Intelligent fault diagnosis method, system and equipment for multiple parts of wind power generator cabin and medium

Abstract

The invention discloses a method, a system, equipment, a medium and a program for intelligently diagnosing faults of multiple parts of a wind power generator cabin, and belongs to the technical field of wind power generation. The method comprises the following steps: collecting a multi-source acoustic signal in a cabin of the wind power generator, preprocessing the multi-source acoustic signal to obtain a preprocessed voiceprint signal, and generating a spectrogram; performing cross-domain feature extraction on the pre-processed voiceprint signal, performing feature dimension reduction and redundancy elimination through principal component analysis, and constructing a voiceprint feature vector; inputting the spectrogram and the voiceprint feature vector into a CNN-LSTM fusion diagnosis model, and performing fault diagnosis to obtain a fault diagnosis result; and matching the fault diagnosis result with the fault-voiceprint knowledge graph to generate a decision suggestion. According to the method, through non-contact voiceprint monitoring and multi-dimensional analysis, high-precision and real-time diagnosis and active decision-making of concurrent faults of multiple components of the wind power generator cabin are realized, and the defects of complex deployment and sound signal interference resistance of traditional vibration monitoring are effectively overcome.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, specifically to a method, system, equipment, medium, and program for intelligent diagnosis of multi-component faults in a wind turbine nacelle. Background Technology

[0002] As a crucial component of clean energy, wind power generation has seen its installed capacity and power generation scale continuously increase. Wind turbine generators, as the core equipment of wind power systems, directly impact the power generation efficiency and economic benefits of the entire wind farm through their stable and safe operation. The wind turbine nacelle, a key component of the wind turbine, integrates numerous critical components such as gearboxes, bearings, and generators. These components are highly susceptible to various failures during long-term operation due to complex mechanical loads, environmental factors, and their own aging. Failure of critical components not only leads to generator shutdown and power loss but can also trigger serious safety accidents, threatening the lives of personnel and the safety of equipment and property. Therefore, timely and accurate condition monitoring of critical components in the wind turbine nacelle, enabling early detection of potential faults and the implementation of effective measures, is of paramount importance for ensuring the safe operation of wind turbine generators. Currently, mainstream technologies for condition monitoring of critical components in wind turbine nacelles include traditional vibration monitoring methods, non-contact monitoring methods based on sound signals, and multi-sensor fusion schemes; however, all of these technologies have significant limitations.

[0003] Traditional vibration monitoring methods primarily rely on contact sensors. During deployment, these sensors need to be precisely installed at specific locations on the monitored components. This installation process is complex and time-consuming, increasing equipment costs and potentially affecting sensor accuracy due to improper installation. Furthermore, contact sensors are constantly exposed to harsh mechanical environments, making them susceptible to damage from impacts and wear, leading to inaccurate or even invalid monitoring data. More critically, this method struggles to comprehensively capture concurrent failures of multiple components within the engine room. Due to the compact space and interconnected nature of components within the engine room, a failure in one component can trigger a chain reaction affecting others. Traditional vibration monitoring methods often only monitor individual components, failing to effectively identify patterns and characteristics of concurrent failures across multiple components.

[0004] While non-contact monitoring methods based on sound signals offer advantages such as ease of deployment and no interference with equipment operation, existing methods typically rely on simple spectral energy analysis. This approach is extremely sensitive to environmental noise interference. In actual wind farm environments, there are multiple interference sources, including wind noise, motor noise, and noise generated by other equipment operation. These noises can mask the weak sound signals generated by faults, making it difficult to effectively separate and identify fault characteristics. Especially for early-stage faults, the resulting sound signals are often very weak and may overlap with normal sound signals. Simple spectral energy analysis cannot accurately extract these early fault characteristics, thus hindering early warning. Furthermore, existing methods generally lack the ability to dynamically analyze the evolution of faults over time, failing to track fault development trends in real time and promptly detect fault deterioration.

[0005] While multi-sensor fusion schemes can theoretically integrate information from multiple sensors to improve the accuracy of fault diagnosis, the heterogeneity of data collected by different sensors (differences in data format, features, and semantics) and the high complexity of fusion algorithms lead to excessively long data processing and analysis times, making it difficult to meet the needs of real-time diagnosis and provide accurate diagnostic results immediately after a fault occurs. Furthermore, methods using single sound signals and static spectral features combined with traditional classifiers for specific component fault identification also have many problems. This method is susceptible to environmental noise interference leading to misjudgments, cannot distinguish overlapping sound sources from different components within the same frequency band, and has limited diagnostic capabilities for non-stationary or progressive faults such as bearing wear, typically providing only a simple normal / abnormal binary judgment, failing to meet the requirements for accurate identification and severe quantification of faults in critical engine room components. Summary of the Invention

[0006] To address the problems of traditional vibration monitoring systems, such as complex deployment and difficulty in capturing concurrent faults in multiple components, poor noise immunity and lack of dynamic analysis in acoustic signal monitoring, difficulty in real-time diagnosis through multi-sensor fusion, and high misjudgment and limited diagnostic capabilities in single-signal recognition, this invention provides an intelligent diagnostic method for multiple component faults in wind turbine nacelles. Through non-contact acoustic signature monitoring and multi-dimensional analysis, it achieves high-precision, real-time diagnosis and proactive decision-making for concurrent faults in multiple components of wind turbine nacelles, effectively overcoming the challenges of complex deployment and interference resistance in traditional vibration monitoring systems.

[0007] To achieve the above objectives, the present invention provides the following technical solution.

[0008] In a first aspect, the present invention provides an intelligent fault diagnosis method for multiple components of a wind turbine nacelle, comprising: Multi-source acoustic signals inside the wind turbine nacelle are collected and preprocessed to obtain preprocessed acoustic signature signals, which are then used to generate a spectrogram. Cross-domain feature extraction is performed on the preprocessed voiceprint signal, and feature dimensionality reduction and redundancy elimination are performed through principal component analysis to construct the voiceprint feature vector; The spectrogram and voiceprint feature vector are input into the CNN-LSTM fusion diagnostic model to perform fault diagnosis and obtain the fault diagnosis result. The fault diagnosis results are matched with the fault-voiceprint knowledge graph to generate decision suggestions.

[0009] As a further improvement of the present invention, the step of collecting multi-source acoustic signals inside the wind turbine nacelle, preprocessing them to obtain preprocessed acoustic signature signals, and generating a spectrogram includes: A microphone array was deployed in the key component area to be collected inside the cabin to collect multi-source acoustic signals. By using a deep learning-based recurrent neural network model, the collected multi-source acoustic signals are separated and processed in real time to obtain the device's operating acoustic fingerprint and form a pre-processed acoustic fingerprint signal. The pure audioprint signal in the equipment operation soundprint is subjected to time-frequency domain normalization processing, and the feature separability of the human ear sensitive frequency band is enhanced by using Mel-scale filter bank to generate a normalized spectrogram.

[0010] As a further improvement of the present invention, the step of performing cross-domain feature extraction on the preprocessed voiceprint signal, and constructing a voiceprint feature vector by performing feature dimensionality reduction and redundancy elimination through principal component analysis includes: Time-domain features, frequency-domain features, and time-frequency-domain features are extracted from the preprocessed speakerprint signal; Principal component analysis is used to reduce feature dimensionality and eliminate redundancy, constructing voiceprint feature vectors for different components; and the voiceprint feature vectors of different components are stored in a multi-dimensional voiceprint template library.

[0011] As a further improvement of the present invention, the step of inputting the spectrogram and voiceprint feature vector into the CNN-LSTM fusion diagnostic model for fault diagnosis and obtaining fault diagnosis results includes: The spectrogram is input into the convolutional neural network in the CNN-LSTM fusion diagnostic model to identify local abnormal patterns in the spectrum. The temporal voiceprint feature vector is input into the long short-term memory network in the CNN-LSTM fusion diagnostic model to capture the changing pattern of fault features over time. Based on the changes in local anomaly patterns and fault characteristics in the spectrum over time, fault diagnosis is performed to obtain fault diagnosis results.

[0012] As a further improvement of the present invention, the fault diagnosis results include the faulty component, fault type, and severity level.

[0013] As a further improvement of the present invention, the step of matching the fault diagnosis results with the fault-voiceprint knowledge graph to generate decision suggestions includes: When an anomaly is detected in the fault diagnosis results, the fault diagnosis results are matched with the fault-voiceprint knowledge graph, and actionable decision suggestions are output.

[0014] Secondly, the present invention provides an intelligent fault diagnosis system for multiple components of a wind turbine nacelle, comprising: Preprocessing module: used to collect multi-source acoustic signals inside the wind turbine nacelle, perform preprocessing, obtain preprocessed acoustic signature signals, and generate a spectrogram; Feature vector module: used to extract cross-domain features from preprocessed voiceprint signals, perform feature dimensionality reduction and redundancy elimination through principal component analysis, and construct voiceprint feature vectors; Diagnostic Results Module: Used to input spectrograms and voiceprint feature vectors into the CNN-LSTM fusion diagnostic model to perform fault diagnosis and obtain fault diagnosis results; Decision suggestion module: This module is used to match fault diagnosis results with the fault-voiceprint knowledge graph to generate decision suggestions.

[0015] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the intelligent diagnosis method for multiple component faults of a wind turbine nacelle.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the intelligent diagnosis method for multiple component faults in a wind turbine nacelle.

[0017] Fifthly, the present invention provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the intelligent diagnosis method for multiple component faults in a wind turbine nacelle.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention collects multi-source acoustic signals from inside a wind turbine nacelle and generates a spectrogram. Combined with cross-domain feature extraction of the pre-processed acoustic signature signal, it constructs a precise acoustic signature feature vector. This multi-dimensional and comprehensive signal analysis and feature construction method can deeply capture subtle changes in the sound characteristics emitted by different components under various operating conditions. Even in complex situations where multiple components fail simultaneously, it can accurately identify and distinguish each fault source, greatly improving the accuracy of fault diagnosis and avoiding misjudgments and omissions caused by single signal identification or limitations of traditional monitoring methods. This lays a solid foundation for timely and accurate fault handling. Secondly, the non-contact acoustic signature monitoring method used in this invention eliminates the need to install numerous complex sensors on various components, simplifying the deployment of the monitoring system and reducing data transmission and processing steps. Simultaneously, the acoustic spectrogram and acoustic signature feature vector are input into a CNN-LSTM fusion diagnostic model. This model combines the powerful spatial feature extraction capability of Convolutional Neural Networks (CNN) with the effective processing capability of Long Short-Term Memory Networks (LSTM) for time-series data. It can quickly process and analyze the input data, providing fault diagnosis results in a short time. This enables real-time monitoring and fault diagnosis of the wind turbine nacelle's operating status, ensuring that faults are detected and addressed immediately upon occurrence, effectively reducing the risk of equipment damage and production losses due to delayed fault handling. Furthermore, through professional preprocessing of acoustic signals and subsequent scientific feature extraction and analysis processes, this invention can minimize the impact of environmental noise and other interference factors on the signals, extracting acoustic signature features that truly reflect the component's fault state. This makes the diagnostic results more reliable and unaffected by complex external environments, enabling stable and accurate fault diagnosis in noisy wind farms or other harsh operating conditions. In addition, this invention matches the fault diagnosis results with a fault-acoustic signature knowledge graph and generates decision suggestions, achieving a leap from fault diagnosis to proactive decision-making. The fault-voiceprint knowledge graph integrates a wealth of fault cases and corresponding voiceprint feature information. By matching them with diagnostic results, it can provide maintenance personnel with detailed and accurate fault cause analysis and targeted maintenance suggestions, guiding the efficient conduct of maintenance work, improving maintenance efficiency and quality, reducing maintenance costs, and realizing proactive management and intelligent maintenance of wind turbine nacelle faults. Attached Figure Description

[0019] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. In the drawings: Figure 1 This is a flowchart illustrating an intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to the present invention. Figure 2 This is a schematic diagram illustrating the specific process of the intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to the present invention. Figure 3 This is a diagram illustrating the working mechanism of a dual-path dynamic diagnostic model for a multi-component intelligent fault diagnosis method for wind turbine nacelles according to the present invention. Figure 4 This is a schematic diagram of the structure of a multi-component intelligent fault diagnosis system for a wind turbine nacelle according to the present invention; Figure 5 This is a schematic diagram of an electronic device in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0021] Unless otherwise defined, 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. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] To address the problems of existing technologies, such as the complexity of traditional vibration monitoring deployment and its difficulty in capturing concurrent faults in multiple components, the poor noise immunity and lack of dynamic analysis in sound signal monitoring, the difficulty in real-time diagnosis through multi-sensor fusion, and the high rate of misjudgments and limited diagnostic capabilities in single-signal recognition, this invention provides an intelligent fault diagnosis method for multiple components of a wind turbine nacelle. Figure 1 As shown, it includes: S100: Collects multi-source acoustic signals inside the wind turbine nacelle, performs preprocessing, obtains preprocessed acoustic signature signals, and generates a spectrogram; S200: Perform cross-domain feature extraction on the preprocessed voiceprint signal, and use principal component analysis to reduce feature dimensionality and eliminate redundancy to construct a voiceprint feature vector; S300: Input the acoustic spectrogram and voiceprint feature vector into the CNN-LSTM fusion diagnostic model to perform fault diagnosis and obtain the fault diagnosis result; S400: Matches fault diagnosis results with the fault-voiceprint knowledge graph to generate decision recommendations.

[0023] This method enables high-precision, real-time diagnosis and proactive decision-making for concurrent faults in multiple components of a wind turbine nacelle through non-contact acoustic fingerprint monitoring and multi-dimensional analysis, effectively overcoming the challenges of complex deployment and interference resistance of traditional vibration monitoring.

[0024] The present invention will be further explained and described below with reference to the accompanying drawings.

[0025] like Figure 2 As shown, this invention provides a method and system for intelligent fault diagnosis of multiple components in a wind turbine nacelle based on acoustic signature analysis, specifically including the following steps: S1: Voiceprint Signal Acquisition and Enhancement Highly directional microphone arrays are deployed in key component areas such as gearboxes, generators, and main bearings within the cabin to simultaneously collect multi-source acoustic signals. An adaptive noise separation module based on deep learning recurrent neural networks (RNNs) is used to separate environmental background noise from equipment operating sound patterns in real time. The pure sound pattern signal is then subjected to time-frequency domain standardization processing, and a Mel-scale filter bank is used to enhance the feature separability of the frequency bands sensitive to human ears, generating a standardized spectrogram.

[0026] S2: Multi-dimensional acoustic feature fusion extraction Cross-domain joint feature extraction is performed on the preprocessed voiceprint signal, covering the three core dimensions of time domain, frequency domain, and time-frequency domain: In time-domain analysis, the envelope harmonic distortion rate of the signal is calculated to quantify the degree of nonlinear distortion. This is combined with short-time zero-crossing rate to detect transient impact events, and a pulse factor is introduced to characterize the intensity of abnormal impacts. In frequency domain analysis, Mel frequency cepstral coefficients (MFCC) are extracted to simulate the characteristics of human hearing. Non-Gaussian fault components are identified by spectral kurtosis, and subband energy entropy is calculated to reflect the degree of disorder in the frequency band energy distribution. In the joint time-frequency analysis, wavelet packet transform is used to decompose the signal into fine frequency bands, extract the transient impact energy distribution of each frequency band, and calculate the wavelet packet energy entropy to quantify the signal complexity.

[0027] The above-mentioned multi-dimensional features together constitute a complete characterization system for device voiceprints.

[0028] Based on the original features extracted across domains, further feature fusion and dimensionality reduction optimization are performed: Principal Component Analysis (PCA) is used to eliminate redundant information, retaining a subset of features with high discriminative power, and constructing dedicated voiceprint feature vectors for different components such as gearboxes, bearings, and generators. Each feature vector integrates time-domain dynamic response, frequency-domain structural characteristics, and time-frequency local details. Finally, the feature vectors are input into a multi-dimensional voiceprint template library, providing a robust and high-resolution input data foundation for subsequent intelligent diagnosis.

[0029] S3: Dual-path dynamic fault diagnosis like Figure 3 As shown, a CNN-LSTM fusion diagnostic model is constructed to achieve dual-path analysis: Static feature path: Input the spectrogram into a convolutional neural network (CNN) to identify local abnormal patterns in the spectrum, such as periodic pulse clusters caused by bearing peeling and sideband modulation caused by broken gear teeth; Dynamic evolution path: Input the temporal voiceprint feature vector into a long short-term memory network (LSTM) to capture the changing pattern of fault features over time, such as the migration of harmonic energy to higher frequencies caused by increased wear; The dual-path outputs are fused through a fully connected layer to generate a three-dimensional fault probability matrix. The dimensions include: faulty component (gearbox / bearing / generator), fault type (stripping / wear / eccentricity, etc.), and severity level (mild / moderate / severe).

[0030] S4: Knowledge Graph-Assisted Decision Making Establish a fault-soundprint knowledge graph, with nodes including: historical fault cases, soundprint feature vectors, maintenance measures, and component life curves; when an anomaly is detected, the system will match the real-time diagnostic results with the graph and output executable decision suggestions: generate a maintenance priority list; combine the soundprint evolution trend with the life decay model in the graph; automatically trigger SCADA system alarms and push diagnostic reports to the operation and maintenance terminal.

[0031] In summary, the voiceprint fingerprint separation and enhancement technology employed in this invention, leveraging the spatial filtering function of a highly directional microphone array, acts like a precise "locator" for sound signal capture, effectively focusing the voiceprint signal emitted by the device. Simultaneously, the noise separation model based on deep learning RNNs acts as an efficient "noise cleaner," accurately separating environmental background noise from the device's voiceprint, successfully solving the long-standing industry problem of aliasing of sound sources within the same frequency band. This technology significantly improves the signal-to-noise ratio for detecting weak fault features, enabling the clear capture of early, minute faults hidden within complex noise, buying valuable time for early fault detection and timely handling, and greatly reducing the risk of further fault deterioration leading to serious equipment damage. Secondly, this invention overcomes the limitations of traditional single-dimensional feature analysis. Through cross-domain joint feature extraction, it encompasses multiple dimensions of feature information, including time-domain envelope harmonic distortion rate, frequency-domain MFCC and spectral kurtosis, and time-frequency domain wavelet packet energy entropy, constructing dedicated voiceprint feature vectors for different components such as gearboxes, bearings, and generators. This comprehensive, multi-faceted feature extraction method can more fully and accurately reflect the operating status and fault characteristics of components. Even in complex situations where multiple components simultaneously experience compound faults, it can accurately identify and distinguish them, effectively avoiding misjudgments and omissions caused by incomplete feature information, and providing a solid and reliable basis for fault diagnosis. Third, this invention uses a CNN path to quickly identify local spatial anomaly patterns from the acoustic spectrogram; while the LSTM path can deeply analyze the temporal evolution of the acoustic signature feature vector. The dual-path output fusion generates a three-dimensional fault probability matrix (component × fault type × severity level), presenting detailed fault information in an intuitive and comprehensive manner. It not only clarifies the component where the fault occurred but also accurately indicates the fault type and severity, providing maintenance personnel with clear and accurate maintenance guidance, greatly improving maintenance efficiency and targeting. Fourth, this invention, through a graph-based knowledge base linking historical cases, maintenance measures, and lifespan curves, accurately maps real-time diagnostic results to graph nodes, enabling rapid output of quantifiable maintenance strategies. This allows maintenance decisions to move beyond reliance on experience-based judgment and instead rely on extensive historical data and scientific analysis, ensuring the rationality and effectiveness of maintenance measures. It also enables the prediction of equipment lifespan, providing strong support for the full lifecycle management of equipment. Furthermore, the system-level protection points of this invention form a complete and efficient diagnostic system architecture, from the signal acquisition layer to the cloud-based diagnostic platform and then to the automatic maintenance strategy output engine. This achieves full automation and intelligence in fault diagnosis, further improving the efficiency and reliability of fault diagnosis and providing strong technical support for the stable operation and sustainable development of the wind power industry.

[0032] The second objective of this invention is to propose an intelligent fault diagnosis system for multiple components of a wind turbine nacelle, such as... Figure 4 As shown, it includes: Preprocessing module 100: used to collect multi-source acoustic signals inside the wind turbine nacelle, perform preprocessing, obtain preprocessed acoustic signature signals, and generate a spectrogram; Feature vector module 200: used to perform cross-domain feature extraction on the preprocessed voiceprint signal, and to construct voiceprint feature vectors by performing feature dimensionality reduction and redundancy elimination through principal component analysis; Diagnosis Result Module 300: Used to input the spectrogram and voiceprint feature vector into the CNN-LSTM fusion diagnosis model to perform fault diagnosis and obtain fault diagnosis results; Decision suggestion module 400: Used to match fault diagnosis results with fault-voiceprint knowledge graph to generate decision suggestions.

[0033] like Figure 5 As shown, a third objective of this invention is to provide an electronic device comprising a processor 501, a memory 502, and a display screen 503. The memory 502 and the display screen 503 are both connected to the processor 501, such as via a bus 504. Optionally, the electronic device may further include a transceiver 505. It should be noted that in practical applications, the transceiver 505 is not limited to one type, and the structure of this electronic device does not constitute a limitation on the embodiments of this application.

[0034] Processor 501 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 501 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0035] Bus 504 may include a pathway for transmitting information between the aforementioned components. Bus 504 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 504 can be divided into address bus, data bus, control bus, etc.

[0036] The memory 502 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0037] The memory 502 is used to store application code that executes the solution of this application, and its execution is controlled by the processor 501. The processor 501 is used to execute the application code stored in the memory 502 to implement the content shown in the foregoing method embodiments.

[0038] Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0039] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the aforementioned functions. Figure 1 The illustrated method embodiments include various processes. For example, a memory may include instructions that can be executed by a processor of an electronic device to perform the described method.

[0040] A computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), staging random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory stick, floppy disk, optical disk, magnetic disk, mechanical encoding device, or any combination thereof.

[0041] A fifth objective of this invention is to provide a computer program product comprising computer instructions that, when executed by a processor, implement the above-described... Figure 1 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0042] Many embodiments and applications beyond the examples provided will be apparent to those skilled in the art upon reading the foregoing description. Therefore, the scope of this teaching should not be determined by reference to the foregoing description, but rather by reference to the foregoing claims and the full scope of their equivalents. For purposes of completeness, all articles and references, including patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the foregoing claims is not intended as a waiver of that subject matter, nor should it be construed as an indication that the applicant has not considered that subject matter as part of the disclosed inventive subject matter.

[0043] The above content provides a further detailed description of the present invention. It should not be construed that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for intelligent diagnosis of faults in multiple components of a wind turbine nacelle, characterized in that, include: Multi-source acoustic signals inside the wind turbine nacelle are collected and preprocessed to obtain preprocessed acoustic signature signals, which are then used to generate a spectrogram. Cross-domain feature extraction is performed on the preprocessed voiceprint signal, and feature dimensionality reduction and redundancy elimination are performed through principal component analysis to construct the voiceprint feature vector; The spectrogram and voiceprint feature vector are input into the CNN-LSTM fusion diagnostic model to perform fault diagnosis and obtain the fault diagnosis result. The fault diagnosis results are matched with the fault-voiceprint knowledge graph to generate decision suggestions.

2. The intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to claim 1, characterized in that, The process involves collecting multi-source acoustic signals from inside the wind turbine nacelle, preprocessing them to obtain preprocessed acoustic signature signals, and generating a spectrogram, including: A microphone array was deployed in the key component area to be collected inside the cabin to collect multi-source acoustic signals. By using a deep learning-based recurrent neural network model, the collected multi-source acoustic signals are separated and processed in real time to obtain the device's operating acoustic fingerprint and form a pre-processed acoustic fingerprint signal. The pure audioprint signal in the equipment operation soundprint is subjected to time-frequency domain normalization processing, and the feature separability of the human ear sensitive frequency band is enhanced by using Mel-scale filter bank to generate a normalized spectrogram.

3. The intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to claim 1, characterized in that, The process of extracting cross-domain features from the preprocessed voiceprint signal, performing feature dimensionality reduction and redundancy elimination through principal component analysis, and constructing a voiceprint feature vector includes: Time-domain features, frequency-domain features, and time-frequency-domain features are extracted from the preprocessed speakerprint signal; Principal component analysis is used to reduce feature dimensionality and eliminate redundancy, constructing voiceprint feature vectors for different components; and the voiceprint feature vectors of different components are stored in a multi-dimensional voiceprint template library.

4. The intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to claim 1, characterized in that, The step of inputting the acoustic spectrogram and voiceprint feature vector into the CNN-LSTM fusion diagnostic model for fault diagnosis, and obtaining the fault diagnosis result, includes: The spectrogram is input into the convolutional neural network in the CNN-LSTM fusion diagnostic model to identify local abnormal patterns in the spectrum. The temporal voiceprint feature vector is input into the long short-term memory network in the CNN-LSTM fusion diagnostic model to capture the changing pattern of fault features over time. Based on the changes in local anomaly patterns and fault characteristics in the spectrum over time, fault diagnosis is performed to obtain fault diagnosis results.

5. The intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to claim 4, characterized in that, The fault diagnosis results include the faulty component, fault type, and severity level.

6. The intelligent fault diagnosis method for multiple components of a wind turbine nacelle according to claim 1, characterized in that, The process of matching fault diagnosis results with a fault-voiceprint knowledge graph to generate decision suggestions includes: When an anomaly is detected in the fault diagnosis results, the fault diagnosis results are matched with the fault-voiceprint knowledge graph, and actionable decision suggestions are output.

7. A multi-component fault intelligent diagnosis system for wind turbine nacelles, characterized in that, include: Preprocessing module: used to collect multi-source acoustic signals inside the wind turbine nacelle, perform preprocessing, obtain preprocessed acoustic signature signals, and generate a spectrogram; Feature vector module: used to extract cross-domain features from preprocessed voiceprint signals, perform feature dimensionality reduction and redundancy elimination through principal component analysis, and construct voiceprint feature vectors; Diagnostic Results Module: Used to input spectrograms and voiceprint feature vectors into the CNN-LSTM fusion diagnostic model to perform fault diagnosis and obtain fault diagnosis results; Decision suggestion module: This module is used to match fault diagnosis results with the fault-voiceprint knowledge graph to generate decision suggestions.

8. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the intelligent diagnosis method for multiple component faults of a wind turbine nacelle as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the intelligent fault diagnosis method for multiple components of a wind turbine nacelle as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the intelligent fault diagnosis method for multiple components of a wind turbine nacelle as described in any one of claims 1-6.

Images