Communication tower bolt loosening positioning system and method based on voiceprint feature extraction

Abstract

The invention discloses a communication tower bolt loosening positioning system and method based on voiceprint feature extraction, belongs to the field of iron tower bolt monitoring, and relates to the voiceprint feature extraction technology, the condition of a communication tower is analyzed through the voiceprint feature extraction technology, excitation sound signals with a certain frequency are applied to the communication tower, and the excitation sound signals are collected through a bone conduction voiceprint sensor; constructing a communication tower voiceprint sample library to form a communication tower bolt loosening knowledge graph library; the voiceprint recognition module loads the target monitoring audio into a bolt looseness classification model; performing classification processing based on a bolt looseness classification model to obtain a classification state result of the target monitoring audio; by means of the mode, the safety and reliability of the communication tower are improved, accidents and equipment damage caused by bolt looseness or breakage are avoided, and therefore casualties and maintenance cost are reduced.

Description

Technical Field

[0001] The present invention belongs to the field of tower bolt monitoring and relates to voiceprint feature extraction technology, and specifically to a communication tower bolt loosening positioning system and method based on voiceprint feature extraction. Background Art

[0002] Steel towers are an important component of transmission lines. As the support of overhead wires, their bearing capacity plays an important role in the normal operation of transmission lines.

[0003] Since communication towers are in a complex environment with various static and dynamic loads for a long time, the tower nodes will suffer varying degrees of damage. The most intuitive manifestation of tower node damage is loose bolts. Usually, when bolts loosen at the tower nodes, the slip load borne by the bolt group will decrease, which will reduce the axial stiffness of the rods and increase the deformation of the rods, resulting in a decrease in the bearing capacity of the communication tower. If the node damage continues to develop, the communication tower may eventually collapse under strong winds. Due to the large number of bolts and complex structure of communication towers, there has been no efficient method for detecting the tightening of communication tower bolts.

[0004] At present, the tightness of the bolts at the bottom of the tower is roughly determined by knocking and listening to the sound, or high-altitude workers climb the tower to conduct sampling inspections to roughly judge the tightness of the bolts of the entire tower. The detection accuracy and work efficiency are extremely low, and the bolt status cannot be monitored in real time, which cannot meet the increasingly increasing requirements for lean operation and inspection of equipment.

[0005] Therefore, the present invention proposes a communication tower bolt loosening positioning system and method based on voiceprint feature extraction. Summary of the invention

[0006] The purpose of this application is to provide a communication tower bolt loosening positioning system and method based on voiceprint feature extraction, which is used to solve the current problem of relying on knocking to identify the tightness of the bolts at the bottom of the tower by listening to the sound, or high-altitude workers climbing the tower to conduct sampling inspection to roughly judge the tightness of the bolts of the entire tower. The detection accuracy and work efficiency are extremely low, and the bolt status cannot be monitored in real time.

[0007] To achieve the above objectives, the present application provides a tower bolt loosening positioning system based on voiceprint feature extraction, including:

[0008] A sound wave generator and a plurality of bone conduction voiceprint sensors are fixedly mounted on the communication tower. The sound wave generator is used to apply excitation sound wave signals of different frequencies to the communication tower, and the bone conduction voiceprint sensors are used to obtain audio information.

[0009] and an intelligent analysis terminal for obtaining audio information sent by the bone conduction voiceprint sensor; the intelligent analysis terminal is used to pre-process the audio information to obtain target monitoring audio;

[0010] A voiceprint recognition module, which is used to perform voiceprint recognition and classification processing on the target monitoring audio, includes:

[0011] The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model;

[0012] and performing classification processing based on the bolt loosening classification model to obtain a classification status result of the target monitoring audio;

[0013] The classification status results include: normal status, slightly loose bolt status, and heavily loose bolt status.

[0014] Preferably, the frequency range of the excitation sound wave signal is 20HZ-20KHZ, and the excitation sound wave signal propagates along different propagation paths along the structure of the communication tower, and the propagation path is determined according to the specific structure of the communication tower.

[0015] Preferably, a plurality of the bone conduction voiceprint sensors are arranged at a plurality of designated positions of the communication tower, and the designated positions of the communication tower are the monitoring point positions of a plurality of bolts that need to be detected, and are set by the operation and maintenance personnel of the transmission line.

[0016] Preferably, it also includes a control module; the control module is used to control the sound wave generator to emit an excitation sound wave signal.

[0017] Preferably, the process of preprocessing the audio information by the intelligent analysis terminal includes:

[0018] Obtain audio information from several monitoring points in real time;

[0019] The pre-trained environment denoising model is called to denoise the audio information of multiple monitoring points to obtain multiple target monitoring audios.

[0020] Preferably, the training process of the environment denoising model includes:

[0021] The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower;

[0022] The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors and the collection time is marked;

[0023] The algorithm model is used to separate the characteristic signal of each acquisition time, and distinguish the voiceprint signal and the environmental noise signal transmitted by the excitation sound wave signal through the communication tower;

[0024] The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environmental denoising model, and the validation set is used to verify the trained environmental denoising model.

[0025] Preferably, the training process of the bolt loosening classification model includes:

[0026] Classify historical target monitoring audio into normal state, state with few bolts loosened, and state with many bolts loosened;

[0027] Capture voiceprint fragments through audio processing software;

[0028] Loading the extracted voiceprint fragments into a preset classification model for voiceprint classification processing;

[0029] The captured voiceprint segment is input into a preset classification model, which identifies the type of the voiceprint segment and outputs a classification status result.

[0030] Preferably, the preset classification model is a deep learning network, specifically a convolutional neural network (CNN) or a recurrent neural network (RNN).

[0031] Preferably, the control module is also connected to a status warning module. When the classification status result identified by the voiceprint recognition module is a state of a small number of loose bolts or a state of a large number of loose bolts, the control module outputs a first warning signal or a second warning signal to the status warning module, and the status warning module notifies the operation and maintenance personnel according to the corresponding first warning signal or second warning signal.

[0032] The method for locating loose bolts of a tower based on voiceprint feature extraction includes the following steps:

[0033] The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower;

[0034] The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors and the collection time is marked;

[0035] The algorithm model is used to separate the characteristic signal of each acquisition time, and distinguish the voiceprint signal and the environmental noise signal transmitted by the excitation sound wave signal through the communication tower;

[0036] The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environmental denoising model;

[0037] The bone conduction voiceprint sensor acquires audio information in real time and sends it to the intelligent analysis terminal, which pre-processes the audio information to obtain the target monitoring audio and feeds it back to the voiceprint recognition module;

[0038] The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model;

[0039] And classification processing is performed based on the bolt loosening classification model to obtain the classification status result of the target monitoring audio.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention analyzes the status of the communication tower through voiceprint feature extraction technology, applies an excitation sound signal of a certain frequency to the communication tower, and uses a bone conduction voiceprint sensor to collect the excitation sound signal, which is the sound signal distorted after being transmitted through the tower body; the abnormal sound of the communication tower is collected and stored in the cloud engine server, and a communication tower voiceprint sample library is constructed, and a knowledge graph library of loose bolts of the communication tower is formed based on the communication tower voiceprint sample library; the voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model; and performs classification processing based on the bolt loosening classification model to obtain the classification status result of the target monitoring audio; this method improves the safety and reliability of the communication tower, avoids accidents and equipment damage caused by loose or broken bolts, and thus reduces casualties and maintenance costs. Secondly, by real-time monitoring and evaluation of the tightening status of the bolts, problems can be discovered and repaired in a timely manner, reducing maintenance costs.

[0042] 2. The present invention uses an environmental denoising model to denoise the audio information of the monitoring point; the control module sets the sound wave generator to sound regularly and emits excitation sound wave signals of different frequencies; the audio information of different monitoring points at different time nodes is obtained through a number of bone conduction soundprint sensors and the collection time is marked; the characteristic signal of each collection time is separated through the algorithm model, and the soundprint signal and the environmental noise signal of the excitation sound wave signal transmitted through the communication tower are distinguished; the acquired soundprint signals of each collection time are divided into a training set and a verification set according to a ratio of 7:3; the training set is used to train the environmental denoising model, and the trained environmental denoising model is verified by the verification set to ensure the accuracy of the environmental denoising model;

[0043] 3. The present invention is provided with a monitoring interaction module including a monitoring interaction interface; the monitoring interaction module is directly connected to the cloud engine server, and is used to display the audio information including frequency, vibration amplitude, and sound wave change data acquired in real time on the monitoring interaction interface, and establish a relationship waveform diagram between time and frequency and vibration amplitude for the acquired frequency and vibration amplitude, analyze the historical development and future trend of the fault from the relationship waveform diagram, and play a role in fault prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the prior art and the drawings required for use in the embodiments are 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 creative work.

[0045] Figure 1 It is a structural block diagram of the tower bolt loosening positioning system based on voiceprint feature extraction of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] Please refer to Figure 1 ,An embodiment of the present invention provides a tower bolt loosening location system based on voiceprint feature extraction, the system includes a sound wave generator;

[0048] The sound wave generator is used to apply excitation sound wave signals of different frequencies to the communication tower; the frequency range of the excitation sound wave signal is 20HZ-20KHZ; the sound wave generator adopts an adaptive active sound generation system, and the sound generation content is automatically analyzed and adapted according to the tower model, size, environment and other conditions, greatly optimizing the efficiency of voiceprint analysis;

[0049] The excitation sound wave signal will propagate along different propagation paths along the structure of the communication tower, wherein the propagation path is determined according to the specific structure of the communication tower;

[0050] and a plurality of bone conduction voiceprint sensors, wherein the plurality of bone conduction voiceprint sensors are arranged at a plurality of designated locations of the communication tower for acquiring audio information;

[0051] The audio information includes frequency, vibration amplitude, sound wave change data, etc.

[0052] In this application, the bone conduction voiceprint sensor is close to the communication tower, has a lower acoustic impedance, shields external noise, detects tower failures, and has higher sensitivity; the bone conduction voiceprint sensor has a built-in DSP voiceprint noise reduction algorithm and a unique sound layering technology, which can decompose the overall noise of multiple devices into the noise of each device. Sampling frequency>=48Khz, sampling accuracy 16bit, audio frequency 20-20khz;

[0053] It should be noted that the designated location of the communication tower is the location of the monitoring points of multiple bolts that need to be inspected, which is set by the operation and maintenance personnel of the transmission line;

[0054] In the present application, several of the bone conduction voiceprint sensors are connected to an intelligent analysis terminal, which is used to obtain the audio information sent by the bone conduction voiceprint sensor and pre-process the received audio information to obtain the target monitoring audio;

[0055] It should be noted that the device further includes a control module; the control module is used to control the sound wave generator to emit an excitation sound wave signal;

[0056] Specifically, the process of preprocessing the audio information by the intelligent analysis terminal includes:

[0057] Obtain audio information from several monitoring points in real time;

[0058] Call the pre-trained environment denoising model to denoise the audio information of multiple monitoring points to obtain multiple target monitoring audios;

[0059] In an embodiment of the present invention, the training process of the environment denoising model includes:

[0060] The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower;

[0061] The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors and the collection time is marked;

[0062] The algorithm model is used to separate the characteristic signal of each acquisition time, and distinguish the voiceprint signal and the environmental noise signal transmitted by the excitation sound wave signal through the communication tower;

[0063] The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environment denoising model, and the validation set is used to verify the trained environment denoising model to ensure the accuracy of the environment denoising model;

[0064] Among them, after the audio information of the monitoring point is denoised, the target monitoring audio is a voiceprint signal without noise signal;

[0065] In the present application, the intelligent analysis terminal is used to send the pre-processed target monitoring audio to the cloud engine server for storage;

[0066] It should be noted that the cloud engine server is connected to a voiceprint recognition module, and the voiceprint recognition module is remotely connected to the cloud engine server;

[0067] The voiceprint recognition module is used to perform voiceprint recognition and classification processing on the target monitoring audio received in real time by the cloud engine server, wherein the specific process includes:

[0068] The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model;

[0069] and performing classification processing based on the bolt loosening classification model to obtain a classification status result of the target monitoring audio;

[0070] The classification status results include: normal status, status of a small number of bolts loosened, and status of a large number of bolts loosened;

[0071] In one embodiment of the present application, the cloud engine server is constructed with a communication tower voiceprint sample library, and a knowledge graph library of loose bolts of communication towers is formed based on the communication tower voiceprint sample library;

[0072] The training process of the bolt loosening classification model includes:

[0073] Classify historical target monitoring audio into normal state, state with few bolts loosened, and state with many bolts loosened;

[0074] The soundprint fragments are intercepted by audio processing software, for example, the sound wave fragments with specific frequency and amplitude generated when the bolts are loosened are intercepted from a long period of audio;

[0075] Loading the extracted voiceprint fragments into a preset classification model for voiceprint classification processing;

[0076] The preset classification model may be a deep learning network, such as a convolutional neural network (CNN) or a recurrent neural network (RNN);

[0077] It can be known that when the intercepted voiceprint segment is input into the preset CNN model, the model will try to identify which type of voiceprint segment this is (such as a small number of loose bolts, a large number of loose bolts, etc.) and output a classification status result.

[0078] In the present application, the control module is also connected to a status warning module. When the classification status result identified by the voiceprint recognition module is a state of a small number of loose bolts or a state of a large number of loose bolts, the control module outputs a first warning signal or a second warning signal to the status warning module. The status warning module notifies the operation and maintenance personnel according to the corresponding first warning signal or second warning signal. The operation and maintenance personnel perform maintenance on the communication tower to ensure the safe operation of the tower and reduce economic losses and safety accidents.

[0079] In one embodiment of the present application, a monitoring interaction module is also included, and the monitoring interaction module includes a monitoring interaction interface; the monitoring interaction module is directly connected to the cloud engine server, and is used to display the audio information including frequency, vibration amplitude, and sound wave change data acquired in real time on the monitoring interaction interface, and establish a relationship waveform diagram between time and frequency and vibration amplitude for the acquired frequency and vibration amplitude, analyze the historical development and future trend of the fault from the relationship waveform diagram, and play a role in fault prevention.

[0080] Another embodiment of the present invention provides a method for locating loose bolts of a tower based on voiceprint feature extraction, the method comprising the following steps:

[0081] The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower;

[0082] The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors, and the collection time is marked; the characteristic signal of each collection time is separated through the algorithm model, and the voiceprint signal of the excitation sound wave signal transmitted through the communication tower and the environmental noise signal are distinguished;

[0083] The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environmental denoising model; the audio information is acquired in real time through the bone conduction voiceprint sensor and sent to the intelligent analysis terminal, which pre-processes the audio information to obtain the target monitoring audio and feeds it back to the voiceprint recognition module;

[0084] The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model; and performs classification processing based on the bolt loosening classification model to obtain a classification status result of the target monitoring audio;

[0085] When the classification status result identified by the voiceprint recognition module is a state of a small number of loose bolts or a state of a large number of loose bolts, the control module outputs a first warning signal or a second warning signal to the status warning module, and the status warning module notifies the operation and maintenance personnel according to the corresponding first warning signal or second warning signal. The operation and maintenance personnel perform maintenance on the communication tower to ensure the safe operation of the tower and reduce economic losses and safety accidents.

[0086] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The tower bolt loosening positioning system based on voiceprint feature extraction is characterized by: include: A sound wave generator and a plurality of bone conduction voiceprint sensors are fixedly mounted on the communication tower. The sound wave generator is used to apply excitation sound wave signals of different frequencies to the communication tower, and the bone conduction voiceprint sensors are used to obtain audio information. and an intelligent analysis terminal for obtaining audio information sent by the bone conduction voiceprint sensor; the intelligent analysis terminal is used to pre-process the audio information to obtain target monitoring audio; The voiceprint recognition module is used to perform voiceprint recognition and classification processing on the target monitoring audio, including: The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model; and performing classification processing based on the bolt loosening classification model to obtain a classification status result of the target monitoring audio; The classification status results include: normal status, slightly loose bolt status, and heavily loose bolt status.

2. The tower bolt loosening positioning system based on voiceprint feature extraction according to claim 1 is characterized in that: The frequency range of the excitation sound wave signal is 20HZ-20KHZ. The excitation sound wave signal propagates along different propagation paths along the structure of the communication tower. The propagation path is determined according to the specific structure of the communication tower.

3. The tower bolt loosening positioning system based on voiceprint feature extraction according to claim 1 is characterized in that: Several of the bone conduction voiceprint sensors are arranged at multiple designated positions of the communication tower. The designated positions of the communication tower are monitoring point positions of multiple bolts that need to be detected and are set by the operation and maintenance personnel of the transmission line.

4. The tower bolt loosening positioning system based on voiceprint feature extraction according to claim 1 is characterized in that: It also includes a control module; the control module is used to control the sound wave generator to send out an excitation sound wave signal.

5. The tower bolt loosening positioning system based on voiceprint feature extraction according to claim 1 is characterized in that: The process of preprocessing the audio information by the intelligent analysis terminal includes: Obtain audio information from several monitoring points in real time; The pre-trained environment denoising model is called to denoise the audio information of multiple monitoring points to obtain multiple target monitoring audios.

6. The tower bolt loosening location system based on voiceprint feature extraction according to claim 5 is characterized in that: The training process of the environment denoising model includes: The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower; The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors and the collection time is marked; The algorithm model is used to separate the characteristic signal of each acquisition time, and distinguish the voiceprint signal and the environmental noise signal transmitted by the excitation sound wave signal through the communication tower; The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environmental denoising model, and the validation set is used to verify the trained environmental denoising model.

7. The tower bolt loosening location system based on voiceprint feature extraction according to claim 1 is characterized in that: The training process of the bolt loosening classification model includes: Classify historical target monitoring audio into normal state, state with few bolts loosened, and state with many bolts loosened; Capture voiceprint fragments through audio processing software; Loading the extracted voiceprint fragments into a preset classification model for voiceprint classification processing; The captured voiceprint segment is input into a preset classification model, which identifies the type of the voiceprint segment and outputs a classification status result.

8. The tower bolt loosening location system based on voiceprint feature extraction according to claim 7 is characterized in that: The preset classification model is a deep learning network, specifically a convolutional neural network or a recurrent neural network.

9. The tower bolt loosening location system based on voiceprint feature extraction according to claim 1 is characterized in that: The control module is also connected to a status warning module. When the classification status result identified by the voiceprint recognition module is a state of a small number of loose bolts or a state of a large number of loose bolts, the control module outputs a first warning signal or a second warning signal to the status warning module, and the status warning module notifies the operation and maintenance personnel according to the corresponding first warning signal or second warning signal.

10. The tower bolt loosening positioning method of the tower bolt loosening positioning system based on voiceprint feature extraction according to any one of claims 1 to 9, characterized in that: The following steps are involved: The control module sets the sound wave generator to emit sound at a fixed time and emits excitation sound wave signals of different frequencies, and the excitation sound wave signals of different frequencies propagate along different propagation paths along the structure of the communication tower; The audio information of different monitoring points and different time nodes is obtained through a number of bone conduction voiceprint sensors and the collection time is marked; The algorithm model is used to separate the characteristic signal of each acquisition time, and distinguish the voiceprint signal and the environmental noise signal transmitted by the excitation sound wave signal through the communication tower; The voiceprint signals acquired at each collection time are divided into a training set and a validation set in a ratio of 7:3; the training set is used to train the environmental denoising model; The bone conduction voiceprint sensor acquires audio information in real time and sends it to the intelligent analysis terminal, which pre-processes the audio information to obtain the target monitoring audio and feeds it back to the voiceprint recognition module; The voiceprint recognition module loads the target monitoring audio into the bolt loosening classification model; And classification processing is performed based on the bolt loosening classification model to obtain the classification status result of the target monitoring audio.

Images