Fan variable pitch bearing intelligent monitoring method and system based on multi-parameter fusion

Abstract

The invention discloses a fan variable pitch bearing intelligent monitoring method and system based on multi-parameter fusion, and the method comprises the steps: respectively calculating angles based on a three-axis acceleration and a three-axis angular velocity, and obtaining a final angle through complementary filtering fusion; carrying out ensemble empirical mode decomposition on the three-axis vibration signals, respectively extracting low-frequency vibration characteristics, high-frequency vibration characteristics and angle variation based on a plurality of signal components, carrying out fusion decision on the low-frequency vibration characteristics, the high-frequency vibration characteristics and the angle variation by adopting a D-S evidence theory to obtain a total confidence coefficient, and judging a fault in combination with a preset fault threshold to obtain a first fault diagnosis result; the low-frequency vibration characteristics, the high-frequency vibration characteristics and the angle variation are input into a deep residual network diagnosis model for processing, and a second fault diagnosis result is obtained; and comparing the first fault diagnosis result with the second fault diagnosis result, outputting a final diagnosis result if the first fault diagnosis result is consistent with the second fault diagnosis result, and performing diagnosis again if the second fault diagnosis result is inconsistent. The limitation of a single sensor frequency band is broken through, a D-S evidence theory is innovatively adopted to carry out multi-source evidence fusion decision, and the fault recognition rate is improved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine generator condition monitoring technology, and in particular to a method and system for intelligent monitoring of wind turbine pitch bearings based on multi-parameter fusion. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Pitch bearings are critical components of wind turbine generators, and their health directly affects the turbine's operational safety and efficiency. During operation, pitch bearings are prone to failures such as fatigue spalling, wear, fracture, and cage damage. However, traditional monitoring methods are limited by installation methods and cannot simultaneously cover both high-frequency and low-frequency defects and damage.

[0004] Currently, existing pitch system monitoring mainly employs single-frequency monitoring technology, relying on a single-frequency sensor, which makes it difficult to comprehensively reflect the three-dimensional motion state of the pitch bearing. During pitch control, the force points frequently switch, and a single sensor cannot fully perceive the information, easily leading to insufficient accuracy in fault diagnosis. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides a method and system for intelligent monitoring of wind turbine pitch bearings based on multi-parameter fusion. It uses spatiotemporal synchronous acquisition of triaxial vibration and triaxial angular velocity, combined with fusion analysis technology for fault monitoring.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention provides an intelligent monitoring method for wind turbine pitch bearings based on multi-parameter fusion, comprising: The dynamic parameters of the pitch bearing of the wind turbine to be monitored are obtained, including triaxial acceleration and triaxial angular velocity. The angle is calculated based on the triaxial acceleration and triaxial angular velocity, and the angles obtained are then fused by complementary filtering based on adaptive filtering coefficients to obtain the final angle. Ensemble empirical mode decomposition is performed on the triaxial vibration signal to generate multiple signal components. Based on the signal components, low-frequency vibration features, high-frequency vibration features, and angle changes are extracted respectively. The DS evidence theory is used to fuse low-frequency vibration characteristics, high-frequency vibration characteristics and angle change to obtain the total confidence. Combined with the preset fault threshold, the fault is judged to obtain the first fault diagnosis result. The low-frequency vibration characteristics, high-frequency vibration characteristics, and angle change are input into the deep residual network diagnostic model for processing to obtain the second fault diagnosis result. The second fault diagnosis result is compared with the first fault diagnosis result. If they are consistent, the final diagnosis result is output. If they are inconsistent, the diagnosis is repeated.

[0007] A further technical solution involves complementary filtering and fusion of triaxial acceleration and triaxial angular velocity, specifically as follows: The acquired triaxial acceleration and triaxial angular velocity signals were preprocessed to obtain the preprocessed triaxial acceleration and triaxial angular velocity; Calculate the angle based on the preprocessed triaxial acceleration, and calculate the angle increment integral based on the preprocessed triaxial angular velocity; Complementary filtering fusion based on angle and angle increment integral yields the fused dynamic parameters.

[0008] A further technical solution is that the preprocessing includes performing gravity field normalization on the triaxial acceleration signal and applying zero-bias compensation to the triaxial angular velocity signal.

[0009] A further technical solution, expressed from a computational perspective, is as follows: ,

[0010] in, Indicates the roll angle. Indicates the pitch angle.

[0011] A further technical solution is to calculate the integral of the angle increment as follows: ,

[0012] in, This indicates the roll angle at the last sampling time. This indicates the pitch angle at the last sampling time. express Angular velocity in the axial direction, express Angular velocity in the axial direction, Indicates the sampling time interval.

[0013] In a further technical solution, both the low-frequency vibration characteristics and the high-frequency vibration characteristics include time-domain characteristics and frequency-domain characteristics.

[0014] A further technical solution involves using DS evidence theory to fuse low-frequency vibration characteristics, high-frequency vibration characteristics, and angular changes for decision-making. Specifically: Multiple sources of evidence are defined based on the input data; Define an identification framework, construct a basic probability assignment function for each source of evidence, and assign basic probabilities. The confidence level is calculated based on the basic probability of each evidence source, and a fusion decision is made based on the confidence levels of each evidence source to obtain the total confidence level.

[0015] Secondly, the present invention provides an intelligent monitoring system for wind turbine pitch bearings based on multi-parameter fusion, comprising: The multi-source data acquisition module is configured to acquire the dynamic parameters of the pitch bearing of the wind turbine to be monitored, including triaxial acceleration and triaxial angular velocity. An angle fusion calculation module is configured to: calculate the angle based on the triaxial acceleration and triaxial angular velocity respectively, and perform complementary filtering fusion on the obtained angles based on adaptive filtering coefficients to obtain the final angle; The feature extraction module is configured to: perform ensemble empirical mode decomposition on the triaxial vibration signal to generate multiple signal components, and extract low-frequency vibration features, high-frequency vibration features and angle changes based on the signal components respectively; The first fault diagnosis module is configured to: use DS evidence theory to fuse low-frequency vibration characteristics, high-frequency vibration characteristics and angle change to obtain the total confidence level, and combine it with the preset fault threshold to judge the fault and obtain the first fault diagnosis result. The second fault diagnosis module is configured to: input low-frequency vibration characteristics, high-frequency vibration characteristics, and angle change into the deep residual network diagnosis model for processing to obtain a second fault diagnosis result; compare the second fault diagnosis result with the first fault diagnosis result; if they are consistent, output the final diagnosis result; if they are inconsistent, perform a re-diagnosis.

[0016] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent monitoring method for wind turbine pitch bearings based on multi-parameter fusion as described in the first aspect.

[0017] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the intelligent monitoring method for wind turbine pitch bearings based on multi-parameter fusion as described in the first aspect.

[0018] The above one or more technical solutions have the following beneficial effects: This invention perceives the motion state of the wind turbine pitch bearing from all dimensions, simultaneously acquiring the bearing's six degrees of freedom motion parameters, breaking through the bandwidth limitations of a single sensor. Furthermore, it employs ensemble empirical mode decomposition to directionally extract high-frequency and low-frequency vibration features, and innovatively uses DS evidence theory for multi-source evidence fusion decision-making, thereby improving the fault identification rate.

[0019] This invention collects multi-source data from multiple sources and solves the problem of difficult multi-source signal coupling, thus improving the accuracy of fault identification. Furthermore, it incorporates the vibration spectrum characteristics of the wind turbine's main drive chain, adapting to a bandwidth of 0-70kHz.

[0020] This invention innovatively adopts a set of empirical mode decomposition methods, which can effectively decompose high and low frequency signal features. Through mode reconstruction and multi-parameter fusion, it significantly improves the accuracy of fault identification and ensures that the health status of bearings can still be comprehensively monitored under complex operating conditions.

[0021] This invention innovatively combines multiple fault diagnosis methods, avoiding misjudgments and false detections caused by a single diagnostic method, and improving the accuracy of fault diagnosis. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of the intelligent monitoring method for wind turbine pitch bearings based on multi-parameter fusion, according to an embodiment of the present invention. Detailed Implementation

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0026] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0027] Example 1 like Figure 1As shown in the figure, this embodiment discloses an intelligent monitoring method for wind turbine pitch bearings based on multi-parameter fusion. The method includes the following steps: S1: Obtain the dynamic parameters of the pitch bearing of the wind turbine to be monitored, including triaxial acceleration and triaxial angular velocity; In this embodiment, dynamic parameters, including triaxial acceleration (...), are collected using a triaxial accelerometer and a triaxial gyroscope. , , ) and triaxial angular velocity ( , , ), sampling frequency is A triaxial accelerometer is deployed at key measuring points in the bearing load area of ​​the wind turbine pitch bearing (such as the bearing housing flange connection) to directly measure the linear vibration acceleration signal of the bearing in a three-dimensional orthogonal coordinate system (X / Y / Z axes). At the same time, a triaxial gyroscope is installed at the same location as the accelerometer to measure the angular velocity components of the bearing around the three-dimensional orthogonal axes.

[0028] The accelerometer and gyroscope use the same hardware clock source to trigger sampling, ensuring strict time alignment of the six-axis data.

[0029] S2: Calculate the angles based on the triaxial acceleration and triaxial angular velocity respectively, and perform complementary filtering and fusion on the obtained angles based on adaptive filtering coefficients to obtain the final angle; In this embodiment, S201: The collected triaxial acceleration and triaxial angular velocity signals are preprocessed to obtain the preprocessed triaxial acceleration and triaxial angular velocity.

[0030] (1) After detecting the stationary state of the device or carrier to which the sensor is attached, the collected triaxial acceleration signals (acc, acceleration, representing the X, Y, and Z axis linear acceleration components measured by the accelerometer) are normalized by gravity field to eliminate the reference deviation caused by the sensor's installation tilt and ensure that the gravity direction is always aligned with the Z axis. The stationary state detection is determined by the following formula: and

[0031] in, , and These are triaxial acceleration components. , and These are the three-axis angular velocity components. A preset threshold is used. After detecting a stationary state, the offset is calculated and updated. The formula is:

[0032] in, Indicates the current time The reference deviation angle, The attitude angle measured by the accelerometer. This is the integral value of the current angular velocity of the gyroscope.

[0033] A dynamic reference correction algorithm is used to apply zero-bias compensation to the three-axis angular velocity signals (gyro and gyroscope represent the X, Y, and Z axis angular velocity components measured by the gyroscope, respectively) to remove drift. The correction formula is:

[0034] in, The current attitude angle, The initial attitude angle is... The angular velocity measured by the gyroscope. This is the zero bias value estimated dynamically.

[0035] (2) Traditional linear compensation ignores the nonlinear error accumulated by integration. This invention adopts an improved nonlinear integral compensation method, which introduces a quadratic nonlinear term. The compensation formula is as follows:

[0036] in, The nonlinear compensation coefficients are fitted based on historical drift data. In this embodiment... The nonlinear compensation coefficient is optimized for the periodic load of the pitch bearing. Compared with the traditional linear integral, this mechanism reduces long-term drift error by 25%, improves the stability of the monitoring system, and is suitable for long-term wind turbine operation scenarios.

[0037] (3) Detect the stationary state at set time intervals and update the offset. .

[0038] S202: Accelerometer angle calculation. Angles are calculated based on the pre-processed triaxial acceleration, including roll and pitch angles, and are expressed as follows: ,

[0039] in, Indicates the roll angle. Indicates the pitch angle.

[0040] S203: Gyroscope integration. Angles are calculated based on the pre-processed three-axis angular velocities, including roll and pitch angles, expressed as: ,

[0041] in, This represents the roll angle at the last sampling time, and this represents the estimated value of the gyroscope at the previous time step. This represents the pitch angle at the last sampling time, and this represents the gyroscope's estimated value at the previous time step. express Angular velocity in the axial direction, express Angular velocity in the axial direction, This represents the sampling time interval, i.e., the time difference between two samples.

[0042] S204: Based on adaptive filter coefficients Complementary filtering fusion is represented as:

[0043]

[0044] in, This represents the roll angle after complementary filtering and fusion, which is the final output angle. It represents the roll angle calculated based on gyroscope integration, reflecting dynamic response characteristics; It represents the roll angle calculated based on the accelerometer, reflecting steady-state accuracy. In this embodiment, the adjustment factor is represented. ; These represent the adaptive filter coefficients, used to balance the dynamic response of the gyroscope and the steady-state accuracy of the accelerometer, based on the pitch angular velocity change rate. Dynamic adjustments are made to optimize drift suppression in high-dynamic scenarios. The original complementary filter uses a fixed... The invention addresses the issue of introducing integral errors during pitch acceleration. It designs adaptive filter coefficients to ensure the steady-state accuracy of the accelerometer at low speeds and enhance the dynamic response of the gyroscope at high speeds.

[0045] Compared to traditional fixed-coefficient complementary filtering, this adaptive mechanism reduces angle estimation error by up to 20% when pitch speed changes drastically, improving the robustness of the six-degree-of-freedom model in dynamic environments. It is suitable for high-frequency switching scenarios involving wind turbine pitch bearings.

[0046] The above technical solution eliminates measurement errors caused by pitching motion by complementary filtering and fusion of triaxial vibration and triaxial angular velocity.

[0047] S3: Perform ensemble empirical mode decomposition on the triaxial vibration signal to generate multiple signal components, and extract time-domain features and frequency-domain features based on the signal components respectively; In this embodiment, the triaxial vibration signal is acquired in real time by a triaxial vibration sensor (such as an accelerometer) installed on the equipment (such as a mechanical system, vehicle or industrial equipment), capturing the vibration data of the equipment on the X, Y and Z axes. Attitude compensation can be performed based on the fusion angle calculated by S202 to correct the signal direction deviation.

[0048] For triaxial vibration signals Perform ensemble empirical mode decomposition (EEMD) to generate Each signal component (IMF component) is represented as:

[0049] in, This represents the residual, i.e., the remaining signal components that cannot be further decomposed after ensemble empirical mode decomposition (EEMD). Based on these signal components, time-domain features (such as mean, peak value, and variance) and frequency-domain features (such as power spectral density and dominant frequency) are extracted respectively. Low-frequency vibration features correspond to low-order IMF components (IMF1-IMF3, typically containing low-frequency signals <100Hz); high-frequency vibration features correspond to high-order IMF components (IMF4-IMF6, typically containing high-frequency signals >100Hz), and are classified according to the IMF intrinsic frequency order or threshold.

[0050] Traditional ensemble empirical mode decomposition uses fixed white noise amplitude, which easily leads to mode aliasing. This invention introduces adaptive noise amplitude, expressed as:

[0051] in, In this embodiment, the scaling factor is represented. Noise injection is dynamically adjusted based on the signal RMS to reduce high-frequency interference in low-frequency components. Adaptive noise injection improves decomposition accuracy by 15%, reduces mode aliasing, ensures that high-frequency IMF more accurately captures bearing microcrack signals, and enhances the reliability of subsequent fusion.

[0052] Time-domain features (peak value, kurtosis) and frequency-domain features (spectral energy, characteristic frequency) are extracted from the IMF components. Each IMF component represents a frequency component in the triaxial vibration signal and is arranged in order from high frequency to low frequency, such as IMF1, IMF2, etc.

[0053] S4: The DS evidence theory is used to fuse low-frequency vibration characteristics, high-frequency vibration characteristics and angle change to obtain the total confidence level. Combined with the preset fault threshold, the fault is judged to obtain the first fault diagnosis result.

[0054] In this embodiment, an improved DS evidence theory is used to fuse low-frequency vibration characteristics, high-frequency vibration characteristics, and gyroscope angle change rate to output fault confidence.

[0055] The specific steps are as follows: Define the evidence sources, including: Evidence 1: Low-frequency vibration characteristics (time-domain and frequency-domain characteristics extracted from IMF1-IMF3 components), used to point to structural deterioration faults; Evidence 2: High-frequency vibration characteristics (time-domain and frequency-domain characteristics extracted from IMF4-IMF6 components), used to point to local damage; Evidence 3: Gyroscope angle change rate ( , of which molecules This is the change in the fused attitude angle calculated based on S202, used to indicate motion instability, i.e., motion jamming or unevenness. The basic probability allocation function (m) of the evidence is fused through the DS synthesis rule, ultimately outputting a fault confidence level (e.g., structural degradation 0.7, local damage 0.2, motion instability 0.1). Based on this output, the fault diagnosis result is obtained by comparing it with a confidence threshold (>0.6), triggering corresponding alarms or maintenance actions.

[0056] Define the identification framework A basic probability assignment function (BPA) is constructed for each source of evidence, and a basic probability is assigned.

[0057] The confidence level is calculated and expressed as: , ,

[0058] in, This indicates the contribution of low-frequency signals (IMF1~IMF3) to fault diagnosis; The power or intensity representing the characteristics of low-frequency vibrations is used to calculate confidence levels; The noise standard deviation representing the characteristics of low-frequency vibration; This indicates the contribution of high-frequency signals (IMF4~IMF6) to fault diagnosis; Power or intensity representing high-frequency vibration characteristics, used to calculate confidence levels; The noise standard deviation representing the characteristics of high-frequency vibration; This represents the confidence level after normalization of the gyroscope's angle change rate, reflecting the dynamic characteristics of the angle change. The change in angle is calculated from gyroscope data at consecutive sampling times; This represents the sampling time interval.

[0059] Traditional DS evidence theory fusion neglects the dynamic weighting of evidence conflict. This invention adds a conflict weighting factor to enhance decision-making stability in high-conflict scenarios, expressed as:

[0060] Based on confidence levels, a fusion decision is made, and the fusion formula is expressed as:

[0061] in, This represents the basic probability allocation function after fusion, i.e., the comprehensive confidence distribution of each fault identification framework. The conflict weighting mechanism reduces the false positive rate by 10% when there is inconsistency in multi-source data, and improves the accuracy of the overall fusion decision, especially in the complex environment of multimodal signal coupling in wind turbine pitch bearings.

[0062] like Greater than the fault threshold If the wind turbine pitch bearing is determined to be in a fault state, i.e., the first fault diagnosis result, the fault threshold in this embodiment is... The value was chosen as 0.95.

[0063] The above technical solution improves the accuracy of fault diagnosis by integrating multi-source data through EEMD decomposition and DS evidence theory.

[0064] S5: Input the time-domain features and frequency-domain features into the deep residual network diagnostic model for processing to obtain the second fault diagnosis result; compare the second fault diagnosis result with the first fault diagnosis result. If they are consistent, output the final diagnosis result; if they are inconsistent, perform a re-diagnosis.

[0065] In this embodiment, a variant of the deep residual network (ResNet) is used for fault diagnosis to improve the model's generalization ability to multi-parameter data. The deep residual network diagnostic model is used to diagnose faults in the pitch system.

[0066] The triaxial vibration signal data is fused with the angle data to form the input tensor. ,in, For batches, For the channel, It is a time series.

[0067] The deep residual network diagnostic model is based on the ResNet-18 architecture and incorporates a spatiotemporal attention module (STA). An attention layer is added after the residual block, as shown below:

[0068] in, , , These are query, key, and value matrices, extracted from fused features and optimized for pitch spatiotemporal synchronization data. The attention dimension is set to 64 in this embodiment. Using the Adam optimizer with cross-entropy as the loss function, and incorporating transfer learning, a deep residual network diagnostic model is pre-trained from a general vibration dataset and then fine-tuned based on wind turbine data to obtain a well-trained deep residual network diagnostic model.

[0069] The system receives time-domain and frequency-domain features, and outputs a fault probability using a trained deep residual network diagnostic model. This fault probability is compared to a set threshold; if it exceeds the threshold, the wind turbine pitch bearing is determined to be faulty, resulting in a second fault diagnosis. The second fault diagnosis is then compared to the first fault diagnosis. If they match, a final diagnosis is output; otherwise, data is re-acquired and a new diagnosis is performed.

[0070] The aforementioned technical features address the challenge of multi-source signal coupling, improve fault diagnosis robustness, and enhance fault identification accuracy. The extended high-frequency detection bandwidth covers 200-70 kHz, enabling the detection of microcracks.

[0071] Example 2 This embodiment discloses an intelligent monitoring system for wind turbine pitch bearings based on multi-parameter fusion, including: The multi-source data acquisition module is configured to acquire the dynamic parameters of the pitch bearing of the wind turbine to be monitored, including triaxial acceleration and triaxial angular velocity. An angle fusion calculation module is configured to: calculate the angle based on the triaxial acceleration and triaxial angular velocity respectively, and perform complementary filtering fusion on the obtained angles based on adaptive filtering coefficients to obtain the final angle; The feature extraction module is configured to: perform ensemble empirical mode decomposition on the triaxial vibration signal to generate multiple signal components, and extract low-frequency vibration features, high-frequency vibration features and angle changes based on the signal components respectively; The first fault diagnosis module is configured to: use DS evidence theory to fuse low-frequency vibration characteristics, high-frequency vibration characteristics and angle change to obtain the total confidence level, and combine it with the preset fault threshold to judge the fault and obtain the first fault diagnosis result. The second fault diagnosis module is configured to: input low-frequency vibration characteristics, high-frequency vibration characteristics, and angle change into the deep residual network diagnosis model for processing to obtain a second fault diagnosis result; compare the second fault diagnosis result with the first fault diagnosis result; if they are consistent, output the final diagnosis result; if they are inconsistent, perform a re-diagnosis.

[0072] The system adopts a multi-core heterogeneous computing architecture, using a single-chip microcontroller (MCU) and a field-programmable gate array (FPGA) for collaborative processing. The MCU is responsible for low-frequency signal processing and communication, while the FPGA implements real-time filtering and feature extraction. It adopts hierarchical storage, with NOR Flash (4GB) storing firmware and SDRAM (64MB) caching high-speed data.

[0073] The sensor device for data acquisition is designed with high reliability protection, featuring an IP67 waterproof and dustproof structure (metal housing and silicone potting), EMI shielding, circuit layer shielding and filtering design, and a signal-to-noise ratio (SNR) improvement of 15dB.

[0074] Example 3 The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method of Embodiment 1.

[0075] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method of Embodiment 1.

[0076] The steps and methods involved in the apparatuses of Embodiments 3 and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0077] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0079] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A wind turbine variable pitch bearing intelligent monitoring method based on multi-parameter fusion, characterized in that, The method comprises the following steps: acquiring dynamic parameters of a wind turbine variable pitch bearing to be monitored, the dynamic parameters comprising three-axis acceleration and three-axis angular velocity; calculating angles based on the three-axis acceleration and three-axis angular velocity respectively, and performing complementary filtering fusion on the angles obtained respectively based on adaptive filtering coefficients to obtain a final angle; performing ensemble empirical mode decomposition on three-axis vibration signals to generate a plurality of signal components, and extracting low-frequency vibration features, high-frequency vibration features and angle change amounts based on the signal components respectively; performing fusion decision on the low-frequency vibration features, high-frequency vibration features and angle change amounts by using D-S evidence theory to obtain a total confidence, and judging a fault in combination with a preset fault threshold to obtain a first fault diagnosis result; inputting the low-frequency vibration features, high-frequency vibration features and angle change amounts into a deep residual network diagnosis model for processing to obtain a second fault diagnosis result; comparing the second fault diagnosis result with the first fault diagnosis result, and if they are consistent, outputting a final diagnosis result, and if they are inconsistent, re-diagnosing.

2. The method of claim 1, wherein the method is based on multi-parameter fusion of the wind turbine bearing. The complementary filtering fusion of the three-axis acceleration and three-axis angular velocity is specifically as follows: respectively pre-processing collected three-axis acceleration and three-axis angular velocity signals to obtain pre-processed three-axis acceleration and three-axis angular velocity; calculating angles based on the pre-processed three-axis acceleration, and calculating angle increment integrals based on the pre-processed three-axis angular velocity; performing complementary filtering fusion based on the angles and the angle increment integrals to obtain fused dynamic parameters.

3. The method of claim 2, wherein the method is based on multi-parameter fusion of the wind turbine bearing. The preprocessing comprises gravity field normalization processing on the three-axis acceleration signals and zero offset compensation on the three-axis angular velocity signals.

4. The method of claim 2, wherein the method is based on multi-parameter fusion of wind turbine bearing intelligent monitoring. The calculation of the angles is expressed as: ， wherein represents a roll angle, represents a pitch angle.

5. The method of claim 2, wherein the method is based on multi-parameter fusion of wind turbine bearing intelligent monitoring. The calculation of the angle increment integrals is expressed as: ， wherein, denotes the roll angle at the last sampling time, denotes the pitch angle at the last sampling time, denotes the angular velocity in the axis direction, denotes the angular velocity in the axis direction, denotes the sampling time interval.

6. The method of claim 1, wherein the method is based on multi-parameter fusion of wind turbine bearing intelligent monitoring. The low-frequency vibration features and the high-frequency vibration features both comprise time domain features and frequency domain features.

7. The method of claim 1, wherein the method is based on multi-parameter fusion of wind turbine bearing intelligent monitoring. The fusion decision on the low-frequency vibration features, high-frequency vibration features and angle change amounts by using D-S evidence theory is specifically as follows: defining a plurality of evidence sources with input data; defining a recognition framework, constructing a basic probability assignment function for each evidence source, and assigning a basic probability; calculating a corresponding confidence degree based on the basic probability of each evidence source, and performing fusion decision based on the confidence degrees of the evidence sources to obtain a total confidence degree.

8. A fan variable pitch bearing intelligent monitoring system based on multi-parameter fusion, characterized in that, The method comprises the following steps: a multi-source data acquisition module configured to acquire dynamic parameters of a wind turbine variable pitch bearing to be monitored, the dynamic parameters comprising three-axis acceleration and three-axis angular velocity; an angle fusion calculation module configured to calculate angles based on the three-axis acceleration and three-axis angular velocity respectively, and perform complementary filtering fusion on the angles obtained respectively based on adaptive filtering coefficients to obtain a final angle; a feature extraction module configured to perform ensemble empirical mode decomposition on three-axis vibration signals to generate a plurality of signal components, and extract low-frequency vibration features, high-frequency vibration features and angle change amounts based on the signal components respectively; a first fault diagnosis module configured to perform fusion decision on the low-frequency vibration features, high-frequency vibration features and angle change amounts by using D-S evidence theory to obtain a total confidence, and judge a fault in combination with a preset fault threshold to obtain a first fault diagnosis result; and a second fault diagnosis module configured to input the low-frequency vibration features, high-frequency vibration features and angle change amounts into a deep residual network diagnosis model for processing to obtain a second fault diagnosis result. The second fault diagnosis module is configured to: input the low-frequency vibration feature, the high-frequency vibration feature, and the angle change amount to a deep residual network diagnosis model for processing to obtain a second fault diagnosis result; compare the second fault diagnosis result with the first fault diagnosis result, and if the two results are consistent, output a final diagnosis result, and if the two results are inconsistent, re-diagnose.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method for intelligent monitoring of a variable-pitch bearing of a wind turbine based on multi-parameter fusion according to any one of claims 1-7.

10. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the method for intelligent monitoring of a variable-pitch bearing of a wind turbine based on multi-parameter fusion according to any one of claims 1-7.

Images