An automatic fault location method for flange fastening bolts of wind turbine generator sets

Abstract

This invention discloses an automatic fault location method for wind turbine flange fastening bolts. The specific implementation steps include: First, ultrasonic sensors are uniformly installed at the flange connection points for multi-point synchronous data acquisition. The axial stress data is preprocessed using the quartile method, and the axial stress data of the same model unit under initial operating conditions is determined as test data. Combining the test data with real-time acquired data, the axial stress threshold is set and optimized using the 3-σ criterion and dynamic adjustment coefficient to achieve preliminary abnormal state identification. A mechanical model is constructed based on the stress condition of the flange fastening bolts. The axial stress distribution characteristics of the bolt group under different operating conditions are obtained using theoretical calculation methods. The axial stress distribution model is optimized using test data, and a multimodal matrix is ​​introduced for characterization to formulate fault assessment criteria. Finally, the stress distribution characteristics of unmonitored bolts are extracted using the weighted average method, and fault location is achieved based on the fault assessment criteria.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine generator condition monitoring, and specifically relates to an automatic fault location method for flange fastening bolts of wind turbine generators. Background Technology

[0002] With the continuous growth of global demand for renewable energy, wind power, as a clean and efficient energy conversion method, has been widely used. However, during long-term operation, especially under severe weather conditions, the reliability and safety of wind turbine components face serious challenges. Among these challenges, the condition monitoring and fault diagnosis of wind turbine flange fastening bolts, as important structural components connecting the blades, hub, and tower, are particularly crucial.

[0003] Traditional methods for monitoring the flange fastening bolts of wind turbines rely heavily on manual inspection and periodic maintenance. This is not only inefficient but also prone to missing faults and failing to detect potential problems in a timely manner. Furthermore, while current monitoring technologies are increasingly moving towards intelligent systems, accurate fault location remains a challenge due to complex operating conditions and diverse fault modes.

[0004] Currently, technologies applied to condition monitoring of wind turbine generators mainly include acoustic monitoring, vibration analysis, and temperature monitoring. However, these methods often cannot comprehensively and in real-time reflect the stress state of bolts and their potential faults, especially in large-scale wind farm environments. Therefore, research and optimization are urgently needed to achieve condition monitoring and fault location of wind turbine flange fastening bolts. Summary of the Invention

[0005] This invention belongs to the field of wind turbine generator condition monitoring, specifically relating to an automatic fault location method for wind turbine flange fastening bolts. The specific content is as follows: First, ultrasonic sensors are installed at the flange fastening bolts in a uniformly distributed manner for simultaneous multi-point data acquisition. Data preprocessing is performed based on a quartile algorithm. The preprocessed axial stress data under initial operating conditions is used as test data. Combined with the test data and real-time acquired data, a threshold is set and adjusted based on the 3-σ criterion and dynamic adjustment coefficient to achieve preliminary anomaly detection. Second, the stress condition of the flange fastening bolts is simplified into a mechanical model. The axial stress distribution under different operating conditions is obtained through theoretical calculations, and the bolt axial stress distribution model is further optimized based on the test data to obtain fault location criteria. Finally, the feature values ​​of unmonitored bolts are extracted using a weighted average method, and bolt fault location is achieved based on the fault location criteria. The overall framework of this invention is as follows: Figure 1 As shown.

[0006] The purpose of this invention is to provide an automatic fault location method for wind turbine flange fastening bolts, to solve the problems of low monitoring efficiency, high missed detection rate, and inaccurate location in the prior art, thereby improving the operational safety and reliability of wind turbine generator sets. This invention mainly includes three modules: data acquisition, automatic anomaly detection, and automatic fault location. The specific functions and implementation steps of each module are as follows:

[0007] The main function of the data acquisition module is to complete the tasks of data acquisition, storage, and preprocessing. The specific implementation steps are as follows:

[0008] S101: Using an ultrasonic sensor, the ultrasonic echo signal of the bolt axial stress is collected. After hardware and software filtering algorithms, noise is removed and the characteristic value of the ultrasonic echo of the bolt axial stress is accurately extracted, thereby calculating the real-time axial stress data of the bolt.

[0009] S102: The axial stress data obtained in S101 is preprocessed by screening according to the quartile method to remove the interfering data;

[0010] The main function of the automatic anomaly detection module is to set and adaptively adjust the bolt axial stress threshold based on the test dataset and real-time acquired data, and use the optimized threshold to achieve preliminary anomaly detection. The specific implementation steps are as follows:

[0011] S201: Based on the real-time wind data collected by the SCADA system and the preprocessing method described in S102, extract and merge the bolt axial stress data of the same type of unit under the initial operating state and the same wind conditions, and use the obtained bolt axial stress data as a test dataset after preprocessing.

[0012] S202: Using the collected test dataset, set the basic threshold according to the 3-σ criterion;

[0013] S203: Collect axial stress data for each monitored bolt, generate an axial stress time series, and introduce a threshold adjustment coefficient to adaptively adjust the axial stress threshold based on real-time data and actual working conditions. If the flange fastening bolt is affected by large load fluctuations, the threshold adjustment coefficient is increased to enhance the threshold's adaptability to fluctuations. For stable operating scenarios, the threshold adjustment coefficient is decreased to avoid excessive false alarms and achieve adaptive threshold adjustment.

[0014] The main functions of the automatic fault location module include: establishing an axial stress distribution model of flange fastening bolts as a fault assessment criterion; calculating the axial stress data of unmonitored bolts using a combined weighted average method and the geometric positional relationship between bolts; determining the axial stress threshold for unmonitored bolts; and finally locating the faulty bolts. The specific process is as follows: Figure 2 As shown, the implementation steps are as follows:

[0015] S301: Based on mechanical analysis methods, establish an axial stress distribution model to obtain bolt axial stress distribution models under different working conditions (normal, single-area loosening or fracture, and multi-area loosening or fracture);

[0016] S302: Based on the principle of multimodal matrix, the stress distribution characteristics of flange fastening bolt group under different working conditions are characterized to obtain the corresponding bolt axial stress distribution matrix, which is the flange fastening bolt failure assessment criterion.

[0017] S303: Based on real-time data, the axial stress estimate of unmonitored bolts is automatically extracted using the weighted average method and the geometric positional relationship between bolts;

[0018] S304: Obtain the axial stress threshold for each bolt based on the threshold setting and optimization algorithm described in S4 and S5;

[0019] S305: By combining the axial stress data of each bolt, the axial stress threshold, and the fault assessment criteria, automatic fault location is performed, and the fault type is determined.

[0020] In summary, this invention achieves real-time, full-process monitoring of the condition of wind turbine flange fastening bolts using ultrasonic sensors, overcoming the limitations of traditional methods. It employs a dataset of initial operating conditions and wind conditions from the same type of turbine as the test dataset, providing bolt axial stress thresholds that better reflect the actual operating conditions of wind turbines. Furthermore, by combining the 3-σ criterion, threshold adjustment coefficient, mechanical model, and multimodal matrix, it provides more scientific and accurate fault assessment criteria, significantly improving monitoring accuracy. The introduction of a weighted average method for feature extraction effectively expands the monitoring range and enhances the ability to locate faults in unmonitored bolts. After implementing this invention, the condition of wind turbine flange fastening bolts can be monitored in real time, accurately locating potential fault points, reducing safety hazards and economic losses caused by bolt failures, and improving the overall operating efficiency and reliability of wind turbines. Transferring bolt fault feature acquisition to the software algorithm level is the foundation for automated fault bolt location, contributing to the automation and intelligence of flange fastening bolt condition monitoring and health assessment systems. Attached Figure Description

[0021] Appendix Figure 1 This is a schematic diagram of the overall framework of the present invention;

[0022] Appendix Figure 2 This is a flowchart of the automatic anomaly detection module of the present invention;

[0023] Appendix Figure 3 This is a flowchart of the automatic fault location module of the present invention. Detailed Implementation

[0024] The data acquisition module comprises three sub-modules: data acquisition, data storage, and data preprocessing. The data acquisition sub-module uses an ultrasonic sensor to collect the ultrasonic echo signal of the bolt's axial stress. Through hardware and software filtering algorithms, noise is removed, and the characteristic values ​​of the bolt preload ultrasonic echo are accurately extracted, thereby calculating the real-time axial stress data of the bolt. The data storage sub-module uses SQL database language to store the axial stress data from each monitoring point of the flange fastening bolt in a database. The data preprocessing sub-module uses the quartile method to filter and preprocess the obtained axial stress data, removing interfering data and improving the signal-to-noise ratio.

[0025] The automatic anomaly detection module comprises three sub-modules: test dataset, threshold adjustment, and anomaly detection. The specific process is as follows: Figure 2 As shown. The test dataset is collected from the initial operating state of the flange fastening bolt group and then preprocessed; the threshold adjustment submodule requires the combined action of the test dataset and real-time acquired data, and automatically adjusts the threshold based on the feature values ​​extracted by the 3-σ criterion and threshold adjustment coefficient; the anomaly detection submodule completes the preliminary anomaly identification based on the adjusted threshold and real-time axial stress data.

[0026] The automatic fault location module mainly consists of four sub-modules: establishing an axial stress distribution model based on mechanical analysis methods, fault assessment criteria, feature extraction of unmonitored bolts, fault location, and output conclusions. The specific process is as follows: Figure 3 As shown, the unmonitored bolt feature extraction submodule automatically extracts the estimated axial stress value of unmonitored bolts based on real-time data and a weighted average method; the axial stress distribution model and fault assessment criterion submodule based on mechanical analysis methods obtains the corresponding axial stress distribution characteristics under different working conditions through the axial stress distribution model, and uses a multimodal matrix to characterize the axial stress distribution characteristics of the flange fastening bolt group under different working conditions, and uses it as a fault assessment criterion. This criterion is the key to achieving automatic fault location.

[0027] First, the collected dataset is preprocessed. The axial stress dataset is sorted in ascending order, divided into four equal parts, and the quartiles are calculated. The quartiles are the data at the three nodes after each division, and the corresponding formulas are as follows:

[0028]

[0029] In the formula

[0030] This represents the first quantile of the dataset at the j-th measurement point;

[0031] This represents the second quantile of the dataset at the j-th measurement point;

[0032] This represents the third quantile of the dataset at the j-th measurement point;

[0033] n represents the total number of data points contained in the dataset;

[0034] and Let represent the (n+1) / 2, n / 2, and (n+2) / 2 numbers in the dataset of the j-th measurement point, respectively;

[0035] m is a positive integer, and its size is determined by the value of n;

[0036] and The m-th, m+1-th, m+2-th, 3m+1-th and 3m+3-th numbers in the dataset of the j-th measurement point, respectively;

[0037] IQR represents the interquartile range of the dataset;

[0038] This represents the lower boundary of the normal values ​​in the dataset of the j-th measurement point;

[0039] This represents the upper boundary of the normal values ​​in the dataset of the j-th measurement point;

[0040] Furthermore, based on the real-time wind data collected by the SCADA system and the preprocessing method, the preprocessed bolt axial stress data of the same type of unit under initial operating conditions and the same wind conditions are merged into a test dataset. Based on the test dataset and the real-time collected data, basic thresholds are set and adjusted, and preliminary anomaly identification is performed. The specific implementation steps are as follows:

[0041] Step 1: Using the collected test dataset, set the basic threshold according to the 3-σ criterion, as shown in the following formula:

[0042]

[0043] In the formula

[0044] This represents the test axial stress dataset at the j-th measuring point, where n is the total number of data points.

[0045] μ j This represents the mean of the test dataset at the j-th measurement point;

[0046] σ j Let represent the variance of the test dataset at the j-th measurement point;

[0047] This represents the basic threshold of the test dataset for the j-th measurement point.

[0048] Step 2: Collect axial stress data for each monitored bolt and generate an axial stress time series. Where N is the number of sensors, the axial stress threshold is dynamically adjusted based on real-time data acquisition and actual working conditions, and the corresponding formula is as follows:

[0049]

[0050] In the formula

[0051] This represents the real-time average axial stress amplitude of the j-th monitored bolt;

[0052] This represents the real-time axial stress dataset at the j-th measuring point;

[0053] α j This represents the threshold adjustment coefficient for the j-th measurement point;

[0054] This represents the maximum value of the axial stress data at the j-th measuring point;

[0055] This represents the minimum value of the axial stress data at the j-th measuring point;

[0056] τ j (t) represents the threshold value after adjusting the axial stress data of the bolt at the j-th measuring point at time t.

[0057] Furthermore, the key to fault location lies in determining the fault assessment criteria and establishing an axial stress distribution model based on mechanical analysis methods. This allows for the acquisition of corresponding axial stress distribution characteristics under different working conditions. The corresponding model formula is as follows:

[0058]

[0059] In the formula

[0060] F z Indicates the axial load of the flange fastening;

[0061] F A(Z)i Indicates in F z The axial force on the i-th bolt under action;

[0062] n s Indicates the number of fastening bolts on a single flange;

[0063] (x i y i () indicates the coordinates of the center of gravity of the flange fastening bolt group;

[0064] I xx This represents the moment of inertia of the flange fastening bolt assembly relative to the swing direction;

[0065] I yy This represents the moment of inertia of the flange fastening bolt assembly relative to the direction of oscillation.

[0066] M x This indicates the bending moment of the flange fastening bolt assembly in the swing direction;

[0067] M y This indicates the bending moment of the flange fastening bolt assembly in the direction of oscillation.

[0068] F A(M)i Indicates in M x and M y The axial force on the i-th bolt under action;

[0069] P i This represents the preload force of the i-th bolt;

[0070] C ib This represents the stiffness of the i-th bolt;

[0071] C im This represents the stiffness of the i-th bolted connection.

[0072] F Ai This represents the total axial force of the i-th bolt;

[0073] Indicates the number of remaining fastening bolts on a single flange under fracture conditions;

[0074] (x * y * () indicates the coordinates of the center of gravity of the flange fastening bolt group under fracture conditions;

[0075] This represents the moment of inertia of the flange fastening bolt assembly relative to the swing direction under fracture conditions.

[0076] This represents the moment of inertia of the flange fastening bolt assembly relative to the direction of oscillation.

[0077] M zx This indicates that the shaft load F is applied due to eccentricity. z The additional waving direction bending moment M generated at that time zx ;

[0078] M zy This indicates that the shaft load F is applied due to eccentricity. z The additional oscillation direction bending moment M generated at that time zx ;

[0079] When indicating the fracture condition, in F z The axial force on the i-th bolt under action;

[0080] When indicating a fracture condition, in M x and M y The axial force on the i-th bolt under action;

[0081] This represents the total axial force of the i-th bolt under fracture conditions;

[0082] Furthermore, based on multimodal matrix theory, the axial stress distribution characteristics of the flange fastening bolt group under different working conditions are characterized. The feature is that the rows of the multimodal matrix represent different working conditions, and the columns represent the axial stress values ​​of each bolt. The specific formula is as follows:

[0083]

[0084] In the formula

[0085] s represents the multimodal matrix of the axial stress distribution characteristics of the flange fastening bolt group under different working conditions;

[0086] [M1...M k ...M L [] indicates different operating conditions during the operation of the flange fastening bolt group, and L indicates the total number of operating conditions;

[0087] This represents the axial stress vector of the flange fastening bolt group when operating under the k-th type of working condition.

[0088] Furthermore, based on real-time data and test datasets, the axial stress distribution characteristics of unmonitored bolts are automatically extracted using the weighted average method and the geometric positional relationship between bolts. The specific extraction algorithm is as follows:

[0089]

[0090] In the formula

[0091] ω ij The correlation between sensor i and the unmonitored bolt j is represented by the weight.

[0092] d ij This represents the Euclidean distance between sensor i and the unmonitored bolt j;

[0093] N represents the number of sensors in a single flange;

[0094] x i (t) represents the measured value of the axial stress of the monitored bolt i at time t;

[0095] This represents the estimated axial stress of bolt j at time t when it was not monitored.

[0096] Finally, by combining fault assessment criteria, the axial stress distribution characteristics of unmonitored bolts, and a model with adjusted and set thresholds, automatic fault location is achieved. The implementation steps are as follows:

[0097] Step 1: Establish an axial stress distribution model based on mechanical analysis methods to obtain bolt axial stress distribution models under different working conditions;

[0098] Step 2: Based on the principle of multimodal matrix, the stress distribution characteristics of the flange fastening bolt group under different working conditions are characterized to obtain the multimodal matrix of bolt axial stress distribution. This matrix is ​​used as the fault assessment criterion for flange fastening bolts.

[0099] Step 3: Based on the preprocessed axial stress data, the estimated axial stress of the unmonitored bolts is automatically extracted using the weighted average method and the geometric positional relationship between the bolts.

[0100] Step 4: Based on the existing threshold setting and optimization model, set and adaptively adjust the axial stress threshold of each bolt;

[0101] Step 5: Combine the axial stress data of each bolt, the axial stress threshold, and the fault assessment criteria to automatically locate the fault and determine the fault type.

Claims

1. An automatic fault location method for flange fastening bolts of a wind turbine generator set, characterized in that, Includes the following steps: Step 1.1: Data acquisition and preprocessing. Ultrasonic sensors are used to acquire ultrasonic echo signals of bolt axial stress. Hardware and software filtering algorithms are used to remove noise and accurately extract ultrasonic echo feature values. Real-time axial stress data of bolts are calculated. Then, the quartile method is used to filter and preprocess the axial stress data to remove interfering data. Step 1.2: Automatic anomaly detection. A test dataset is constructed based on the real-time wind data collected by the SCADA system and the preprocessing method in Step 1.

1. The basic threshold is set by the 3-σ criterion, and an adaptive threshold adjustment coefficient is introduced in combination with the real-time collected data to complete the anomaly detection. Step 1.3: Automatic fault location. Based on mechanical analysis methods, an axial stress distribution model under different working conditions is established. A multimodal matrix is ​​introduced to characterize the stress distribution characteristics of the bolt group and serve as a fault assessment criterion. The axial stress distribution characteristics of unmonitored bolts are extracted using the weighted average method and the geometric positional relationship between bolts. By combining the axial stress data of all bolts, the adaptively adjusted threshold, and the fault assessment criterion, automatic fault location and fault type determination are achieved.

2. The automatic fault location method for flange fastening bolts of a wind turbine generator set according to claim 1, characterized in that, The specific process of data acquisition and preprocessing in step 1.1 is as follows: Step 2.1: Use an ultrasonic sensor to collect the ultrasonic echo signal of the bolt's axial stress; Step 2.2: Using hardware and software filtering algorithms, noise is removed and the characteristic values ​​of the ultrasonic echo of the bolt axial stress are accurately extracted, thereby calculating the real-time axial stress data of the bolt. Step 2.3: The axial stress data obtained in Step 2.2 is preprocessed using the quartile method to remove interfering data. The corresponding formula is as follows: In the formula This represents the first quantile of the dataset at the j-th measurement point; This represents the second quantile of the dataset at the j-th measurement point; This represents the third quantile of the dataset at the j-th measurement point; n represents the total number of data points contained in the dataset; , and Let represent the (n+1) / 2, n / 2, and (n+2) / 2 numbers in the dataset of the j-th measurement point, respectively; m is a positive integer, and its size is determined by the value of n; , , , , and The m-th, m+1-th, m+2-th, 3m+1-th, and 3m+3-th numbers in the dataset of the j-th measurement point, respectively; Represents the interquartile range of the dataset; This represents the lower boundary of the normal values ​​in the dataset of the j-th measurement point; This represents the upper boundary of the normal values ​​in the dataset of the j-th measurement point.

3. The automatic fault location method for flange fastening bolts of a wind turbine generator set according to claim 1, characterized in that, The specific process of constructing the test dataset and adjusting the threshold in step 1.2 is as follows: (1) Test dataset construction: Wind speed and wind direction information are collected by SCADA system and input into flange fastening bolt monitoring system. Bolt axial stress data in the same wind speed and wind direction range under the initial operating state of each unit in the wind farm are extracted. After preprocessing, the bolt axial stress data of the same type of unit under the same wind conditions are merged to obtain the test dataset. (2) Threshold adjustment: Step a: Using the collected test dataset, set the basic threshold according to the 3-σ criterion, as shown in the following formula: In the formula This represents the test axial stress dataset at the j-th measuring point; This represents the mean of the test dataset at the j-th measurement point; Let represent the variance of the test dataset at the j-th measurement point; This represents the basic threshold of the test dataset for the j-th measurement point; Step b: Collect axial stress data for each monitored bolt and generate an axial stress time series. Where N is the number of sensors, a threshold adjustment coefficient is introduced to adaptively adjust the axial stress threshold based on real-time data acquisition and actual working conditions. If the flange fastening bolts are affected by large load fluctuations, the threshold adjustment coefficient is increased to enhance the threshold's adaptability to fluctuations. For stable operating scenarios, the threshold adjustment coefficient is decreased to avoid excessive false alarms, thus achieving adaptive threshold adjustment. The corresponding formula is as follows: In the formula This represents the real-time average axial stress amplitude of the j-th monitored bolt; This represents the real-time axial stress dataset at the j-th measuring point; This represents the threshold adjustment coefficient for the j-th measurement point; This represents the maximum value of the axial stress data at the j-th measuring point; This represents the minimum value of the axial stress data at the j-th measuring point; This represents the threshold value after adjusting the axial stress data of the bolt at the j-th measuring point at time t.

4. The automatic fault location method for flange fastening bolts of a wind turbine generator set according to claim 1, characterized in that, The process of establishing the axial stress distribution model and constructing the multimodal matrix in step 1.3 is as follows: (1) Establishment of axial stress distribution model: In the formula Indicates the axial load of the flange fastening; Indicates in The axial stress of the i-th bolt under action; Indicates the number of fastening bolts on a single flange; ( , () indicates the coordinates of the center of gravity of the flange fastening bolt group; This represents the moment of inertia of the flange fastening bolt assembly relative to the swing direction; This represents the moment of inertia of the flange fastening bolt assembly relative to the direction of oscillation. This indicates the bending moment of the flange fastening bolt assembly in the swing direction; This indicates the bending moment of the flange fastening bolt assembly in the direction of oscillation. Indicates in and The axial stress of the i-th bolt under action; This represents the preload force of the i-th bolt; This represents the stiffness of the i-th bolt; This represents the stiffness of the i-th bolted connection. Denote the total axial stress of the i-th bolt; Indicates the number of remaining fastening bolts on a single flange under fracture conditions; ( , () indicates the coordinates of the center of gravity of the flange fastening bolt group under fracture conditions; This represents the moment of inertia of the flange fastening bolt assembly relative to the swing direction under fracture conditions. This represents the moment of inertia of the flange fastening bolt assembly relative to the direction of oscillation. This indicates that the shaft load F is applied due to eccentricity. z The additional waving direction bending moment M generated at that time zx ; This indicates that the shaft load F is applied due to eccentricity. z The additional oscillation direction bending moment M generated at that time zx ; When indicating fracture conditions The axial force on the i-th bolt under action; When indicating fracture conditions and The axial force on the i-th bolt under action; This represents the total axial force of the i-th bolt under fracture conditions; (2) Construction of multimodal matrix: In the formula A multimodal matrix representing the axial stress distribution characteristics of flange fastening bolt groups under different working conditions; This indicates different operating conditions during the operation of the flange fastening bolt group, where L represents the total number of operating conditions. This represents the axial stress vector of the flange fastening bolt group when operating under the k-th type of working condition.

5. The automatic fault location method for flange fastening bolts of a wind turbine generator set according to claim 1, characterized in that, The specific algorithm for extracting the features of unmonitored bolts in step 1.3 is as follows: In the formula The correlation between sensor i and the unmonitored bolt j is represented by the weight. This represents the Euclidean distance between sensor i and the unmonitored bolt j; N represents the number of sensors per blade; This represents the measured axial stress value of the monitored bolt i at time t; This represents the estimated axial stress of bolt j at time t when it was not monitored.

6. The automatic fault location method for flange fastening bolts of a wind turbine generator set according to claim 1, characterized in that, The specific implementation process of automatic fault location in step 1.3 is as follows: First, model data under different working conditions is obtained through the axial stress distribution model established in claim 4. Then, the fault assessment criteria are determined through the multimodal matrix constructed in claim 4. The axial stress dataset of all bolts is obtained by combining the feature extraction algorithm of unmonitored bolts in claim 5. The axial stress threshold of each bolt is set and adaptively adjusted according to the threshold adjustment algorithm in claim 3. Finally, the axial stress data of all bolts, the axial stress threshold and the fault assessment criteria are combined to complete the automatic fault location and fault type determination.

Images