A method, apparatus, equipment and storage medium for detecting abnormalities in wind turbines.

By integrating and filtering the fan noise signal, combining preset signal length and step size for gear segmentation, and dynamically adjusting the filter weight, the problem of inaccurate signal segmentation caused by noise interference is solved, thereby improving the accuracy and reliability of fan anomaly detection.

CN117345678BActive Publication Date: 2026-11-10长三角集成电路工业应用技术创新中心 +2
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202311448565.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-02
Publication Date
2026-11-10
Estimated Expiration
2043-11-02

AI Technical Summary

Technical Problem

In existing methods for detecting abnormalities in wind turbines, noise interference leads to inaccurate signal segmentation, making it impossible to capture continuous changes and resulting in missed detections, thus reducing the accuracy and reliability of the detection.

Method used

By integrating environmental noise signals and fan noise signals, filtering is used to remove specific noise. Gear segmentation is performed by combining preset signal length and step size, and an anomaly detection model is used for independent detection. The filter weights are dynamically adjusted to adapt to different noise environments.

Benefits of technology

It improves the accuracy and reliability of wind turbine anomaly detection, reduces missed detections, and enhances the ability to capture signal timing characteristics and dynamic changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117345678B_ABST
    Figure CN117345678B_ABST
Patent Text Reader

Abstract

This application relates to a method, apparatus, equipment, and storage medium for detecting fan anomalies, applied in the field of fan fault detection. The method includes: acquiring environmental noise signals from an environmental audio acquisition device and fan noise signals from a fan audio acquisition device; integrating the fan noise signals and environmental noise signals according to a method that minimizes error to generate a desired noise signal; filtering specific noise remaining in the desired noise signal; segmenting the filtered desired noise signal into several test segments according to a preset signal length and a preset step size; obtaining the detection results of the test segments using a segment-based anomaly detection model; and determining whether the fan status is abnormal based on whether the detection results are abnormal. The technical advantages of this application are: reducing the possibility of missed detections and improving the accuracy and reliability of fan anomaly detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the technical field of wind turbine fault detection, and in particular to a method, apparatus, equipment and storage medium for wind turbine anomaly detection. Background Technology

[0002] Fans play a vital role in industry, widely used in air conditioning systems, power equipment, and industrial manufacturing. However, due to prolonged operation or other reasons, fans may experience various abnormal conditions, such as mechanical failures, blade damage, and bearing wear. These abnormalities can lead to decreased fan performance, increased energy consumption, or even equipment damage. In fan anomaly detection, noise from the environment, the equipment itself, and other adjacent equipment, including vibrations and sounds, can mask or confuse abnormal signals, making detection difficult. Therefore, effective noise reduction is crucial for accurate fan anomaly detection.

[0003] Currently, the anomaly detection of wind turbines first uses filtering technology to preprocess the collected signals, selectively removing noise within a specific frequency range; then, the signals are divided into several levels by setting thresholds; finally, the corresponding level-specific anomaly detection model is used to identify the abnormal conditions of each level of the wind turbine.

[0004] Using thresholds to segment signals is usually done manually based on experience, experiments, or domain knowledge, which introduces a degree of subjectivity. Furthermore, the threshold method is suitable for segmenting signals into discrete levels. If the operating status of the wind turbine changes continuously between different levels, using threshold segmentation may result in signal loss and failure to capture such continuous changes. This can lead to missed detections when performing level-based anomaly detection on the segmented signals, thereby reducing the accuracy and reliability of wind turbine anomaly detection. Summary of the Invention

[0005] To improve the accuracy and reliability of wind turbine anomaly detection, this application provides a wind turbine anomaly detection method, apparatus, equipment, and storage medium.

[0006] Firstly, this application provides a method for detecting abnormalities in wind turbines, employing the following technical solution: The method includes:

[0007] The ambient noise signal from the environmental audio acquisition device and the fan noise signal from the fan audio acquisition device were acquired respectively.

[0008] The wind turbine noise signal and the environmental noise signal are integrated in a way that minimizes the error to generate the desired noise signal;

[0009] Filter the specific noise remaining in the desired noise signal;

[0010] The filtered desired noise signal is segmented into several levels according to the preset signal length and preset step size to generate several levels to be detected.

[0011] The detection results of several gears to be detected are obtained using a gear-fixing anomaly detection model. The detection results include normal status and abnormal status. The status of the fan is determined to be abnormal based on whether the status of the detection results is abnormal.

[0012] The above technical solution integrates the acquired environmental noise signal and the fan noise signal, filters the integrated noise signal to remove interference from specific noises, segments the filtered desired noise signal into gears, and performs independent anomaly detection for each gear. Compared with using thresholds to segment noise signals, gear segmentation based on preset signal length and preset step size can more accurately analyze the signal characteristics within each signal length and perform segmentation based on signal characteristics, alleviating the subjectivity caused by human experience judgment and improving the accuracy of fan anomaly detection. Judging one by one according to the preset step size can better capture the temporal characteristics and dynamic changes of the signal, even if there may be overlap between adjacent gears, thus mitigating the signal loss and inability to capture continuous changes caused by using threshold methods. In this way, when performing gear-specific anomaly detection, the occurrence of missed detections can be reduced, thereby improving the accuracy and reliability of fan anomaly detection.

[0013] In one specific implementation, the system further includes a filter for adjusting the wind turbine noise signal, wherein integrating the wind turbine noise signal and the ambient noise signal in a manner that minimizes error to generate the desired noise signal includes:

[0014] The filter is used to filter the wind turbine noise signal under the initial weights to generate a predicted noise signal;

[0015] Calculate the noise error between the ambient noise signal and the predicted noise signal;

[0016] If the noise error is less than a preset error threshold, the predicted noise signal is removed from the wind turbine noise signal, and the wind turbine noise signal after removing the predicted noise signal is confirmed as the desired noise signal; otherwise, the weights of the filter are adjusted, and the filter is used again to filter the predicted noise signal under the adjusted weights.

[0017] By integrating environmental noise signals and fan noise signals, dynamically adjusting filter weights, and canceling the fan noise signal with the filtered predicted noise signal, a more accurate fan noise signal can be obtained, reducing environmental noise interference and helping to determine the fan's status, thus improving the accuracy of subsequent fan anomaly detection. The filter weights in this solution are adjusted based on noise error. By calculating the noise error between the environmental noise signal and the predicted noise signal and comparing it with a preset error threshold, the filter weights can be dynamically adjusted. This dynamic adjustment allows the filter to more effectively adapt to different noise environments and improves the accuracy of the predicted noise signal.

[0018] In one specific implementation, the step of using the filter to filter the wind turbine noise signal under default weights to generate a predicted noise signal includes:

[0019] Define the initial weight vector of the filter and the wind turbine noise vector corresponding to the wind turbine noise signal;

[0020] The predicted noise signal is obtained by performing a dot product operation on the weight vector and the wind turbine noise vector.

[0021] Through the above technical solution, the initial weight vector of the filter can be defined according to specific needs and expected filtering effects. By adjusting the value of the weight vector, filtering of noise signals of different frequencies, amplitudes, or phases can be achieved, thereby meeting specific application requirements. By performing a dot product operation between the weight vector and the fan noise vector, a predicted noise signal can be obtained. This predicted noise signal can be used to analyze, monitor, or control noise problems in the system. Through filtering, noise interference can be reduced, signal quality and clarity can be improved, thereby improving the performance and reliability of the system.

[0022] In one specific implementation, adjusting the weights of the filter includes:

[0023] The weight increment is determined by combining the preset control factor, the noise error, and the wind turbine noise signal. The control factor is used to control the degree of adjustment of the filter weight.

[0024] The weights of the filter are adjusted based on the weight increment.

[0025] The above technical solution considers a preset control factor, which can control the adjustment range of the filter. By combining the control factor, the speed and stability of the filter adjustment can be balanced, avoiding excessively fast or slow changes in weights, thereby improving the performance and stability of the filter. By considering noise error and fan noise signal to determine the weight increment, the filter can more adaptively adapt to different noise environments and fan conditions, thereby improving the filter's noise processing effect. This weight adjustment strategy can improve the filter's adaptability and stability, thereby improving the filtering effect and the accuracy of the desired noise signal.

[0026] In one specific implementation, the specific noise remaining in the desired noise signal includes: high-frequency noise, low-frequency noise, narrowband noise, and broadband noise.

[0027] By using the above technical solutions, specific noises, including high-frequency noise, low-frequency noise, narrowband noise, and broadband noise, can be filtered to reduce interference components in the desired noise signal and improve its quality and purity. This processing helps to more accurately reflect the actual noise characteristics of the wind turbine, thereby improving the accuracy and reliability of wind turbine anomaly detection.

[0028] In one specific implementation, the step of segmenting the filtered desired noise signal into levels according to a preset signal length and a preset step size includes:

[0029] The filtered desired noise signal is divided into several regions to be judged according to the preset signal length.

[0030] The short-time Fourier transform algorithm is used to obtain amplitude information of several regions to be judged, and the amplitude information includes the amplitude corresponding to each time step of the region to be judged.

[0031] The region to be judged is divided into several time sub-regions according to a preset step size, and the starting time of several time sub-regions is determined as several time points.

[0032] Calculate the amplitude difference between the amplitude corresponding to each time point and the average amplitude of the region to be judged;

[0033] The time point with the largest amplitude difference is determined as the intercept point, and the desired noise signal is segmented into gears based on the intercept point.

[0034] The above technical solution uses a short-time Fourier transform algorithm to convert the desired noise signal from the time domain to the frequency domain, obtaining the amplitude signal of each region to be judged. This allows for analysis of the desired noise signal in the frequency domain, extracting the amplitude signal. This analysis provides a more comprehensive understanding of the frequency characteristics of the noise signal, helping to more accurately determine the time point of amplitude abrupt changes. By calculating the difference between the amplitude and the average amplitude, the time point with the largest amplitude difference can be found, accurately locating the time point of amplitude abrupt changes and improving the accuracy of gear segmentation. By determining the interception point and performing gear segmentation using a preset signal length and preset step size, different gears can be adaptively divided. Even if there is overlap between adjacent gears, the temporal characteristics and dynamic changes of the signal can be better captured, alleviating the signal loss and inability to capture continuous changes caused by using threshold methods, thereby improving the accuracy and reliability of wind turbine anomaly detection.

[0035] In one specific implementation scheme, determining whether the fan status is abnormal based on whether the detection result is abnormal includes:

[0036] If any abnormal status is detected in the test results of the gear to be tested, the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal.

[0037] With the above technical solution, if the detection result of at least one of the test gears is abnormal, the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal. The fan status can be determined through a simple and intuitive logical judgment method. By considering the detection results of multiple test gears, multiple signals can be comprehensively considered. Compared with relying solely on the detection result of a single gear, comprehensive judgment of multiple gears can reduce the possibility of missed detections, thereby improving the accuracy and reliability of fan anomaly detection.

[0038] Secondly, this application provides a wind turbine anomaly detection device, which adopts the following technical solution: the device includes:

[0039] The noise signal acquisition module is used to acquire the environmental noise signal from the environmental audio collector and the fan noise signal from the fan audio collector, respectively.

[0040] The noise signal integration module is used to integrate the wind turbine noise signal and the environmental noise signal in a way that minimizes error, and generate a desired noise signal;

[0041] A filtering module is used to filter specific noise remaining in the desired noise signal;

[0042] The gear segmentation module is used to segment the filtered desired noise signal into gears according to a preset signal length and a preset step size, generating several gears to be detected.

[0043] An anomaly detection module is used to obtain the detection results of several gears to be detected using a gear-specific anomaly detection model. The detection results include normal status and abnormal status. The fan status is determined to be abnormal based on whether the status of the detection results is abnormal.

[0044] Thirdly, this application provides a computer device that adopts the following technical solution: it includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as any of the above-described wind turbine anomaly detection methods.

[0045] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: storing a computer program that can be loaded by a processor and executed by any of the above-mentioned wind turbine anomaly detection methods.

[0046] In summary, this application has the following beneficial technical effects:

[0047] This application integrates the acquired environmental noise signal and the fan noise signal, filters the integrated noise signal to remove interference from specific noises, segments the filtered desired noise signal into levels, and performs independent anomaly detection for each level. Compared with using thresholds to segment the noise signal, segmenting based on preset signal length and preset step size allows for more precise analysis of signal characteristics within each signal length. Segmentation based on these signal characteristics alleviates subjectivity caused by human experience judgment and improves the accuracy of fan anomaly detection. Judging each level individually according to the preset step size, even if there is overlap between adjacent levels, better captures the temporal characteristics and dynamic changes of the signal, mitigating the signal loss and inability to capture continuous changes caused by threshold methods. This reduces missed detections when performing level-specific anomaly detection for each level, thereby improving the accuracy and reliability of fan anomaly detection. Attached Figure Description

[0048] Figure 1 This is a flowchart of the fan anomaly detection method in the embodiments of this application;

[0049] Figure 2 This is a schematic diagram illustrating the filter in an embodiment of this application;

[0050] Figure 3 This is a schematic diagram of the fan detection device in the embodiments of this application;

[0051] Figure 4 This is a schematic diagram used to illustrate a computer device in the embodiments of this application.

[0052] Reference numerals: 301, Noise signal acquisition module; 302, Noise signal integration module; 303, Filtering module; 304, Gear segmentation module; 305, Anomaly detection module. Detailed Implementation

[0053] The following is in conjunction with the appendix Figure 1-4 This application will be described in further detail.

[0054] Currently, anomaly detection for wind turbines typically involves the following steps: First, filtering techniques are used to preprocess the acquired signal, selectively removing noise within a specific frequency range. Second, a threshold is set to divide the signal into multiple levels. Finally, a corresponding level-based anomaly detection model is used to identify anomalies at each level of the wind turbine. However, threshold settings are usually based on human experience, experiments, or domain knowledge, which introduces a degree of subjectivity. Furthermore, while threshold methods are suitable for dividing signals into discrete levels, if the wind turbine's operating state changes continuously between different levels, threshold segmentation may lead to signal loss, failing to capture such continuous changes and resulting in missed detections. This could reduce the accuracy and reliability of wind turbine anomaly detection.

[0055] To address the aforementioned issues, this application discloses a method for detecting fan anomalies. This method is applied to a fan anomaly detection system, which includes at least an environmental audio acquisition unit for acquiring ambient noise signals and a fan audio acquisition unit for acquiring fan noise signals. The environmental audio acquisition unit and the fan audio acquisition unit are devices or sensors used to capture sound signals in the environment. They can convert sound into digital signals for further processing and analysis. For example, the environmental audio acquisition unit and the fan audio acquisition unit can be microphones, sound vibration sensors, etc. The fan audio acquisition unit is installed near the fan but not in direct contact with it. This mitigates the problem of signal interference or distortion caused by mechanical vibrations generated during fan operation being transmitted to the sensor through direct contact, resulting in more accurate noise acquisition results. The environmental audio acquisition unit is installed away from the fan, ensuring accurate measurement of ambient background noise without interference from fan noise. The fan anomaly detection system should also include a filter for adjusting the frequency characteristics of the fan noise signal.

[0056] like Figure 1 As shown, the method includes the following steps:

[0057] S10, acquire the environmental noise signal from the environmental audio collector and the fan noise signal from the fan audio collector respectively.

[0058] Specifically, environmental noise signals can provide the sound level and characteristics of the background environment, serving as a comparative reference for fan noise signals. By understanding the level of environmental noise, the relative impact of fan noise can be better assessed. Fan noise is the sound generated by the fan equipment. The normal operation of the fan usually produces specific noise characteristics. By monitoring and analyzing fan noise signals, abnormal noise signals of the fan can be detected.

[0059] S20 integrates the fan noise signal and the environmental noise signal in a way that minimizes the error to generate the desired noise signal.

[0060] Specifically, minimizing error is an optimization objective used to measure the difference between the expected output and the actual output. In this case, by minimizing the error, the wind turbine noise signal and the environmental noise signal are integrated together to generate a desired noise signal that accurately reproduces the noise situation during wind turbine operation, thereby providing a more reliable reference for subsequent analysis, evaluation, or decision-making.

[0061] S30 filters out specific noise residue from the desired noise signal.

[0062] Specifically, residual specific noise refers to specific noise components that still exist in the desired noise signal. Even after integration and processing, the desired noise signal may still contain certain types of noise, which may be related to wind turbine operation or the environment. Filtering is a signal processing technique used to adjust or change the frequency response characteristics of a signal to remove or reduce specific frequency components. In this case, filtering the residual specific noise in the desired noise signal aims to remove or reduce the influence of these specific noises, thereby obtaining a cleaner and more accurate noise signal. Filtering can use different filtering methods, such as digital filters, moving average filters, low-pass filters, and high-pass filters. The specific selection of a filtering method may be based on the specific type of noise to be removed, frequency characteristics, and signal quality requirements. This will help improve the accuracy and reliability of the data and provide a clearer noise signal for subsequent processing and applications.

[0063] S40 divides the filtered desired noise signal into several gears based on the preset signal length and preset step size, generating several gears to be detected.

[0064] Specifically, the filtered desired noise signal is first used as input, which can be a digital sequence or a vector of signals representing the processed noise data. Based on the signal length and step size, the boundary of each gear is calculated, thus determining the start and end positions of each gear. According to the calculated gear boundaries, the desired noise signal is segmented, with each gear containing data between its corresponding start and end positions. Finally, each segmented gear is used as the gear to be detected for subsequent fan anomaly detection. The signal length refers to the length by which the desired noise signal is segmented, representing its time range. The step size is the time interval between each movement during filtering or other operations.

[0065] S50 uses a fixed-gear anomaly detection model to obtain the detection results of several gears to be detected. The detection results include normal status and abnormal status. The status of the fan is determined to be abnormal based on whether the status of the detection results is abnormal.

[0066] Specifically, for the anomaly detection model, one approach is to extract the power spectrum or fast Fourier spectrum features of the signal, observe the amplitude characteristics at specific frequency points based on industry experience, and determine whether anomalies exist through methods such as peak-to-trough difference and adjacent peak difference. Another approach is based on machine learning, which identifies abnormal gears through feature extraction, model training, and model prediction. A third approach involves collecting a large amount of labeled audio data for deep learning model training, and then inputting subsequent data into the model to predict anomalies. The choice of method depends on the actual application scenario. The first approach is cost-effective and suitable for scenarios with low computing power. The second approach has good generalization and can be deployed on embedded platforms. The third approach has high accuracy, good generalization, and robustness, but requires high computing power and has high initial data collection costs. This solution can determine the appropriate anomaly detection model based on the specific scenario.

[0067] The wind turbine anomaly detection method integrates acquired environmental noise signals and wind turbine noise signals. The integrated noise signal is then filtered to remove specific noise interference. The filtered desired noise signal is then segmented by level, and independent anomaly detection is performed for each level. Compared to segmenting noise signals using thresholds, level segmentation based on preset signal lengths and step sizes allows for more precise analysis of signal characteristics within each signal length. Segmentation based on these characteristics alleviates subjectivity caused by human experience and improves the accuracy of wind turbine anomaly detection. By judging each level sequentially according to the preset step size, even if there is overlap between adjacent levels, the temporal characteristics and dynamic changes of the signal can be better captured, mitigating signal loss and the inability to capture continuous changes caused by threshold methods. This level-specific anomaly detection reduces missed detections, thereby improving the accuracy and reliability of wind turbine anomaly detection.

[0068] In one embodiment, to minimize the difference between the wind turbine noise signal and the ambient noise signal, after acquiring the ambient noise signal and the wind turbine noise signal, the step of integrating the wind turbine noise signal and the ambient noise signal according to a method that minimizes error to generate the desired noise signal can be specifically performed as follows:

[0069] First, the fan noise signal is filtered using a filter with default weights to generate a predicted noise signal. Then, the noise error between the ambient noise signal and the predicted noise signal is calculated. This noise error measures how closely the predicted noise signal approximates the actual ambient noise. If the noise error is less than a preset error threshold, it indicates that the predicted noise signal is relatively close to the ambient noise signal. In this case, the predicted noise signal is removed from the fan noise signal, resulting in relatively pure fan noise. This fan noise signal is then identified as the desired noise signal, which can be used to analyze, monitor, or control noise problems in the system. Otherwise, it indicates that the predicted noise signal differs significantly from the ambient noise signal, and the filter's default weights may not be effective in handling the noise. Therefore, the filter weights are adjusted, and the filter is used again with the adjusted weights to better adapt to the noise environment. For example: Figure 2As shown, let the environmental noise signal be the desired signal d(n), the fan noise signal be the reference signal x(n), and y(n) be the output signal of x(n) after digital filtering; e(n) is the error signal, representing the difference between the desired signal and the filter output signal y(n), which is also the main basis for the adaptive module to modify the transverse filter parameters; w(n) is the filter weight vector, representing the filter parameters; μ is the learning rate, used to control the update speed of the weights. A smaller μ value slows down the update speed of the weights, while a larger μ value accelerates the update speed of the weights; X(n) is the desired noise signal after noise reduction; when applying the minimum error method, it can be calculated according to the following steps:

[0070] First, initialize the filter by setting the weights w(n) to 0. Then, perform iterative calculations based on the reference and desired signals: calculate the output signal after the adaptive filter processes the reference signal: y(n) = w(n) * x(n). T Find the residual error signal: e(n) = d(n) - y(n); Adjust the weights of the filter according to the error: w(n+1) = w(n) + μ*x(n)*e(n); After n+1 iterations of the weight w(n+1), if the error signal satisfies the convergence condition (the convergence condition can be determined according to actual needs), the desired noise signal after denoising is: X(n) = x(n) - y(n).

[0071] By integrating environmental noise signals and fan noise signals, dynamically adjusting filter weights, and subtracting the filtered predicted noise signal from the fan noise signal, a more accurate and cleaner fan noise signal can be obtained. This reduces environmental noise interference, helps determine the fan's status, and improves the accuracy of subsequent fan anomaly detection. The filter weights in this scheme are adjusted based on noise error. By calculating the noise error between the environmental noise signal and the predicted noise signal and comparing it with a preset error threshold, the filter weights can be dynamically adjusted. This dynamic adjustment allows the filter to adapt more effectively to different noise environments and improves the accuracy of the predicted noise signal.

[0072] In one embodiment, to reduce noise interference, a filter is used to filter the wind turbine noise signal with default weights. The step of generating the predicted noise signal can be specifically performed as follows:

[0073] Define the initial weight vector of the filter and the corresponding wind turbine noise vector of the wind turbine noise signal; perform a dot product operation on the weight vector and the wind turbine noise vector to obtain the predicted noise signal. The filter achieves the filtering effect by weighting and combining the input signals. In this step, it is necessary to define the initial weight vector of the filter, where each element represents the weight value at the corresponding position of the filter; at the same time, it is necessary to convert the wind turbine noise signal into a wind turbine noise vector, which contains the values ​​of each sampling point of the wind turbine noise signal; perform a dot product operation on the defined initial weight vector and the wind turbine noise vector, specifically, multiply each weight value in the weight vector by the value at the corresponding position in the wind turbine noise vector to obtain the value of the predicted noise signal; for example, the initial weight vector of the filter is defined as (w1, w2, ..., w k The fan noise vector corresponding to the fan noise signal is defined as (x1, x2, ..., x). k The weight vector and the wind turbine noise vector are multiplied by a dot product to obtain the predicted noise signal:

[0074] y(n)=w n (x1+x2+…+x k )

[0075] Y = (y1, y2, ..., y n )

[0076] The initial weight vector of the filter can be defined according to specific needs and expected filtering effects. By adjusting the value of the weight vector, filtering of noise signals of different frequencies, amplitudes, or phases can be achieved to meet specific application requirements. By performing a dot product operation between the weight vector and the fan noise vector, a predicted noise signal can be obtained. This predicted noise signal can be used to analyze, monitor, or control noise problems in the system. Through filtering, noise interference can be reduced, signal quality and clarity can be improved, thereby improving the performance and reliability of the system.

[0077] In one embodiment, to improve the filter's noise handling performance, the step of adjusting the filter weights can be specifically performed as follows:

[0078] The weight increment is determined by combining the preset control factor, noise error, and fan noise signal. The weight increment represents the amount by which the default weight of the filter needs to be adjusted. The default weight of the filter is adjusted based on the weight increment. Specifically, the weight increment is added to the default weight to obtain the adjusted weight. In this way, the characteristics and response of the filter will change according to the weight adjustment, making it more adaptable to the current noise environment and fan noise signal.

[0079] By considering preset control factors, the adjustment range of the filter can be controlled. By combining control factors, the speed and stability of filter adjustment can be balanced, avoiding excessively fast or slow changes in weights, thereby improving the performance and stability of the filter. By considering noise error and fan noise signal to determine the weight increment, the filter can more adaptively adapt to different noise environments and fan conditions, thereby improving the filter's noise processing effect. This weight adjustment strategy can improve the filter's adaptability and stability, thereby improving the filtering effect and the accuracy of the desired noise signal.

[0080] In one embodiment, in order to reduce the interference components in the desired noise signal, the specific noise remaining in the desired noise signal includes: high-frequency noise, low-frequency noise, narrowband noise, and broadband noise.

[0081] High-frequency noise refers to noise signals that vary within a higher frequency range. This type of noise typically contains short periods or rapidly changing vibrational components. High-frequency noise may originate from switching circuits in electronic devices, electromagnetic interference, or other rapidly changing signal sources. Low-frequency noise refers to noise signals that vary within a lower frequency range. This type of noise typically contains longer periods or slowly changing vibrational components. Low-frequency noise may originate from power fluctuations, mechanical vibrations, or other slowly changing signal sources. Narrowband noise is noise signals that occur within a relatively narrow frequency range. This type of noise is usually concentrated near a specific frequency point or band and may have a high amplitude. Narrowband noise can originate from interference from specific devices or systems, such as signal interference from wireless communication devices. Wideband noise is noise signals that occur within a relatively wide frequency range. This type of noise has a certain degree of energy distribution throughout the entire spectrum and is not clearly concentrated at a specific frequency point or band. Wideband noise can originate from various uncertain noise sources, such as atmospheric noise in the natural environment or thermal noise in electronic devices.

[0082] By filtering specific noises, including high-frequency noise, low-frequency noise, narrowband noise, and broadband noise, interference components in the desired noise signal can be reduced, improving its quality and purity. This processing helps to more accurately reflect the actual noise characteristics of the wind turbine, thereby improving the accuracy and reliability of wind turbine anomaly detection.

[0083] In one embodiment, to mitigate the problems of signal loss and inability to capture continuous changes caused by using a threshold method, the step of segmenting the filtered desired noise signal into levels according to a preset signal length and a preset step size can be specifically performed as follows:

[0084] The filtered desired noise signal is divided into several regions to be judged based on a preset signal length. A short-time Fourier transform (SFT) algorithm is used to obtain amplitude information for each region to be judged. This amplitude information includes the amplitude at each moment within the region. The SFT can perform local analysis of the signal in time and frequency to obtain the spectral information at each time point. The regions to be judged are divided into several time sub-regions according to a preset step size, and the starting times of these sub-regions are determined as several time points. The amplitude difference between the amplitude at each time point and the average amplitude of the region to be judged is calculated. This amplitude difference indicates the degree of deviation of the amplitude of the sub-region relative to the average amplitude of the entire region to be judged. The time point with the largest amplitude difference is determined as the intercept point. This intercept point indicates that the amplitude deviation is greatest at this location, which is likely the boundary position of the noise signal. Based on the intercept point, the desired noise signal is segmented into different levels to identify the boundary positions of the noise.

[0085] It should be noted that this application describes a gear segmentation algorithm based on a preset signal length and step size, which uses short-time Fourier transform and difference analysis to determine the cut-off point. However, the specific implementation of this algorithm may vary depending on the application scenario and requirements. In addition to the gear segmentation algorithm exemplified in this application, other suitable gear segmentation algorithms can be used, such as classification methods based on support vector machines (SVM). Furthermore, the gear segmentation algorithm and the gear anomaly detection algorithm are interrelated and can be implemented simultaneously in the program design to improve real-time performance. By introducing an anomaly detection algorithm during gear segmentation, abnormal situations can be detected and marked in a timely manner to alert the system operator or automated program to handle them, thereby enhancing the system's real-time performance and responsiveness, and enabling timely response to anomalies that may affect system performance. By performing gear segmentation and gear anomaly detection simultaneously, the program design can be improved in terms of real-time performance. This parallel processing method allows the system to detect and respond to possible anomalies while processing the desired noise signal, which is crucial for real-time processing and control of noise signals and can improve the robustness and reliability of the system.

[0086] The short-time Fourier transform algorithm is used to transform the desired noise signal from the time domain to the frequency domain, obtaining the amplitude information of each region to be judged. The desired noise signal can be analyzed in the frequency domain to extract amplitude information. This analysis can provide a more comprehensive understanding of the frequency characteristics of the noise signal and help to more accurately determine the time point of amplitude change. By calculating the difference between the amplitude and the average amplitude, the time point with the largest amplitude difference can be found, accurately locating the time point of amplitude change and improving the accuracy of gear segmentation. By determining the interception point and performing gear segmentation by setting the preset signal length and preset step size, different gears can be adaptively divided. Even if there may be overlap between adjacent gears, the temporal characteristics and dynamic changes of the signal can be better captured, alleviating the problems of signal loss and inability to capture continuous changes caused by using the threshold method, thereby improving the accuracy and reliability of wind turbine anomaly detection.

[0087] In one embodiment, to reduce the possibility of missed detections, the step of determining whether the fan status is abnormal based on whether the detection result is abnormal can be specifically performed as follows:

[0088] Obtain the detection results corresponding to each gear to be tested; if there are abnormal detection results in the detection results of the gear to be tested, such as excessive amplitude or abnormal temperature, then the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal.

[0089] If at least one of the test gears shows an abnormal result, the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal. The fan status can be determined through a simple and intuitive logical judgment method. By considering the test results of multiple test gears, multiple signals can be comprehensively considered. Compared with relying solely on the test result of a single gear, comprehensive judgment of multiple gears can reduce the possibility of missed detections, thereby improving the accuracy and reliability of fan anomaly detection.

[0090] Figure 1 This is a flowchart illustrating a wind turbine anomaly detection method in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows; unless explicitly stated otherwise, there is no strict order requirement for the execution of these steps, and they can be executed in other orders; and Figure 1 At least some of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0091] Based on the above method, this application also discloses a fan anomaly detection device.

[0092] Reference Figure 3 The device includes the following modules:

[0093] The noise signal acquisition module 301 is used to acquire the environmental noise signal from the environmental audio collector and the fan noise signal from the fan audio collector, respectively.

[0094] The noise signal integration module 302 is used to integrate the fan noise signal and the environmental noise signal in a way that minimizes the error, and generate the desired noise signal.

[0095] The filtering module 303 is used to filter specific noise remaining in the desired noise signal;

[0096] The gear segmentation module 304 is used to segment the filtered desired noise signal into gears according to the preset signal length and preset step size, and generate several gears to be detected.

[0097] The anomaly detection module 305 is used to obtain the detection results of several gears to be detected using the fixed-gear anomaly detection model. The detection results include normal status and abnormal status. The fan status is determined to be abnormal based on whether the status of the detection results is abnormal.

[0098] In one embodiment, the noise signal integration module 302 is specifically used to filter the fan noise signal using a filter with default weights to generate a predicted noise signal; calculate the noise error between the ambient noise signal and the predicted noise signal; if the noise error is less than a preset error threshold, remove the fan noise signal from the predicted noise signal and confirm the fan noise signal after removing the predicted noise signal as the desired noise signal; otherwise, adjust the weights of the filter and use the filter again to filter the predicted noise signal with the adjusted weights.

[0099] In one embodiment, the noise signal integration module 302 is specifically used to determine the weight increment by combining a preset control factor, noise error and fan noise signal; and to adjust the default weight of the filter based on the weight increment.

[0100] In one embodiment, the filtering module 303 is specifically used to process specific noises remaining in the desired noise signal, including: high-frequency noise, low-frequency noise, narrowband noise, and broadband noise.

[0101] In one embodiment, the gear segmentation module 304 is specifically used to divide the filtered desired noise signal into several regions to be judged according to a preset signal length; use a short-time Fourier transform algorithm to obtain the amplitude information of several regions to be judged, the amplitude information including the amplitude corresponding to each moment of the region to be judged; divide the regions to be judged into several time sub-regions according to a preset step size, and determine the starting time of several time sub-regions as several time points; calculate the amplitude difference between the amplitude corresponding to each time point and the average amplitude of the region to be judged; determine the time point with the largest amplitude difference as the intercept point, and perform gear segmentation on the desired noise signal according to the intercept point.

[0102] In one embodiment, the anomaly detection module 305 is specifically used to determine that if there is an abnormal detection result in the detection result of the gear to be detected, the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal.

[0103] The fan malfunction detection device provided in this application embodiment can be applied to the fan malfunction detection method provided in the above embodiment. For relevant details, please refer to the above method embodiment. The implementation principle and technical effect are similar, and will not be repeated here.

[0104] It should be noted that the wind turbine anomaly detection device provided in this embodiment is only illustrated by the above-described division of functional modules / units when performing wind turbine anomaly detection. In practical applications, the above functions can be assigned to different functional modules / units as needed, that is, the internal structure of the wind turbine anomaly detection device can be divided into different functional modules / units to complete all or part of the functions described above. Furthermore, the implementation method of the wind turbine anomaly detection method provided in the above-described method embodiments and the implementation method of the wind turbine anomaly detection device provided in this embodiment belong to the same concept. The specific implementation process of the wind turbine anomaly detection device provided in this embodiment is detailed in the above-described method embodiments and will not be repeated here.

[0105] This application also discloses a computer device.

[0106] Specifically, such as Figure 4As shown, the computer device can be a desktop computer, laptop computer, handheld computer, or cloud server, etc. The computer device may include, but is not limited to, a processor and memory. The processor and memory can be connected via a bus or other means. The processor can be a Central Processing Unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, graphics processing units (GPUs), embedded neural network processing units (NPUs) or other dedicated deep learning coprocessors, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0107] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above embodiments of this application. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the methods in the above embodiments. The memory may include a program storage area and a data storage area, wherein the program storage area may store the control unit and the application program required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0108] This application also discloses a computer-readable storage medium.

[0109] Specifically, the computer-readable storage medium is used to store a computer program, which, when executed by a processor, implements the methods described in the above-described method embodiments. Those skilled in the art will understand that implementing all or part of the processes in the methods described in the above-described embodiments of this application can be accomplished by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0110] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for detecting fan malfunctions, characterized in that, The method is applied to a wind turbine detection system, the system including an environmental audio acquisition unit for acquiring environmental noise signals and a wind turbine audio acquisition unit for acquiring wind turbine noise signals, the method including: The ambient noise signal from the environmental audio acquisition device and the fan noise signal from the fan audio acquisition device were acquired respectively. The wind turbine noise signal and the environmental noise signal are integrated in a way that minimizes the error to generate the desired noise signal; Filter the specific noise remaining in the desired noise signal; The filtered desired noise signal is segmented into several levels according to the preset signal length and preset step size to generate several levels to be detected. The detection results of several gears to be detected are obtained using a gear-fixing anomaly detection model. The detection results include normal status and abnormal status. The status of the fan is determined to be abnormal based on whether the status of the detection results is abnormal. The step of segmenting the filtered desired noise signal into levels according to a preset signal length and a preset step size includes: The filtered desired noise signal is divided into several regions to be judged according to the preset signal length. The short-time Fourier transform algorithm is used to obtain amplitude information of several regions to be judged, and the amplitude information includes the amplitude corresponding to each time step of the region to be judged. The region to be judged is divided into several time sub-regions according to a preset step size, and the starting time of several time sub-regions is determined as several time points. Calculate the amplitude difference between the amplitude corresponding to each time point and the average amplitude of the region to be judged; The time point with the largest amplitude difference is determined as the intercept point, and the desired noise signal is segmented into gears based on the intercept point.

2. The method according to claim 1, characterized in that, The system further includes a filter for adjusting the wind turbine noise signal, wherein integrating the wind turbine noise signal and the ambient noise signal in a manner that minimizes error to generate the desired noise signal includes: The filter is used to filter the wind turbine noise signal under the initial weights to generate a predicted noise signal; Calculate the noise error between the environmental noise signal and the predicted noise signal; If the noise error is less than a preset error threshold, the predicted noise signal is removed from the wind turbine noise signal, and the wind turbine noise signal after removing the predicted noise signal is confirmed as the desired noise signal; otherwise, the weights of the filter are adjusted, and the filter is used again to filter the predicted noise signal under the adjusted weights.

3. The method according to claim 2, characterized in that, The step of using the filter to filter the wind turbine noise signal under default weights to generate a predicted noise signal includes: Define the initial weight vector of the filter and the wind turbine noise vector corresponding to the wind turbine noise signal; The predicted noise signal is obtained by performing a dot product operation on the weight vector and the wind turbine noise vector.

4. The method according to claim 2, characterized in that, The adjustment of the filter weights includes: The weight increment is determined by combining the preset control factor, the noise error, and the wind turbine noise signal. The control factor is used to control the degree of adjustment of the filter weight. The weights of the filter are adjusted based on the weight increment.

5. The method according to claim 1, characterized in that, The specific noise remaining in the desired noise signal includes: high-frequency noise, low-frequency noise, narrowband noise, and broadband noise.

6. The method according to claim 1, characterized in that, The step of determining whether the fan status is abnormal based on whether the detection result is abnormal includes: If any abnormal status is detected in the test results of the gear to be tested, the fan status is determined to be abnormal; otherwise, the fan status is determined to be normal.

7. A fan testing device, characterized in that, The device includes: The noise signal acquisition module (301) is used to acquire the environmental noise signal from the environmental audio collector and the fan noise signal from the fan audio collector, respectively. The noise signal integration module (302) is used to integrate the fan noise signal and the environmental noise signal in a way that minimizes the error, and generate a desired noise signal; The filtering module (303) is used to filter the specific noise remaining in the desired noise signal; The gear segmentation module (304) is used to perform gear segmentation on the filtered desired noise signal according to a preset signal length and a preset step size, generating several gears to be detected. Specifically, it is used to divide the filtered desired noise signal into several regions to be judged according to the preset signal length; use a short-time Fourier transform algorithm to obtain the amplitude information of several regions to be judged, the amplitude information including the amplitude corresponding to each moment of the region to be judged; divide the region to be judged into several time sub-regions according to a preset step size, and determine the starting time of several time sub-regions as several time points; calculate the amplitude difference between the amplitude corresponding to each time point and the average amplitude of the region to be judged; determine the time point with the largest amplitude difference as the intercept point, and perform gear segmentation on the desired noise signal according to the intercept point. An anomaly detection module (305) is used to obtain the detection results of several gears to be detected using a fixed-gear anomaly detection model. The detection results include normal status and abnormal status. The fan status is determined to be abnormal based on whether the status of the detection results is abnormal.

8. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed according to any one of claims 1 to 6.

Citation Information

Patent Citations

  • Method for detecting abnormity of fan in outdoor unit, control device and air conditioner outdoor unit

    CN116907029A