Voiceprint-based converter valve saturable reactor screw looseness detection system and method

Abstract

The invention provides a voiceprint-based converter valve saturable reactor screw looseness detection system and method, and the method comprises the steps: collecting a sound wave signal during the operation of a saturable reactor based on a sound wave signal collection module, and carrying out the preprocessing of the collected sound wave signal based on a preprocessing module, the method comprises the following steps: preprocessing a sound wave signal, carrying out time-frequency analysis on the preprocessed sound wave signal based on a wavelet transformation module to obtain time-frequency domain analysis data, converting the time-frequency domain analysis data into a gray fusion image based on a gray fusion image generation module, building a ResNet model based on a ResNet model module, and training the ResNet model based on the obtained gray fusion image to obtain a trained sound wave signal; and inputting a gray fusion image of a sound wave signal to be detected into the trained ResNet model to realize looseness detection. The device can accurately capture the symptom of screw loosening at an early stage, improves the detection sensitivity and reliability, can adapt to a complex operation environment, and is convenient to use.

Description

Technical Field

[0001] The present invention relates to the technical field of high-voltage direct current power transmission, and in particular to a system and method for detecting screw loosening of a saturated reactor of a converter valve based on voiceprint. Background Art

[0002] With the rapid development of high-voltage direct current transmission technology, the importance of stable operation of converter valves as its core equipment has become increasingly prominent. Among them, saturated reactors, as key components in converter valves, play a vital role. The technical means currently used to monitor the health status of converter valve equipment on the market mainly include traditional vibration monitoring and electrical parameter detection. However, these methods have obvious shortcomings and limitations when detecting minor faults in the mechanical structure of the equipment (such as loose screws). The current mainstream means of mechanical fault monitoring is to capture the vibration signal of the equipment during operation through a vibration sensor. However, the collection of vibration signals has low sensitivity for detecting minor faults such as loose screws, especially during the operation of saturated reactors, where the vibration amplitude is small, and conventional vibration monitoring is difficult to capture the characteristic signals of early looseness. In addition, the vibration sensor is complex to install and is easily interfered by external environmental noise, resulting in reduced detection accuracy and increased false alarm rate. Another common monitoring method is to indirectly judge mechanical faults by detecting changes in the electrical parameters of the equipment. However, loose screws often do not have a significant impact on electrical parameters in the early stage, which makes the electrical parameter detection method ineffective in early fault identification, and usually only has a significant response when the fault intensifies. This lag makes equipment maintenance lack of predictability and can easily cause more serious equipment damage.

[0003] In recent years, voiceprint monitoring technology has received widespread attention and rapid development. With the advancement of signal processing algorithms and artificial intelligence technology, the application potential of voiceprint monitoring in the field of equipment fault diagnosis has gradually emerged. Compared with traditional vibration monitoring and electrical parameter monitoring technologies, voiceprint monitoring technology has many significant advantages, especially in complex equipment status monitoring, showing high sensitivity and adaptability. At the same time, the cost of voiceprint monitoring is relatively low and the requirements for use scenarios are relatively small. In addition, voiceprint monitoring has good anti-electromagnetic interference characteristics and will not be affected by external electromagnetic signals, ensuring the stability and accuracy of data acquisition. At present, voiceprint monitoring technology has gradually become one of the important technologies for intelligent equipment status monitoring due to its high sensitivity, low cost and wide adaptability. Therefore, it is very necessary to design a system and method for detecting screw loosening of converter valve saturated reactor based on voiceprint. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a system and method for detecting screw loosening of a saturated reactor of a converter valve based on voiceprint.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a screw loosening detection system for a saturated reactor of a converter valve based on voiceprint, comprising: an acoustic wave signal acquisition module, a preprocessing module, a wavelet transform module, a grayscale fusion image generation module and a ResNet model module. The acoustic wave signal acquisition module is arranged on the saturated reactor, the acoustic wave signal acquisition module is connected to the preprocessing module, the preprocessing module is connected to the wavelet transform module, the wavelet transform module is connected to the grayscale fusion image generation module, and the grayscale fusion image generation module is connected to the ResNet model module.

[0007] Preferably, the sound wave signal acquisition module includes a high-sensitivity microphone array, a signal amplification circuit, an analog-to-digital converter, an embedded processor and a communication module. The high-sensitivity microphone array is arranged on the saturated inductor, the high-sensitivity microphone array is connected to the signal amplification circuit, the signal amplification circuit is connected to the analog-to-digital converter, the analog-to-digital converter is connected to the embedded processor, and the embedded processor is connected to the preprocessing module via the communication module.

[0008] Preferably, the high-sensitivity microphone array is fixedly arranged on the surface of the saturable reactor via a vibration isolation bracket.

[0009] The present invention also provides a method for detecting screw loosening of a converter valve saturated reactor based on voiceprint, comprising:

[0010] Step 100: collecting the acoustic wave signal of the saturated reactor in operation based on the acoustic wave signal collection module;

[0011] Step 200: preprocessing the collected sound wave signal based on the preprocessing module;

[0012] Step 300: performing time-frequency analysis on the preprocessed sound wave signal based on a wavelet transform module to obtain time-frequency domain analysis data;

[0013] Step 400: converting the time-frequency domain analysis data into a grayscale fusion image based on a grayscale fusion image generation module;

[0014] Step 500: Building a ResNet model based on the ResNet model module, and training the ResNet model based on the obtained grayscale fusion image;

[0015] Step 600: Input the grayscale fusion image of the acoustic wave signal to be detected into the trained ResNet model to achieve looseness detection.

[0016] Preferably, in step 200, the collected sound wave signal is preprocessed based on the preprocessing module, specifically:

[0017] The high-frequency noise of the collected sound wave signal is removed by a low-pass filter.

[0018] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0019] The present invention provides a saturated reactor screw loosening detection system and method based on voiceprint, the method comprises collecting the acoustic wave signal in the operation of the saturated reactor based on the acoustic wave signal acquisition module, preprocessing the collected acoustic wave signal based on the preprocessing module, performing time-frequency analysis on the preprocessed acoustic wave signal based on the wavelet transform module, obtaining time-frequency domain analysis data, converting the time-frequency domain analysis data into a grayscale fusion image based on the grayscale fusion image generation module, building a ResNet model based on the ResNet model module, and training the ResNet model based on the obtained grayscale fusion image, inputting the grayscale fusion image of the acoustic wave signal to be detected into the trained ResNet model, and realizing loosening detection. The present invention has high sensitivity, and can detect extremely small mechanical anomalies through the combination of wavelet transform and deep learning model; high real-time performance, can monitor online, collect and analyze data in real time, and provide a rapid alarm function; high robustness, can still work normally in a complex electromagnetic environment, and ensure the accuracy of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 A system structure diagram provided for an embodiment of the present invention;

[0022] Figure 2 A flow chart of a method provided by an embodiment of the present invention;

[0023] Figure 3a This is the grayscale fusion image when the screw is not loose;

[0024] Figure 3b This is the grayscale fused image when the screw is loose. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] The purpose of the present invention is to provide a system and method for detecting loose screws of a saturated reactor of a converter valve based on voiceprint, which can accurately capture the signs of screw loosening in the early stage, improve the sensitivity and reliability of detection, and at the same time enhance the anti-interference ability of the system, reduce false alarms, adapt to complex operating environments, and realize online monitoring of loose screws of a saturated reactor of a converter valve by voiceprint.

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1 As shown, the present invention provides a screw loosening detection system for a saturated reactor of a converter valve based on voiceprint, comprising: an acoustic wave signal acquisition module, a preprocessing module, a wavelet transform module, a grayscale fusion image generation module and a ResNet model module. The acoustic wave signal acquisition module is arranged on the saturated reactor, the acoustic wave signal acquisition module is connected to the preprocessing module, the preprocessing module is connected to the wavelet transform module, the wavelet transform module is connected to the grayscale fusion image generation module, and the grayscale fusion image generation module is connected to the ResNet model module.

[0029] The sound wave signal acquisition module includes a high-sensitivity microphone array, a signal amplification circuit, an analog-to-digital converter, an embedded processor and a communication module. The high-sensitivity microphone array is arranged on the saturated inductor, the high-sensitivity microphone array is connected to the signal amplification circuit, the signal amplification circuit is connected to the analog-to-digital converter, the analog-to-digital converter is connected to the embedded processor, and the embedded processor is connected to the preprocessing module via the communication module.

[0030] The high-sensitivity microphone array is fixedly arranged on the surface of the saturated reactor through a vibration isolation bracket, which can avoid interference of environmental noise and mechanical vibration on the microphone.

[0031] Multiple high-sensitivity microphones are distributed on the surface of the saturated reactor housing to collect real-time acoustic signals generated by the saturated reactor during operation. These microphones can capture tiny sound changes caused by mechanical vibrations (such as loose screws).

[0032] Signal amplification circuit: The collected sound signal may be weak, so it is first amplified by the amplifier to ensure the stability of subsequent signal processing. The amplifier uses a low-noise operational amplifier to avoid introducing too much noise;

[0033] Analog-to-digital converter (ADC): The amplified analog sound signal is converted into a digital signal through an analog-to-digital converter (ADC) for subsequent digital processing. The sampling rate of the ADC is set to 44.1kHz to meet the needs of collecting sound signals within the hearing range of the human ear, while ensuring sufficient resolution to capture the weak signals generated by loose screws;

[0034] Embedded processor: The processor uses a high-performance ARM Cortex-M4 or equivalent chip with powerful digital signal processing (DSP) capabilities for real-time processing of sound signals. The processor is connected to multiple sensor channels and can simultaneously collect and process signals from different microphones.

[0035] Communication module: Transmits processed data to a host computer or remote server for storage, analysis and alarm via wired or wireless communication modules (such as Ethernet or Wi-Fi).

[0036] like Figure 2 As shown, the present invention also provides a method for detecting screw loosening of a saturated reactor of a converter valve based on voiceprint, comprising:

[0037] Step 100: collecting the acoustic wave signal of the saturated reactor in operation based on the acoustic wave signal collection module;

[0038] Step 200: preprocessing the collected sound wave signal based on the preprocessing module;

[0039] Step 300: performing time-frequency analysis on the preprocessed sound wave signal based on a wavelet transform module to obtain time-frequency domain analysis data;

[0040] Step 400: converting the time-frequency domain analysis data into a grayscale fusion image based on a grayscale fusion image generation module;

[0041] Step 500: Building a ResNet model based on the ResNet model module, and training the ResNet model based on the obtained grayscale fusion image;

[0042] Step 600: Input the grayscale fusion image of the acoustic wave signal to be detected into the trained ResNet model to achieve looseness detection.

[0043] In step 100, the acoustic wave signal in the operation of the saturated reactor is collected based on the acoustic wave signal collection module, specifically:

[0044] The acoustic wave signal acquisition module collects the acoustic wave signals of the reactor in operation in real time through the microphone array distributed on the shell of the saturated reactor; it should be noted that the acoustic waves caused by mechanical vibrations, especially the vibrations caused by loose screws, are transmitted to the microphone through the air to form specific soundprint signals.

[0045] In step 200, the collected sound wave signal is preprocessed based on the preprocessing module, specifically:

[0046] The collected sound signal is filtered through a low-pass filter to remove high-frequency noise (the filter cutoff frequency is set to 10kHz) to eliminate high-frequency interference in the environment.

[0047] In step 300, the pre-processed sound wave signal is subjected to time-frequency analysis based on the wavelet transform module to obtain time-frequency domain analysis data, specifically:

[0048] The wavelet transform module performs time-frequency analysis on the preprocessed acoustic signal, and uses Daubechies (db4) as the wavelet basis function, which can effectively decompose the different frequency bands of the acoustic signal. The wavelet transform can extract the short-term mutation features in the signal, which are often associated with abnormalities in the mechanical structure (such as loose screws).

[0049] In step 400, the time-frequency domain analysis data is converted into a grayscale fusion image based on a grayscale fusion image generation module, specifically:

[0050] The time-frequency domain data after wavelet transformation is converted into a two-dimensional grayscale image. Each pixel represents the energy intensity of a certain frequency band at a specific moment. The data of different microphones are fused to generate multiple grayscale images. These images reflect the voiceprint characteristics of the reactor under the operating state;

[0051] The present invention provides two schematic diagrams, wherein: Figure 3a This is the grayscale fusion image when the screw is not loose. Figure 3b This is the grayscale fused image when the screw is loose.

[0052] In step 500, a ResNet model is built based on the ResNet model module, and the ResNet model is trained based on the obtained grayscale fusion image, specifically:

[0053] Using a large number of collected normal and abnormal data sets, the grayscale images are input into the ResNet deep learning model for training. The input layer of the model is a multi-channel convolutional layer that can process grayscale images from multiple microphones in parallel;

[0054] The training process includes the following parameter settings:

[0055] Learning rate: The initial learning rate is set to 0.001 and gradually decreases during training to increase the model convergence speed;

[0056] Batch size: set to 32 to improve computational efficiency while ensuring training stability;

[0057] Loss function: The cross entropy loss function is used to measure the difference between the predicted results and the actual annotations.

[0058] In step 600, the grayscale fusion image of the acoustic wave signal to be detected is input into the trained ResNet model to implement looseness detection, specifically:

[0059] After the model training is completed, it enters the real-time monitoring mode, transforms the newly collected sound signal through wavelet transform and generates a grayscale image, which is input into the trained ResNet model in real time;

[0060] The ResNet model predicts the current state of the reactor based on the input grayscale image and determines whether there is any loose screw phenomenon; if the model detects the signal characteristics of loose screws, the system immediately sends an alarm signal and sends the alarm information and related data to the host computer or operation and maintenance personnel through the communication module.

[0061] The present invention provides an embodiment to define the process conditions of the system described in the present invention:

[0062] 1. Environmental requirements: The system is suitable for various operating environments, including high temperature, high humidity and high electromagnetic interference places. To ensure the stability of the system, the microphone and processor need to have a protection level (such as IP67).

[0063] 2. Signal processing time: The processing time of each sound signal segment is controlled within 10 milliseconds to ensure real-time performance.

[0064] 3. Detection sensitivity: The system can detect screw loosening below 0.1mm, with a frequency range of 20Hz to 10kHz.

[0065] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0066] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A converter valve saturated reactor screw loosening detection system based on voiceprint, characterized in that: include: An acoustic wave signal acquisition module, a preprocessing module, a wavelet transform module, a grayscale fusion image generation module and a ResNet model module. The acoustic wave signal acquisition module is arranged on the saturated reactor, the acoustic wave signal acquisition module is connected to the preprocessing module, the preprocessing module is connected to the wavelet transform module, the wavelet transform module is connected to the grayscale fusion image generation module, and the grayscale fusion image generation module is connected to the ResNet model module.

2. The system according to claim 1, characterized in that The sound wave signal acquisition module includes a high-sensitivity microphone array, a signal amplification circuit, an analog-to-digital converter, an embedded processor and a communication module. The high-sensitivity microphone array is arranged on the saturated inductor, the high-sensitivity microphone array is connected to the signal amplification circuit, the signal amplification circuit is connected to the analog-to-digital converter, the analog-to-digital converter is connected to the embedded processor, and the embedded processor is connected to the preprocessing module via the communication module.

3. The system according to claim 2, characterized in that The high-sensitivity microphone array is fixedly arranged on the surface of the saturated reactor through a vibration isolation bracket.

4. A method for detecting screw loosening of a converter valve saturated reactor based on voiceprint, characterized in that: include: Step 100: collecting the acoustic wave signal of the saturated reactor in operation based on the acoustic wave signal collection module; Step 200: preprocessing the collected sound wave signal based on the preprocessing module; Step 300: performing time-frequency analysis on the preprocessed sound wave signal based on a wavelet transform module to obtain time-frequency domain analysis data; Step 400: converting the time-frequency domain analysis data into a grayscale fusion image based on a grayscale fusion image generation module; Step 500: Building a ResNet model based on the ResNet model module, and training the ResNet model based on the obtained grayscale fusion image; Step 600: Input the grayscale fusion image of the acoustic wave signal to be detected into the trained ResNet model to achieve looseness detection.

5. The method according to claim 4, characterized in that In step 200, the collected sound wave signal is preprocessed based on the preprocessing module, specifically: The high-frequency noise of the collected sound wave signal is removed by a low-pass filter.

Images