High-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology

Abstract

The invention relates to the technical field of high-voltage switch fault diagnosis systems, and discloses a high-voltage switch fault diagnosis system based on an artificial intelligence audio recognition technology, and the system comprises at least one audio collection device which is used for collecting audio signals generated in the operation process of a high-voltage switch; the data preprocessing module is used for preprocessing the audio signals, and the feature extraction module is used for extracting feature vectors from the preprocessed audio signals. According to the high-voltage switch fault diagnosis system based on the artificial intelligence audio recognition technology, by using the high-sensitivity MEMS microphone and the advanced audio processing technology, weak abnormal sound in the operation process of a high-voltage switch can be accurately captured, the fault diagnosis precision is improved, various sensor data such as a temperature sensor and a vibration sensor are combined, and the fault diagnosis accuracy is improved. The accuracy and reliability of fault diagnosis are improved, the system can automatically adjust the sensitivity parameters of the model according to the environmental noise level, and false alarms are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of high-voltage switch fault diagnosis systems, and in particular to a high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology. Background Art

[0002] The high-voltage switch fault diagnosis system is an intelligent system based on artificial intelligence audio recognition technology. It aims to identify and diagnose the fault type of the high-voltage switch by collecting and analyzing the audio signals generated during the operation of the high-voltage switch. However, the existing technology: traditional fault diagnosis methods rely on fixed thresholds or simple statistical methods, which are easily affected by environmental noise, resulting in a high false alarm rate. In addition, traditional fault diagnosis methods can usually only identify some obvious fault types, and have weak recognition capabilities for complex fault types. At the same time, data processing methods usually take a long time to complete data preprocessing and feature extraction, affecting the real-time performance of the system. Summary of the invention

[0003] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology includes:

[0006] At least one audio acquisition device, used to collect audio signals generated during the operation of the high-voltage switch;

[0007] A data preprocessing module, used for preprocessing the audio signal;

[0008] a feature extraction module for extracting feature vectors from the preprocessed audio signal;

[0009] a fault diagnosis module, using a pre-trained deep learning model to identify a fault type of the high-voltage switch according to the feature vector;

[0010] A result output module is used to display the diagnosis results.

[0011] As a further solution of the present invention: the at least one audio acquisition device for collecting audio signals generated during the operation of the high-voltage switch includes:

[0012] Select the SPH0645LM4H-B MEMS microphone, which has an operating frequency range of 20Hz to 20kHz and a signal-to-noise ratio of up to 62dB;

[0013] Install the microphone near the input and output terminals of the high-voltage switch and in the middle of the switch body, and use a shock-proof bracket to fix the microphone on the housing of the high-voltage switch;

[0014] The sampling rate was set to 44.1kHz, 16-bit quantization was selected to balance data quality and storage requirements, and the collected data was saved in WAV format;

[0015] Add a preamplifier at the microphone output, set the gain to 20dB, and add a low-pass filter before the ADC with a cutoff frequency of 20kHz.

[0016] The audio data is transmitted to a data processing center using wired or wireless means, and at the data processing center, the received audio data is stored on a NAS or cloud storage platform.

[0017] As a further solution of the present invention: the data preprocessing module is used to preprocess the audio signal, including:

[0018] Read audio data saved in WAV format from NAS or cloud storage platforms, and use Python's scipy.io.wavfile library to read WAV files;

[0019] Use spectral subtraction to reduce the noise of the audio signal. The expression is:

[0020] y(n)=x(n)-α·N(n);

[0021] Where y(n) represents the signal after denoising, x(n) represents the original signal, N(n) represents the estimated noise signal, and α represents the noise reduction coefficient;

[0022] Use a bandpass filter to filter the audio signal to remove high-frequency and low-frequency noise. The expression is:

[0023] y[n]=b0x[n]+b1x[n-1]+b2x[n-2]-a1y[n-1]-a2y[n-2];

[0024] y[n] represents the filtered signal, x[n] represents the input signal, b0, b1, b2 and a1, a2 are the filter coefficients;

[0025] Normalize the audio signal to the range of [-1,1] to eliminate amplitude differences. The expression is:

[0026]

[0027] y[n] represents the normalized signal, and x[n] represents the original signal;

[0028] The audio signal is divided into multiple short-time frames, with the frame length set to 20ms and the frame shift set to 10ms.

[0029] As a further solution of the present invention: the feature extraction module is used to extract feature vectors from the preprocessed audio signal, including:

[0030] Calculate the Mel-frequency cepstral coefficient (MFCC) of each audio frame, the expression is:

[0031]

[0032] Where X[k] represents the frequency domain signal, x[n] represents the time domain signal, NN represents the frame length, m represents the Mel frequency, f represents the linear frequency, Em represents the logarithmic energy of the mth Mel filter, Hm(k) represents the frequency response of the mth Mel filter, cn represents the nth MFCC coefficient, and M represents the number of Mel filters;

[0033] Calculate the zero crossing rate (ZCR) of each audio frame, the expression is:

[0034]

[0035] Where ZCR represents the zero crossing rate, NN represents the frame length, and 1{x[n]x[n-1]<0} represents the indicator function;

[0036] Calculate the short-time energy (STE) of each audio frame, the expression is:

[0037]

[0038] Where STE represents short-time energy, NN represents the frame length, and x[n] represents the nth sample in the audio frame.

[0039] As a further solution of the present invention: the fault diagnosis module uses a pre-trained deep learning model to identify the fault type of the high-voltage switch according to the feature vector, including:

[0040] Use convolutional neural networks (CNN) or recurrent neural networks (RNN) as deep learning models;

[0041] During model training, the cross entropy loss function and Adam optimizer are used;

[0042] The model training dataset includes audio feature vectors of normal operating conditions and various fault conditions.

[0043] As a further solution of the present invention: the result output module is used to display the diagnosis result, including:

[0044] Display fault type, fault location and severity through a graphical interface;

[0045] Provide written instructions and suggested repair measures;

[0046] Support voice broadcast function, which is convenient for on-site operators to use.

[0047] As a further solution of the present invention: the system also includes a multimodal data fusion module for integrating data from at least two different sensors to improve the accuracy of fault diagnosis.

[0048] As a further solution of the present invention: the system also includes an adaptive learning module, which can automatically adjust the sensitivity parameters of the model according to the environmental noise level to reduce false alarms.

[0049] A computer device comprises a memory and a processor, wherein the memory stores a computer program.

[0050] A computer-readable storage medium having a computer program stored thereon.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By using highly sensitive MEMS microphones and advanced audio processing technology, it is possible to accurately capture the faint abnormal sounds during the operation of high-voltage switches, improve the accuracy of fault diagnosis, and combine multiple sensor data, such as temperature sensors and vibration sensors, to improve the accuracy and reliability of fault diagnosis. The system can automatically adjust the sensitivity parameters of the model according to the ambient noise level, reduce false alarms, and improve the robustness of the system. It transmits audio data to the data processing center via wired or wireless means to achieve real-time monitoring and remote management, facilitate maintenance and troubleshooting, and provide intuitive fault diagnosis results and recommended maintenance measures through a graphical interface and voice broadcast function, which is convenient for on-site operators to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is the system module connection diagram of the high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, the specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0057] Example

[0058] See also Figure 1 , is an embodiment of the present invention, which provides a high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology, including:

[0059] Preparation of test subjects:

[0060] Select two high-voltage switchgears of the same model and mark them as A (prior art group) and B (invention group);

[0061] A MEMS microphone model SPH0645LM4H-B is installed near the incoming and outgoing terminals of each high-voltage switchgear and in the middle of the switch body. Its operating frequency range is 20Hz to 20kHz, and its signal-to-noise ratio is as high as 62dB;

[0062] Use a shock-proof mount to fix the microphone to the housing of the high-voltage switch to reduce the impact of external vibration on the audio signal;

[0063] The sampling rate was set to 44.1kHz, 16-bit quantization was selected to balance data quality and storage requirements, and the collected data was saved in WAV format;

[0064] Add a preamplifier at the microphone output, set the gain to 20dB, and add a low-pass filter before the ADC with a cutoff frequency of 20kHz.

[0065] The audio data is transmitted to the data processing center by wired means, and the received audio data is stored on the NAS platform in the data processing center;

[0066] Data collection:

[0067] Under normal operating conditions of the high-voltage switchgear, 1 hour of audio data was collected;

[0068] Under the simulated fault conditions of the high-voltage switchgear (such as partial discharge, poor contact, etc.), collect audio data for 1 hour;

[0069] The collected data includes audio signals of normal operating conditions and various fault conditions;

[0070] Data preprocessing:

[0071] Read audio data saved in WAV format from the NAS platform and use Python's scipy.io.wavfile library to read WAV files;

[0072] A convolutional neural network (CNN) is used as the deep learning model. During the model training process, the cross entropy loss function and the Adam optimizer are used.

[0073] The model training dataset includes audio feature vectors of normal operating conditions and various fault conditions.

[0074] The extracted feature vector is input into the trained deep learning model to identify the fault type of the high-voltage switch;

[0075] Result output:

[0076] Display fault type, fault location and severity through a graphical interface;

[0077] Provide written instructions and suggested repair measures;

[0078] Support voice broadcast function, which is convenient for on-site operators to use;

[0079]

[0080] In summary, the false alarm rate of the prior art group A is 15.2%, while the false alarm rate of the present invention group B is only 3.1%, which is reduced by 79.6%. This shows that the adaptive learning module of the present invention can automatically adjust the sensitivity parameters of the model according to the environmental noise level, effectively reducing false alarms. The fault detection accuracy of the prior art group A is 78.5%, while the fault detection accuracy of the present invention group B reaches 94.3%, which is an increase of 19.9%. This is mainly because the present invention adopts a multimodal data fusion module, combines a variety of sensor data, and improves the accuracy of fault diagnosis. The data processing time of the prior art group A is 12.4 seconds, while the data processing time of the present invention group B is only 7.8 seconds, which is a reduction of 37.1%. This is due to the edge computing unit of the present invention, which can perform preliminary data processing and fault judgment locally, reducing data transmission delay. The environmental noise impact of the prior art group A is 52.3dB, while the environmental noise impact of the present invention group B is only 38.7dB, which is a reduction of 25.9%. This shows that the spectral subtraction and bandpass filter of the present invention can more effectively remove environmental noise and improve the quality of audio signals.

[0081] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology, characterized by: include, At least one audio acquisition device, used to collect audio signals generated during the operation of the high-voltage switch; A data preprocessing module, used for preprocessing the audio signal; a feature extraction module for extracting feature vectors from the preprocessed audio signal; a fault diagnosis module, using a pre-trained deep learning model to identify a fault type of the high-voltage switch according to the feature vector; A result output module is used to display the diagnosis results.

2. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The at least one audio acquisition device, used to collect audio signals generated during the operation of the high-voltage switch, includes: Select the SPH0645LM4H-B MEMS microphone, which has an operating frequency range of 20Hz to 20kHz and a signal-to-noise ratio of up to 62dB; Install the microphone near the input and output terminals of the high-voltage switch and in the middle of the switch body, and use a shock-proof bracket to fix the microphone on the housing of the high-voltage switch; The sampling rate was set to 44.1kHz, 16-bit quantization was selected to balance data quality and storage requirements, and the collected data was saved in WAV format; Add a preamplifier at the microphone output, set the gain to 20dB, and add a low-pass filter before the ADC with a cutoff frequency of 20kHz. The audio data is transmitted to a data processing center using wired or wireless means, and at the data processing center, the received audio data is stored on a NAS or cloud storage platform.

3. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The data preprocessing module is used to preprocess the audio signal, including: Read audio data saved in WAV format from NAS or cloud storage platforms, and use Python's scipy.io.wavfile library to read WAV files; Use spectral subtraction to reduce the noise of the audio signal. The expression is: y(n)=x(n)-α·N(n); Where y(n) represents the signal after denoising, xn) represents the original signal, Nn) represents the estimated noise signal, and α represents the denoising coefficient; Use a bandpass filter to filter the audio signal to remove high-frequency and low-frequency noise. The expression is: y[n]=b0x[n]+b1x[n-1]+b2x[n-2]-a1y[n-1]-a2y[n-2]; y[n] represents the filtered signal, x[n] represents the input signal, b0, b1, b2 and a1, a2 are the filter coefficients; Normalize the audio signal to the range of [-1,1] to eliminate amplitude differences. The expression is: y[n] represents the normalized signal, and x[n] represents the original signal; The audio signal is divided into multiple short-time frames, with the frame length set to 20ms and the frame shift set to 10ms.

4. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 2 is characterized in that: The feature extraction module is used to extract feature vectors from the preprocessed audio signal. include, Calculate the Mel-frequency cepstral coefficient (MFCC) of each audio frame, the expression is: Where X[k] represents the frequency domain signal, x[n] represents the time domain signal, NN represents the frame length, m represents the Mel frequency, f represents the linear frequency, Em represents the logarithmic energy of the mth Mel filter, Hmk) represents the frequency response of the mth Mel filter, cn represents the nth MFCC coefficient, and M represents the number of Mel filters; Calculate the zero crossing rate (ZCR) of each audio frame, the expression is: Where ZCR represents the zero crossing rate, NN represents the frame length, and 1{x[n]x[n-1]<0} represents the indicator function; Calculate the short-time energy (STE) of each audio frame, the expression is: Where STE represents short-time energy, NN represents the frame length, and x[n] represents the nth sample in the audio frame.

5. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The fault diagnosis module uses a pre-trained deep learning model to identify the fault type of the high-voltage switch according to the feature vector, including: Use convolutional neural networks (CNN) or recurrent neural networks (RNN) as deep learning models; During model training, the cross entropy loss function and Adam optimizer are used; The model training dataset includes audio feature vectors of normal operating conditions and various fault conditions.

6. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The result output module is used to display the diagnosis results. include, Display fault type, fault location and severity through a graphical interface; Provide written instructions and suggested repair measures; Support voice broadcast function, which is convenient for on-site operators to use.

7. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The system also includes a multimodal data fusion module for integrating data from at least two different sensors to improve the accuracy of fault diagnosis.

8. The high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology according to claim 1 is characterized in that: The system also includes an adaptive learning module that can automatically adjust the model's sensitivity parameters based on the ambient noise level to reduce false alarms.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the high-voltage switch fault diagnosis system based on artificial intelligence audio recognition technology described in any one of claims 1 to 7 are implemented.

Images