source data data acquisition layer, and connecting to the tower bolt simulation model, the bolt displacement data acquisition device and the bolt stress data acquisition device; A multi-source data fusion layer is constructed, and a wind load normalization module, a modal decomposition module, a modal component clustering module, and a sensitive component screening module are set up; A wind turbine tower bolt fault diagnosis layer is constructed, and a correlation comparison module, a fault judgment module, and a bolt fault diagnosis module are set up; The multi-source data data acquisition layer, the multi-source data fusion layer, and the wind turbine tower bolt fault diagnosis layer are sequentially connected to obtain the wind turbine tower bolt fault diagnosis architecture.

3. The fan tower drum bolt fault diagnosis method based on multi-source data fusion according to claim 2, characterized in that: The multi-source data includes simulated bolt axial stress data, measured bolt displacement data, and first measured bolt axial stress data.

4. The fan tower drum bolt fault diagnosis method based on multi-source data fusion according to claim 3, characterized in that: The data acquisition layer of the wind turbine tower bolt fault diagnosis architecture is used to collect multi-source data of the wind turbine tower bolt, and pre-processing is performed to obtain multi-source bolt axial stress data, including the following steps: Under the wind load and pre-tightening force, a bolt displacement data acquisition device is used to collect measured bolt displacement data of the wind turbine tower, and a bolt stress data acquisition device is used to collect corresponding first measured bolt axial stress data; Under the wind load and pre-tightening force, a tower bolt simulation model is used for simulation to obtain simulated bolt axial stress data; The data acquisition layer of the wind turbine tower bolt fault diagnosis architecture is used to collect simulated bolt axial stress data, measured bolt displacement data, and first measured bolt axial stress data to obtain multi-source data of the wind turbine tower bolt; The measured bolt displacement data is sequentially denoised and converted to obtain corresponding second measured bolt axial stress data; The simulated bolt axial stress data corresponding to the wind load and pre-tightening force, the first measured bolt axial stress data, and the second measured bolt axial stress data are integrated to obtain multi-source bolt axial stress data.

5. The fan tower drum bolt fault diagnosis method based on multi-source data fusion according to claim 4, characterized in that: The multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture is used to process and fuse the multi-source bolt axial stress data to obtain multi-source sensitive modal components, including the following steps: The multi-source bolt axial stress data is input into the multi-source data fusion layer of the wind turbine tower bolt fault diagnosis architecture; The wind load normalization module is used to normalize the multi-source bolt axial stress data to obtain wind load normalized multi-source bolt axial stress data; The modal decomposition module is used to decompose the wind load normalized multi-source bolt axial stress data to obtain multi-source modal components; The modal component clustering module is used to cluster the multi-source modal components to obtain a multi-source scatter plot; The sensitive component screening module is used to screen the multi-source modal components in the multi-source scatter plot to obtain multi-source sensitive modal components.

6. The fan tower drum bolt fault diagnosis method based on multi-source data fusion according to claim 5, characterized in that: The bolt fault diagnosis module is provided with a bolt fault diagnosis model, which is constructed based on an LSTM-Attention-MLP algorithm, and includes a feature extractor constructed based on an LSTM algorithm, a weighted fusion device constructed based on an Attention mechanism, and a bolt fault diagnosis device constructed based on an MLP algorithm, which are sequentially connected. 7.A wind turbine tower bolt fault diagnosis system based on multi-source data fusion, configured to implement the wind turbine tower bolt fault diagnosis method according to any one of claims 1-6. The system includes an architecture construction unit, a data acquisition unit, a data processing and fusion unit, and a bolt fault diagnosis unit, which are sequentially connected.

Images