A method and system for identifying and monitoring the operating state of a bridge arm reactor

Abstract

The application discloses a kind of bridge arm reactor operating state identification and monitoring method and system, belong to the monitoring technical field of bridge arm reactor, method includes: collecting the sound signal in the operating environment of bridge arm reactor and pre-processing, obtain pre-processed audio signal;Utilize pre-training model BEATs to carry out feature extraction, obtain high-dimensional feature vector;Establish the benchmark of the initial operating state of bridge arm reactor including average initial state vector, average cosine similarity and average fluctuation amplitude;The operating state of bridge arm reactor is dynamically monitored by iteration monitoring module, the change score of operating state is calculated and compared with preset threshold, to determine whether the operating state of bridge arm reactor changes.The application integrates the method of iteratively updating in the monitoring method, to adapt to the slow-changing working condition in the operation process of bridge arm reactor, to efficiently and accurately monitor the equipment, improve the safety and reliability of equipment operation.

Description

Technical Field

[0001] This invention belongs to the field of bridge arm reactor monitoring technology, specifically relating to a method and system for identifying and monitoring the operating status of bridge arm reactors. Background Technology

[0002] As a core component of high-voltage converters, the operating status of bridge arm reactors directly affects the stability and safety of power electronic equipment. Abnormal or altered operation can not only lead to equipment failure but also impact the surrounding environment. Therefore, research on the operating status of bridge arm reactors has significant engineering value and practical implications.

[0003] Existing research primarily focuses on theoretical derivation, finite element analysis, and field testing, mainly addressing electromagnetic forces, mechanical vibrations, and the resulting noise characteristics. For example, some scholars have explored the electromagnetic force variations of bridge arm reactors under different operating conditions, while others have conducted systematic analyses of the noise characteristics of reactors used in flexible DC systems. Some studies have proposed inter-turn short-circuit diagnosis methods based on vibration distribution, achieving fault location; other literature has investigated modal characteristics and vibration mechanisms, providing a basis for avoiding resonance problems. Furthermore, there are already considerable research findings on reactor temperature rise patterns, structural segmentation optimization, and insulation structure design.

[0004] However, most of these studies focus on noise and structural optimization analysis under normal operating conditions, with limited attention paid to dynamic monitoring of operating status, and a particular lack of application of deep learning methods. In recent years, deep learning technology has shown significant advantages in equipment health monitoring and mechanical vibration pattern recognition. For example, some studies have used convolutional neural networks to perform pattern recognition on partial discharge signals of electrical equipment, achieving intelligent monitoring; others have proposed models based on recurrent neural networks for vibration pattern analysis and anomaly prediction; and still others have explored Transformer-based architectures, demonstrating good performance in denoising and feature extraction of mechanical vibration data. However, in the current identification and monitoring of the operating status of bridge arm reactors, it is difficult to achieve accurate identification and dynamic tracking of equipment status under complex operating conditions, especially in the slowly changing operating conditions of bridge arm reactors, and the safety and reliability of equipment operation need further improvement. Summary of the Invention

[0005] This invention provides a method and system for identifying and monitoring the operating status of bridge arm reactors. The purpose is to solve the problem that in the current identification and monitoring of the operating status of bridge arm reactors, it is difficult to achieve accurate identification and dynamic tracking of equipment status under complex operating conditions, especially in the slow-changing operating conditions of bridge arm reactors, and the safety and reliability of equipment operation need to be further improved.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a method for identifying and monitoring the operating status of a bridge arm reactor, comprising the following steps: S1. Collect the acoustic signal in the operating environment of the bridge arm reactor, preprocess the acoustic signal to obtain the preprocessed audio signal; S2. Based on the preprocessed audio signal, feature extraction is performed using the pre-trained model BEATs to obtain a high-dimensional feature vector. Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. S3. Based on high-dimensional feature vectors, establish a benchmark for the initial operating state of the bridge arm reactor, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. S4. Based on the initial operating state of the bridge arm reactor, the operating state of the bridge arm reactor is dynamically monitored through the iterative monitoring module. The change score of the operating state is calculated and compared with the preset threshold to determine whether the operating state of the bridge arm reactor has changed, thus completing the identification and monitoring of the operating state of the bridge arm reactor.

[0007] In some implementations, in S1, the acquisition of acoustic signals in the operating environment of the bridge arm reactor is based on: arranging multiple measuring points in the operating environment of the bridge arm reactor, pre-setting the number and distance of each measuring point, and placing a single-channel microphone at each measuring point, pre-setting the height, distance, and frequency of the microphone.

[0008] In some implementations, in S1, preprocessing includes converting the original audio signal into Fbank features according to the following formula (5): (5); in, For the extracted high-dimensional feature sequences, The original audio signal. To calculate the Fbank features of the audio signal.

[0009] Furthermore, in S2, feature extraction using the pre-trained model BEATs includes: inputting the pre-processed audio signal into the BEATs model, extracting multi-scale time-frequency features, and obtaining the final feature vector through average pooling using the following formula (6): (6); in, For time steps, For the final feature vector: Let be the feature of the t-th time step.

[0010] In some implementations, in S3, establishing the baseline for the initial operating state of the bridge arm reactor includes: The entire test audio segment is divided into audio segments of a preset time. The first few audio samples of the initial running state are taken. Each audio sample is mapped into a high-dimensional feature vector using a pre-trained BEATs model. The average of the feature vectors of all samples is then taken to obtain the average initial state vector.

[0011] Furthermore, in S3, the average initial state vector is calculated using the following formula (9): (9); in, This is a mathematical representation of the initial operating state of the bridge arm reactor, where N represents the first few audio features. No. A high-dimensional feature vector of an audio feature.

[0012] Furthermore, S3 also includes calculating the average cosine similarity and average fluctuation amplitude between each pair of the first N audio features; The mean cosine similarity is calculated using the following formulas (10) and (11): (10); (11); The average fluctuation range is calculated using the following formula (12): (12); in, S For the front N A set of cosine similarities between each pair of audio feature vectors. Z i and Z j It is the first i The and the first j The final feature vector of each audio file. The average cosine similarity, |S| For set S The number of elements in the middle. S i and S j for and The average cosine similarity between each sample and the remaining samples; M This represents the maximum number of samples.

[0013] In some implementations, in S4, calculating the state change fraction includes: Extract feature vectors from newly acquired audio samples; calculate the cosine similarity between the feature vector and the current average initial state vector; define the absolute value of the difference between the cosine similarity and the average fluctuation amplitude as the state change score.

[0014] In some implementations, in S4, when the state change score exceeds a preset threshold, it is determined that the device operating state has changed; otherwise, it is determined that the device state remains unchanged, and the current sample is included in the reference set to update the baseline.

[0015] This invention also provides a system for identifying and monitoring the operating status of a bridge arm reactor. The system is used to implement the aforementioned method for identifying and monitoring the operating status of a bridge arm reactor. The system includes a signal acquisition and preprocessing module, a feature extraction module, a reference establishment module, and a status determination and identification module, wherein: Signal acquisition and preprocessing module: used to acquire acoustic signals in the operating environment of the bridge arm reactor, preprocess the acoustic signals, and obtain preprocessed audio signals; Feature extraction module: Used to extract features from pre-trained BEATs models based on pre-processed audio signals to obtain high-dimensional feature vectors; Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. Benchmark establishment module: used to establish a benchmark of the initial operating state of the bridge arm reactor based on high-dimensional feature vectors, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. Status determination and identification module: Based on the initial operating status of the bridge arm reactor, the module dynamically monitors the operating status of the bridge arm reactor through an iterative monitoring module, calculates the change score of the operating status and compares it with a preset threshold, thereby determining whether the operating status of the bridge arm reactor has changed, and completing the identification and monitoring of the operating status of the bridge arm reactor.

[0016] Compared with the prior art, the present invention provides a method and system for identifying and monitoring the operating status of a bridge arm reactor, which has the following advantages: This invention provides a method for identifying and monitoring the operating status of a bridge arm reactor. It integrates an iteratively updated method into the monitoring method to adapt to slowly changing operating conditions during bridge arm reactor operation, thereby enabling effective monitoring of changes in operating status. This invention applies the BEATs model to the acoustic signal analysis of the bridge arm reactor, focusing on the differences in sound field characteristics under different operating states. It utilizes the attention mechanism of deep learning to learn the global dependencies of time-frequency features, thereby enhancing the ability to identify other states based on the acoustic information of the equipment in one state. Based on the BEATs pre-trained model and combined with the iterative update mechanism, the operating status of the bridge arm reactor is monitored to achieve accurate identification and dynamic tracking of the equipment status under complex operating conditions. Experimental results show that, compared with existing methods, the monitoring method of this invention has superior performance and effectiveness in identifying and monitoring the operating status of bridge arm reactors, effectively improving the safety and reliability of equipment operation and demonstrating excellent practicality. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and constitute a part of this invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation thereof. Figure 1 This is a flowchart illustrating a method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention.

[0018] Figure 2 This is a schematic diagram of the operation status monitoring process in an embodiment of the method for identifying and monitoring the operation status of a bridge arm reactor according to the present invention; Figure 3 This is a schematic diagram of the measuring room in the method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention; Figure 4 This is a schematic diagram of the t-SNE feature distribution of Fbank features in the method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention; Figure 5 This is a schematic diagram of the t-SNE feature distribution after BEATs processing in the method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention. Figure 6 This is a schematic diagram of the cosine similarity curve in the method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to indicate that selected embodiments of the invention are based on the embodiments in this invention. All other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, elements defined by the phrase "comprising one..." do not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functions, and operations of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this respect, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0021] How can we provide a method that combines deep learning models with reactor noise characteristics to identify and monitor the operating status of bridge arm reactors, thereby improving the safety and reliability of equipment operation?

[0022] like Figure 1 As shown, the present invention provides a method for identifying and monitoring the operating status of a bridge arm reactor, comprising the following steps: S1. Collect the acoustic signal in the operating environment of the bridge arm reactor, preprocess the acoustic signal to obtain the preprocessed audio signal; S2. Based on the preprocessed audio signal, feature extraction is performed using the pre-trained model BEATs to obtain a high-dimensional feature vector. Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. S3. Based on high-dimensional feature vectors, establish a benchmark for the initial operating state of the bridge arm reactor, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. S4. Based on the initial operating state of the bridge arm reactor, the operating state of the bridge arm reactor is dynamically monitored through the iterative monitoring module. The change score of the operating state is calculated and compared with the preset threshold to determine whether the operating state of the bridge arm reactor has changed, thus completing the identification and monitoring of the operating state of the bridge arm reactor.

[0023] This invention provides a method for identifying and monitoring the operating status of a bridge arm reactor. By collecting and preprocessing acoustic signals, the operating status of the bridge arm reactor is transformed into processable audio data, avoiding the complex hardware layout and interference problems required by traditional methods that rely on electromagnetic force or vibration measurements. Next, feature extraction is performed using a pre-trained model BEATs (Bidirectional Encoder representation from Audio Transformers). This model, based on the Transformer architecture and pre-trained on a large-scale audio dataset, can automatically learn multi-scale time-frequency features and global semantic information in acoustic signals, making it easier to capture differences between different operating states. This invention quantifies the operating state by establishing a benchmark including the average initial state vector, average cosine similarity, and average fluctuation amplitude, achieving accurate modeling of the initial state and ensuring the reliability of state determination. Finally, based on this benchmark, the state change score is dynamically calculated through an iterative monitoring module and compared with a preset threshold, achieving real-time tracking and automatic determination of the operating state. This enables timely identification of abnormal changes in equipment status, improving the response speed and accuracy of monitoring. This invention achieves efficient and automated monitoring through acoustic signal acquisition, deep learning feature extraction, benchmark establishment, and iterative monitoring, thereby improving the safety and reliability of equipment operation.

[0024] like Figure 2 As shown, in some actual working conditions, specifically, the method for identifying and monitoring the operating status of a bridge arm reactor according to the present invention is carried out according to the following steps: The propagation of sound in air is described by Stokes' equations of fluid dynamics, which state the medium's equations of state, continuity, and motion during the generation and propagation of sound waves. These equations collectively describe the relationships between pressure, sound intensity, and density as sound waves propagate through air. The following are the sound field equations: (1); In the formula: This represents the sound pressure at different locations in space. For fluid (air) density, For the speed of sound, The vibration velocity.

[0025] Taking into account room reflections, sound absorption, and geometric gain, the effective sound pressure distribution of a single bridge arm reactor can be calculated using the following formula: (2); Among them: points within the room r Receive signal at location r s As the origin of sound pressure, This represents the actual distance the sound wave travels from the sound source to the measuring point. For reference distance, As the attenuation factor, Related to factors such as frequency, materials, and environment, This describes the amplitude attenuation that occurs as the propagation distance increases. Gain representing the direction of sound wave propagation. P 0 represents the reference sound pressure level at the reference location, and the coefficient is... R This represents the room's reflection gain, which is related to the sound absorption characteristics of the room's boundaries.

[0026] The total effective sound pressure of multiple bridge arm reactors can be calculated using the following formula: (3); In the formula: P i The measurement of the first i Effective sound pressure of each bridge arm reactor.

[0027] The sound pressure level of the received room signal can be expressed by the following formula: (4); in, I Σ The total sound pressure level after multiple sound sources are superimposed; N The number of sound sources; P i Indicates the first i The effective sound pressure (unit: Pa) of a sound source at the receiving point; P ref The reference sound pressure is 20µPa, which is the reference sound pressure level for the human ear's hearing threshold in the air.

[0028] This invention arranges multiple measuring points in the operating environment of the bridge arm reactor, uses a microphone to collect sound signals under different operating conditions and performs preprocessing to obtain the time-frequency representation of the original audio signal.

[0029] The pre-trained model BEATs is used to perform end-to-end feature extraction on the pre-processed audio signal to generate a high-dimensional feature sequence. This model is pre-trained on a large-scale audio dataset and can capture local and global patterns in the audio signal, thereby improving the discriminative power of the features.

[0030] Specifically, to construct audio features, the continuous time-domain audio signal is first... Convert to Fbank time-frequency feature matrix ,in T For time frames, D This represents the dimension of the Mel filter. Inputting this matrix into the BEATs model extracts high-dimensional feature sequences. z The process is formally represented as: (5); in, For the extracted high-dimensional feature sequences, x(t) The original audio signal. Fbank(·) To calculate audio signals Fbank feature, BEATS The model is a deep feature extractor, which can extract features from... Fbank High-dimensional feature sequences from feature matrix X extract advanced semantic information. z , z∈R T×d for BEATs The extracted high-dimensional feature sequences can be used for subsequent calculations, where T For time frames, d for BEATs The feature dimensions output by the model. Then at time step T The final feature vector is obtained by performing average pooling on the top layer. Z : (6); in, T For time steps, Z Final feature vector: z t Features of the time step; z t ∈R d Indicates the first t Features of each time step.

[0031] First, the entire test audio segment is divided into 10-second segments, and the first 10 seconds of the initial running state are taken. N Audio sample of the bar: (7); Using pre-training BEATs Deep models map raw audio into high-dimensional feature vectors: (8); The average initial state vector is obtained by averaging the feature vectors of all samples. (9); in, This provides a mathematical representation of the initial operating state of the bridge arm reactor. N These are the first few audio features. Zi For the first i A high-dimensional feature vector of an audio feature.

[0032] Before calculation N The average cosine similarity between each pair of audio features and its average fluctuation rangeA : (10); (11); (12); in, S For the front N A set of cosine similarities between each pair of audio feature vectors. Z i and Z j It is the first i The and the first j The final feature vector of each audio file. The average cosine similarity, |S| For set S The number of elements in the middle. S i and S j for and The average cosine similarity between each sample and the remaining samples; M This represents the maximum number of samples.

[0033] This invention employs an iteratively updated monitoring method for the dynamic tracking of the operating status of bridge arm reactors. During equipment operation, the characteristic distribution of the initial state audio is continuously updated and used as a discrimination criterion to identify potential changes in the operating status.

[0034] First, the mean cosine similarity Used to describe the average trend and average fluctuation range under normal conditions. A This is used to indicate the acceptable range of fluctuations. In the subsequent operational status monitoring phase, newly acquired audio samples also undergo... BEATs Model extracts features Z n+1 And calculate the cosine similarity with the current average initial state vector. S' : (13); Similarity of new samples Compared with average similarity The degree of deviation is defined as the state change fraction. γ And set twice the benchmark fluctuation range 2 A To determine the threshold: (14); (15); in, If the score exceeds the set threshold, the device's operating status is considered to have changed; if the score is within the threshold range, the device's status is considered to remain unchanged. (16); If the standby state remains unchanged, then the sample is included in the reference set: (12); Then, formulas (9), (10), (11), and (12) are repeated to update the average initial state vector and related statistics in real time. Through this iterative update mechanism, the model can continuously adapt to the slowly changing operating conditions of the equipment, thereby achieving stable tracking of the operating state.

[0035] This invention also provides a system for identifying and monitoring the operating status of a bridge arm reactor. The system is used to implement the aforementioned method for identifying and monitoring the operating status of a bridge arm reactor. The system includes a signal acquisition and preprocessing module, a feature extraction module, a reference establishment module, and a status determination and identification module, wherein: Signal acquisition and preprocessing module: used to acquire acoustic signals in the operating environment of the bridge arm reactor, preprocess the acoustic signals, and obtain preprocessed audio signals; Feature extraction module: Used to extract features from pre-trained BEATs models based on pre-processed audio signals to obtain high-dimensional feature vectors; Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. Benchmark establishment module: used to establish a benchmark of the initial operating state of the bridge arm reactor based on high-dimensional feature vectors, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. Status determination and identification module: Based on the initial operating status of the bridge arm reactor, the module dynamically monitors the operating status of the bridge arm reactor through an iterative monitoring module, calculates the change score of the operating status and compares it with a preset threshold, thereby determining whether the operating status of the bridge arm reactor has changed, and completing the identification and monitoring of the operating status of the bridge arm reactor.

[0036] The following detailed description of a method and system for identifying and monitoring the operating status of a bridge arm reactor according to the present invention will be provided through specific embodiments.

[0037] Experimental setup: This embodiment collects audio data from the bridge arm reactor during its initial operating state and the stages of changing operating states. The data is downsampled to 16kHz and then input into the BEATs model for feature extraction to generate high-quality audio embeddings. To visually demonstrate the ability of BEATs to represent audio features, the original acoustic features (Fbank) are compared with the BEATs features, and t-SNE is used for dimensionality reduction and visualization. This allows observation of the differences in feature distribution in the low-dimensional space under different operating states. For the monitoring results, indices such as ACC, AUC, and pAUC are calculated.

[0038] Data preparation: like Figure 3 As shown, a total of 6 bridge arm reactors were arranged in the field measurement room. The distance between two adjacent bridge arm reactors was 2.5 meters, and the distance between two adjacent bridge arm reactors was 4 meters. To collect acoustic data, 9 measuring points were set up at different locations, each 3 meters apart. A single-channel microphone was placed at each measuring point, with a height of 0.9 meters and a distance of 7.24 meters from the bridge arm reactor, for acoustic signal acquisition. The sampling frequency was set to 48kHz, and the acoustic signals of the bridge arm reactors throughout their entire operating process, from initial operation to changes in operating status, were recorded.

[0039] The experimental results are shown in Table 1 below; Table 1 Performance evaluation of different pre-trained models

[0040] For all test audio, calculate the ACC, AUC, and pAUC evaluation metrics and visualize the feature distribution.

[0041] In addition to the BEATs model, two other pre-trained audio feature extraction models, AST and VGGish, were introduced to further compare the performance differences of different pre-trained models in the task of monitoring the operating status of bridge arm reactors. Table 1 summarizes the evaluation metrics of each model under the operating status monitoring task, including ACC, AUC, and pAUC. Experimental results show that the BEATs model performs excellently in all metrics, outperforming AST and VGGish, indicating that BEATs has stronger discriminative ability in capturing audio features and distinguishing different operating states.

[0042] like Figure 4 and Figure 5 As shown, Figure 4 and Figure 5The visualization results of the original acoustic features Fbank and the high-dimensional features extracted by BEATs in two-dimensional space are presented respectively. Experimental results show that in the two-dimensional image after t-SNE dimensionality reduction, the distribution of the original acoustic features cannot effectively distinguish the two operating states of the bridge arm reactor, and the samples are highly mixed in the feature space; however, after dimensionality reduction of the high-dimensional features extracted by BEATs, the samples of different operating states are clearly separated in two-dimensional space, indicating that the pre-trained model BEATs used in this invention can capture effective discriminative features, thereby improving the ability to monitor state changes.

[0043] like Figure 6 As shown, the curves of the cosine similarity calculated from the feature vector and the average initial state vector of each test sample during the test process are displayed. It can be seen that when the bridge arm reactor is in the initial operation, the curve is relatively stable and the cosine similarity remains above 0.9. When the operating state changes, the cosine similarity curve will show a significant decrease.

[0044] In summary, this invention provides a method and system for identifying and monitoring the operating state of a bridge arm reactor. Based on a pre-trained model BEATs, it captures the differences between different operating states and enhances the representation of extracted features, thereby achieving the monitoring of state changes. The pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture, pre-trained on the Audioset dataset, and capable of learning multi-scale time-frequency features and global semantic information of acoustic signals. This invention utilizes the pre-trained model to generate high-dimensional feature representations and construct a feature benchmark for the initial operating state, and dynamically corrects the benchmark features through an iterative update mechanism. During monitoring, acoustic signals are acquired in real time and features are extracted. Similarity is compared with the benchmark features to obtain a state change score, and a preset threshold is used to determine whether the operating state of the bridge arm reactor has changed. Final experimental results demonstrate that the method of this invention exhibits superior performance in identifying and monitoring operating states.

[0045] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for identifying and monitoring the operating status of a bridge arm reactor, characterized in that, Includes the following steps: S1. Collect the acoustic signal in the operating environment of the bridge arm reactor, preprocess the acoustic signal to obtain the preprocessed audio signal; S2. Based on the preprocessed audio signal, feature extraction is performed using the pre-trained model BEATs to obtain a high-dimensional feature vector. Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. S3. Based on high-dimensional feature vectors, establish a benchmark for the initial operating state of the bridge arm reactor, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. The baseline for establishing the initial operating state of the bridge arm reactor includes: The entire test audio segment is divided into audio segments of a preset time. The first few audio samples of the initial running state are taken. Each audio sample is mapped into a high-dimensional feature vector using a pre-trained BEATs model. The average of the feature vectors of all samples is then taken to obtain the average initial state vector. It also includes calculating the average cosine similarity and average fluctuation amplitude between each pair of the first N audio features; The mean cosine similarity is calculated using the following formulas (10) and (11): （10）； （11）； The average fluctuation range is calculated using the following formula (12): （12）； in, S For the front N A set of cosine similarities between each pair of audio feature vectors. Z i and Z j It is the first i The and the first j The final feature vector of each audio file. The average cosine similarity, |S| For set S The number of elements in the middle. S i and S j for and The average cosine similarity between each sample and the remaining samples; M The maximum number of samples; S4. Based on the initial operating state of the bridge arm reactor, the operating state of the bridge arm reactor is dynamically monitored through the iterative monitoring module. The change score of the operating state is calculated and compared with the preset threshold to determine whether the operating state of the bridge arm reactor has changed, thus completing the identification and monitoring of the operating state of the bridge arm reactor. The calculation of the fraction of state changes includes: Extract feature vectors from newly acquired audio samples; calculate the cosine similarity between this feature vector and the current average initial state vector; define the absolute value of the difference between the cosine similarity and the average cosine similarity as the state change score; and set the average fluctuation amplitude to twice 2. A The preset threshold; When the state change score exceeds the preset threshold, it is determined that the device operating state has changed; otherwise, it is determined that the device state remains unchanged, and the current sample is included in the reference set to update the baseline.

2. The method for identifying and monitoring the operating status of the bridge arm reactor according to claim 1, characterized in that, In S1, the acquisition of acoustic signals in the operating environment of the bridge arm reactor is based on: arranging multiple measuring points in the operating environment of the bridge arm reactor, pre-setting the number and distance of each measuring point, and placing a single-channel microphone at each measuring point, pre-setting the height, distance and frequency of the microphone.

3. The method for identifying and monitoring the operating status of the bridge arm reactor according to claim 1, characterized in that, In S1, the preprocessing includes converting the original audio signal into Fbank features according to the following formula (5): （5）； in, For the extracted high-dimensional feature sequences, The original audio signal. To calculate the Fbank features of the audio signal.

4. The method for identifying and monitoring the operating status of the bridge arm reactor according to claim 3, characterized in that, In S2, feature extraction using the pre-trained model BEATs includes: inputting the pre-processed audio signal into the BEATs model, extracting multi-scale time-frequency features, and obtaining the final feature vector through average pooling using the following formula (6): （6）； in, For time steps, For the final feature vector: Let be the feature of the t-th time step.

5. The method for identifying and monitoring the operating status of the bridge arm reactor according to claim 1, characterized in that, In S3, the average initial state vector is calculated using the following formula (9): （9）； in, This provides a mathematical representation of the initial operating state of the bridge arm reactor. These are the first few audio features. No. A high-dimensional feature vector of an audio feature.

6. A system for identifying and monitoring the operating status of a bridge arm reactor, characterized in that, The system is used to implement the method for identifying and monitoring the operating status of the bridge arm reactor as described in any one of claims 1-5. The system includes a signal acquisition and preprocessing module, a feature extraction module, a benchmark establishment module, and a status determination and identification module, wherein: Signal acquisition and preprocessing module: used to acquire acoustic signals in the operating environment of the bridge arm reactor, preprocess the acoustic signals, and obtain preprocessed audio signals; Feature extraction module: Used to extract features from pre-trained BEATs models based on pre-processed audio signals to obtain high-dimensional feature vectors; Among them, the pre-trained model BEATs is a self-supervised audio pre-training model based on the Transformer architecture. It is pre-trained on the Audioset dataset and is used to extract multi-scale time-frequency features and global semantic information of acoustic signals. Benchmark establishment module: used to establish a benchmark of the initial operating state of the bridge arm reactor based on high-dimensional feature vectors, including the average initial state vector, average cosine similarity, and average fluctuation amplitude. Status determination and identification module: Based on the initial operating status of the bridge arm reactor, the module dynamically monitors the operating status of the bridge arm reactor through an iterative monitoring module, calculates the change score of the operating status and compares it with a preset threshold, thereby determining whether the operating status of the bridge arm reactor has changed, and completing the identification and monitoring of the operating status of the bridge arm reactor.

Images