CN-122014532-A - Fan bearing detection method and device for wind power generation equipment

CN122014532ACN 122014532 ACN122014532 ACN 122014532ACN-122014532-A

Abstract

The embodiment of the application reversely inverts physical parameters from a small amount of real observation signals and generates a high-fidelity synthetic fault data set constrained by a dynamic equation based on the real parameters, thereby solving the problems of model overfitting and generalization difference caused by scarcity of real fault data under the condition of a small sample, and simultaneously overcoming the physical distortion of a pure data driving generation method and the excessive dependence of traditional simulation on prior parameters, and realizing high-precision and strong-robustness fan bearing fault grading detection.

Inventors

WU SHIDONG
LIN ZHIHUA
HAN JUNFEI
ZHU XIAOYI
PAN HAINING
GUO SHUAINAN
WANG QIAN
SU YIBO
LI YUXIN
ZHANG XIAOLONG

Assignees

中国长江三峡集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (10)

1. The method for detecting the fan bearing of the wind power generation equipment is characterized by comprising the following steps of: Collecting an observation signal sequence of a fan bearing, and executing standardization processing and time sequence slicing actions on the observation signal sequence to generate a standardization sample sequence set; Determining actual engineering working condition parameters of the fan bearing, wherein the actual engineering working condition parameters comprise operation condition information representing the operation condition of the fan bearing in a real operation environment, geometric parameters representing the geometric characteristics of the fan bearing and material mechanical properties of the fan bearing; Determining a kinetic equation constraint condition based on the normalized sample sequence set, the operating condition information, the geometric parameters, and the material mechanical properties; Inverting an inversion physical parameter vector which is hidden in the observation signal sequence and is constrained by a physical feasible region from the observation signal sequence by adopting an inverse physical information neural network through the standardized sample sequence set and the constraint condition of the dynamic equation; Generating a multi-working condition fault parameter set constrained by the dynamic equation constraint condition based on the inversion physical parameter vector; constructing a fan bearing fault data set based on the extended multi-condition fault parameter set, and generating a fan bearing fault grading target model by utilizing the fan bearing fault data set; And controlling the fan bearing fault grading target model, and generating a detection result aiming at the target fan bearing based on target data aiming at the target fan bearing.
2. The method of claim 1, wherein the step of determining a kinetic equation constraint based on the normalized sample sequence set, the operating condition information, the geometric parameters, and the material mechanical properties comprises: constructing a degree-of-freedom coupling nonlinear dynamics equation based on the standardized sample sequence set, the running condition information, the geometric parameters and the material mechanical properties; calculating time-varying meshing stiffness and bearing supporting stiffness of the fan bearing by utilizing the degree-of-freedom coupling nonlinear dynamics equation, and calculating bending, shearing and axial compression deformation energy by potential energy method integration to generate a nonlinear time-varying stiffness matrix; Determining gear eccentric errors, multi-order machining error harmonic components and bearing geometric dimension deviations, and adjusting the nonlinear time-varying stiffness matrix through the gear eccentric errors, the multi-order machining error harmonic components and the bearing geometric dimension deviations to generate an enhanced dynamic system motion equation; carrying out time integral numerical solution on the motion equation of the enhanced dynamics system by adopting a backward differential formula method to generate time domain acceleration response curves of the input shaft, the intermediate shaft and the output shaft in multiple directions; Performing fast Fourier transform on the time domain acceleration response curve to generate a single-side amplitude spectrum; performing frequency domain gain weighting processing on the single-side amplitude spectrum to generate an enhanced spectrum signal for highlighting fault impact characteristics; An input feature set is constructed based on the enhanced-spectrum signal and the observed signal sequence.
3. The method of claim 2, wherein inverting the implicit physical feasibility domain constrained inversion physical parameter vector in the sequence of observed signals from the sequence of observed signals using a reverse physical information neural network through the set of normalized sample sequences and the kinetic equation constraint comprises: generating a time sequence feature representation which fuses short-time local impact details and long-time global period dynamics by inputting the input feature set into a shared asymmetric convolution residual block comprising a response head branch and a physical head branch in parallel; applying a self-adaptive normalization layer and an activation function to the time sequence feature representation, reserving a leachable scale and a translation parameter through batch normalization, and weighting statistics in a time window to generate a normalized deep feature vector; And generating an inversion physical parameter vector constrained by a physical feasible region by utilizing the normalized deep feature vector through fully connecting layer regression physical parameters.
4. A method according to claim 3, wherein the step of generating a multi-operating fault parameter set constrained by the kinetic equation constraint based on the inverted physical parameter vector comprises: encoding the inversion physical parameter vector into a feature space, and constructing a reverse parameter encoding mechanism to take damping coefficients, linear stiffness, nonlinear stiffness parameters, impact amplitude and fault characteristic frequency as trainable vectors to generate an implicit physical quantity encoding capable of expressing complex fault behaviors; Calculating data consistency loss based on the standardized sample sequence set and the input feature set, predicting the difference between response and a real measurement result through the data consistency loss quantization model, and generating primarily optimized dynamic system parameters; introducing regularization items to apply soft constraint to the preliminarily optimized dynamic system parameters, and generating a reverse dynamic parameter set with reasonable physical state by utilizing priori structural information to shrink a solution space; And determining the inverse dynamics parameter set as a basic parameter, determining a plurality of groups of typical rotating speed disturbance values, a plurality of loads random disturbance coefficients, a plurality of fault level adjustment attenuation coefficients and impact period coefficients, and adding the plurality of groups of typical rotating speed disturbance values, the plurality of loads random disturbance coefficients, the plurality of fault level adjustment attenuation coefficients and the impact period coefficients into the basic parameter to generate an expanded multi-working-condition fault parameter set.
5. The method of claim 2, wherein the step of performing frequency domain gain weighting on the single-sided amplitude spectrum to generate an enhanced spectrum signal for highlighting fault impact features comprises: Identifying a fundamental frequency position of the meshing frequency in the single-side amplitude spectrum and each order frequency multiplication position of the fundamental frequency position of the meshing frequency, and generating a key fault concern frequency point set based on the fundamental frequency position of the meshing frequency and each order frequency multiplication position; constructing a Gaussian or rectangular window type weighting factor function based on the key fault focus frequency point set, and calculating a gain weighting vector with the same length as the single-side amplitude spectrum; Multiplying the gain weight vector with the single-side amplitude spectrum element by element to generate a weight spectrum with amplitude distribution for highlighting fault impact characteristics; An inverse fast fourier transform is performed on the weighted spectrum, generating an enhanced spectrum signal for highlighting the fault impact feature.
6. The method of claim 4, wherein the step of constructing a fan bearing failure data set based on the extended multiple-event failure parameter set comprises: Driving a generation model to execute forward time integral simulation based on the extended multi-condition fault parameter set to obtain a time domain vibration response sequence conforming to the constraint condition of the nonlinear dynamic equation; performing standardization processing and sliding window slicing on the time domain vibration response sequence to obtain a standardized fault sample sequence set under multiple working conditions; performing frequency domain transformation and gain weighting processing on the standardized fault sample sequence set to obtain an enhanced fault spectrum signal set with outstanding early damage impact and modulation characteristics; Carrying out channel dimension superposition or characteristic splicing on the enhanced fault spectrum signal set and the standardized fault sample sequence set to obtain a multi-mode fault data sample set fused with time domain and frequency domain information; And adding controllable noise disturbance and working condition label labels to the multi-mode fault data sample set to obtain a fan bearing fault data set.
7. The method of claim 6, wherein the step of controlling the fan bearing failure classification target model to generate a detection result for a target fan bearing based on target data for the target fan bearing comprises: Acquiring a real-time vibration acceleration signal sequence of a target fan bearing, and determining the real-time vibration acceleration signal sequence as a target observation signal sequence of the target fan bearing, wherein the target observation signal sequence is original time domain vibration acceleration data continuously acquired at a preset position of target wind power generation equipment according to a preset sampling frequency; Executing standardization processing and time sequence slicing actions on the target observation signal sequence of the target fan bearing to generate a standardized sample sequence set aiming at the fan bearing fault grading target model; inputting the standardized target sample sequence set into the fan bearing fault grading target model to generate a prediction probability distribution aiming at fault types and fault severity; And determining a detection result aiming at the fault type and severity degree of the target fan bearing according to the prediction probability distribution.
8. The utility model provides a fan bearing detection device of wind power generation equipment which characterized in that includes: the standardized sample sequence set generating module is used for acquiring an observation signal sequence of the fan bearing, and executing standardized processing and time sequence slicing actions on the observation signal sequence to generate a standardized sample sequence set; The actual engineering working condition parameter determining module is used for determining the actual engineering working condition parameters of the fan bearing, wherein the actual engineering working condition parameters comprise operation condition information representing the operation condition of the fan bearing in a real operation environment, geometric parameters representing the geometric characteristics of the fan bearing and material mechanical properties of the fan bearing; A kinetic equation constraint condition determining module configured to determine a kinetic equation constraint condition based on the normalized sample sequence set, the operating condition information, the geometric parameter, and the material mechanical property; The inversion physical parameter vector inversion module is used for inverting the inversion physical parameter vector which is hidden in the observation signal sequence and is constrained by a physical feasible region from the observation signal sequence by adopting a reverse physical information neural network through the standardized sample sequence set and the constraint condition of the dynamic equation; The multi-working-condition fault parameter set generation module is used for generating a multi-working-condition fault parameter set constrained by the constraint condition of the dynamic equation based on the inversion physical parameter vector; the fan bearing fault grading target model generation module is used for constructing a fan bearing fault data set based on the extended multi-station fault parameter set and generating a fan bearing fault grading target model by utilizing the fan bearing fault data set; The detection result generation module is used for controlling the fan bearing fault grading target model and generating a detection result aiming at the target fan bearing based on target data aiming at the target fan bearing.
9. An electronic device comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, which program or instruction when executed by the processor implements the method of claims 1-7.
10. A readable storage medium, characterized in that it stores thereon a program or instructions, which when executed by a processor, implements the method according to claims 1-7.

Description

Fan bearing detection method and device for wind power generation equipment Technical Field The invention relates to the technical field of wind turbine bearing detection of wind power generation equipment, in particular to a wind turbine bearing detection method of wind power generation equipment, a wind turbine bearing detection device of wind power generation equipment, electronic equipment and a readable storage medium. Background Along with the continuous and rapid increase of the installed capacity of wind power generation, the fan bearing is used as a core component of a transmission chain, is subjected to multiple adverse factors such as complex variable load, variable rotation speed, temperature fluctuation, lubrication degradation and the like for a long time, and becomes a weak link of the complete machine which is most prone to faults such as fatigue peeling, abrasion, cracks and the like. Once the bearing is damaged or fails early, not only can the unplanned shutdown of the unit be caused and the power generation efficiency be reduced sharply, but also major safety accidents such as main shaft breakage, gear box damage and even complete machine overturning can be caused, and the operation and maintenance cost and the power grid stability are seriously threatened. In recent years, a fan bearing fault diagnosis method based on deep learning has been significantly developed, and manual intervention can be reduced to a certain extent through end-to-end feature learning. However, this type of method is highly dependent on a large number of high-quality labeling fault samples, while actual wind farm operation data is mainly in a healthy state, actual fault samples (especially middle and late serious faults) are extremely rare, and even enough samples are difficult to collect for training. The model is easy to fit under the condition of a small sample, has poor generalization capability and high false alarm/missing report rate, and cannot meet the requirement of the engineering field on early warning of high-precision and strong-robustness faults. Disclosure of Invention Embodiments of the present invention provide a method, an apparatus, an electronic device, and a readable storage medium for detecting a fan bearing of a wind power generation device, so as to overcome or at least partially solve the above-mentioned problems. In order to solve the technical problems, the application is realized as follows: In a first aspect, an embodiment of the present application provides a method for detecting a fan bearing of a wind power generation device, including: Collecting an observation signal sequence of a fan bearing, and executing standardization processing and time sequence slicing actions on the observation signal sequence to generate a standardization sample sequence set; Determining actual engineering working condition parameters of the fan bearing, wherein the actual engineering working condition parameters comprise operation condition information representing the operation condition of the fan bearing in a real operation environment, geometric parameters representing the geometric characteristics of the fan bearing and material mechanical properties of the fan bearing; Determining a kinetic equation constraint condition based on the normalized sample sequence set, the operating condition information, the geometric parameters, and the material mechanical properties; Inverting an inversion physical parameter vector which is hidden in the observation signal sequence and is constrained by a physical feasible region from the observation signal sequence by adopting an inverse physical information neural network through the standardized sample sequence set and the constraint condition of the dynamic equation; Generating a multi-working condition fault parameter set constrained by the dynamic equation constraint condition based on the inversion physical parameter vector; constructing a fan bearing fault data set based on the extended multi-condition fault parameter set, and generating a fan bearing fault grading target model by utilizing the fan bearing fault data set; And controlling the fan bearing fault grading target model, and generating a detection result aiming at the target fan bearing based on target data aiming at the target fan bearing. Optionally, the step of determining a kinetic equation constraint condition based on the normalized sample sequence set, the operating condition information, the geometric parameter, and the material mechanical property comprises: constructing a degree-of-freedom coupling nonlinear dynamics equation based on the standardized sample sequence set, the running condition information, the geometric parameters and the material mechanical properties; calculating time-varying meshing stiffness and bearing supporting stiffness of the fan bearing by utilizing the degree-of-freedom coupling nonlinear dynamics equation, and calculating bending, shearing and axial compression