A communication room-oriented audio line state identification method and device

Abstract

This invention discloses a method and apparatus for audio line status identification in communication equipment rooms. The method includes audio data acquisition, acoustic physical constraint modeling, multi-region acoustic feature importance classification, audio line status identification model design, constraint loss dynamic balance optimization, acoustic feature loss function construction, and audio line status identification. This invention belongs to the field of data processing technology, specifically referring to a method and apparatus for audio line status identification in communication equipment rooms. This scheme models the sound wave propagation in the equipment room based on a three-dimensional acoustic wave equation to suppress non-fault noise interference; it divides the equipment room into zones, quantitatively allocates the number of monitoring points and regional loss weights, focuses on core areas, and significantly improves the fault detection rate of critical lines; it introduces the measured background noise sound pressure level of the equipment room to construct a noise correction coefficient, significantly improving the robustness of fault identification; and it designs an exponential dynamic relaxation coefficient, combined with benchmark weight allocation, to further enhance the early fault detection capability in critical areas.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and apparatus for identifying the status of audio lines in communication equipment rooms. Background Technology

[0002] Audio line status identification methods refer to a class of monitoring methods that automatically identify the working status of electrical lines, such as normal operation, poor contact, discharge, and aging, by collecting audio signals during operation and then using feature extraction and classification models. However, general audio line status identification methods suffer from several drawbacks. They fail to distinguish between critical equipment areas, densely populated line areas, and historically high-fault areas, resulting in insufficient accuracy in identifying high-value areas, wasted computing power in non-critical areas, and low efficiency in fault location and early warning. Furthermore, these methods do not consider the characteristics of strong background noise and large differences in loss magnitude in communication equipment rooms, making it impossible to distinguish between critical line areas and ordinary areas. Critical faults are easily smoothed out or weakened, further leading to poor audio line status identification results. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a method and apparatus for audio line status identification in communication equipment rooms. Addressing the problems of general audio line status identification methods failing to distinguish between critical equipment areas, densely populated line areas, and historically high-fault areas, resulting in insufficient accuracy in identifying high-value areas, wasted computing power in non-critical areas, and low efficiency in fault location and early warning, this solution models sound wave propagation in the equipment room based on a three-dimensional acoustic wave equation. This adapts to the complex reverberation and multi-reflection environment of the equipment room, suppresses non-fault noise interference, and maintains stable feature expression even at low signal-to-noise ratios. The equipment room is divided into zones, and the number of monitoring points and regional loss weights are quantitatively allocated. Focusing on the high-value, high-fault, and densely populated core areas of the communication equipment room, this achieves precise allocation of monitoring resources, making early faults with small signals easier to identify and significantly improving the detection of critical line faults. To address the issue that conventional audio line status identification methods fail to consider the strong background noise and significant differences in loss magnitude in communication equipment rooms, making it impossible to distinguish between critical and ordinary areas, and leading to the smoothing and weakening of critical faults, resulting in poor audio line status identification performance, this solution introduces a noise correction coefficient based on the measured background noise sound pressure level of the equipment room. This coefficient dynamically adjusts the proportion of physical and edge constraints, automatically strengthening physical constraints and suppressing noise interference under high background noise conditions in communication equipment rooms, significantly improving the robustness of identifying line discharge and poor contact faults at low signal-to-noise ratios. Furthermore, an exponential dynamic relaxation coefficient is designed, combined with baseline weight allocation, to enhance generalization ability and fault sound feature purity. Higher loss weights are assigned to densely populated and high-fault areas in the equipment room, further strengthening the early fault detection capability in critical areas, thereby improving the audio line status identification performance.

[0004] The technical solution adopted by this invention is as follows: This invention provides a method for identifying the status of audio lines in communication equipment rooms, the method comprising the following steps: Step S1: Audio data acquisition; Step S2: Acoustic physical constraint modeling; Step S3: Classification of the importance of acoustic features in multiple regions; Step S4: Design of audio line status recognition model; Step S5: Dynamic balance optimization of constraint loss; Step S6: Constructing the acoustic feature loss function; Step S7: Audio line status identification.

[0005] Furthermore, in step S1, the audio data acquisition is performed by acquiring audio data under different line conditions through sensors; the acquired audio data is preprocessed to finally obtain an audio spatiotemporal coordinate dataset.

[0006] Furthermore, in step S2, the acoustic physical constraint modeling is based on the three-dimensional wave equation to describe the propagation mechanism of sound waves in the computer room, and the edge absorption conditions are established by combining the acoustic impedance characteristics of different materials in the computer room.

[0007] Furthermore, in step S3, the multi-region acoustic feature importance division is based on the layout of the equipment room and audio characteristics, dividing the entire monitoring area into zones of different importance; after dividing the zones, the number of monitoring points is allocated to different zones, and the regional acoustic feature weights are calculated.

[0008] Furthermore, in step S4, the audio line state recognition model design adopts a dual-channel feature fusion network as its basic architecture, including: a feature extraction branch, which extracts sound texture features through convolutional layers; a physical reasoning branch, which learns the implicit representation of the sound pressure field through a multilayer perceptron; a fusion decision head, which concatenates the features of the two branches and outputs the final line state probability through a fully connected layer; and an output layer that uses Softmax classification.

[0009] Furthermore, in step S5, the constraint loss dynamic balance optimization is to design a dynamic weighting strategy based on the benchmark loss calibration: a forward propagation is performed before training to calculate the benchmark magnitude of various losses and assign reverse benchmark weights. Then, a relaxation coefficient is introduced to gradually weaken the weight of edge conditions in the later stage of training, and a computer room noise correction term is introduced to improve the anti-interference ability of physical constraints under low signal-to-noise ratio.

[0010] Furthermore, in step S6, the acoustic feature loss function construction involves constructing a loss function that integrates acoustic physical constraints, multi-region weighting, and dynamic balance, and completing the training of the audio line state recognition model.

[0011] Furthermore, in step S7, the audio line status recognition is based on the trained audio line status recognition model. Audio data is acquired in real time, preprocessed, and then input into the audio line status recognition model. The model outputs a Softmax probability, and the category with the highest probability is taken as the current line status. If the current line status is a fault type, an early warning is issued to the management personnel, and the location coordinate information is reported.

[0012] The present invention provides an audio line status identification device for communication equipment rooms, comprising an audio data acquisition module, an acoustic physical constraint modeling module, a multi-region acoustic feature importance division module, an audio line status identification model design module, a constraint loss dynamic balance optimization module, an acoustic feature loss function construction module, and an audio line status identification module; The audio data acquisition module acquires audio data under different line conditions, and after preprocessing, constructs an audio spatiotemporal coordinate dataset. The acoustic physical constraint modeling module describes the sound wave propagation mechanism in the computer room based on the three-dimensional wave equation and establishes edge absorption conditions to complete the acoustic physical constraint modeling. The multi-region acoustic feature importance division module divides the region into different weighted partitions, assigns monitoring points to each partition, and calculates the regional acoustic feature weights. The audio line status recognition model design module adopts a dual-channel feature fusion network, combining feature extraction branch, physical inference branch and fusion decision head to design the audio line status recognition model; The constraint loss dynamic balance optimization module determines the benchmark magnitude of various losses and assigns benchmark weights through forward propagation, and introduces noise correction terms and dynamic weight relaxation coefficients to achieve constraint loss dynamic balance optimization. The acoustic feature loss function construction module constructs three types of losses: basic acoustic features, acoustic physical constraints, and edge conditions. It combines multi-region weighting and regional acoustic feature weights to establish a total loss function and complete the training of the audio line state recognition model. The audio line status recognition module performs real-time audio line status recognition based on the trained audio line status recognition model.

[0013] The beneficial effects achieved by the present invention using the above solution are as follows: (1) In view of the problem that general audio line status identification methods do not distinguish between key equipment areas, dense line areas and historical high fault areas, resulting in insufficient identification accuracy of high-value areas, wasted computing power in non-critical areas, and low efficiency of fault location and early warning, this solution is based on three-dimensional acoustic wave equation modeling of sound wave propagation in the computer room, adapting to the complex reverberation and multi-reflection environment of the computer room, suppressing non-fault noise interference, and maintaining stable feature expression under low signal-to-noise ratio; dividing the computer room into zones, quantifying the allocation of monitoring points and regional loss weights, focusing on the core areas of high value, high fault, and dense line in the communication computer room, realizing precise tilting of monitoring resources, making it easier to identify early faults of small signals, and significantly improving the fault detection rate of key lines.

[0014] (2) In view of the fact that general audio line status identification methods do not take into account the characteristics of strong background noise and large differences in loss magnitude in communication equipment rooms, they cannot distinguish between critical line areas and ordinary areas in the equipment room, and critical faults are easily smoothed and weakened, resulting in poor audio line status identification effect, this solution introduces the measured background noise sound pressure level of the equipment room to construct a noise correction coefficient, dynamically adjusts the proportion of physical constraints and edge constraints, automatically strengthens physical constraints and suppresses noise interference under high background noise in the communication equipment room, and significantly improves the robustness of line discharge and poor contact fault identification under low signal-to-noise ratio; designs an exponential dynamic relaxation coefficient, combined with the baseline weight allocation, to improve the generalization ability and fault sound feature purity, and assigns higher loss weights to the dense line area and high fault area of ​​the equipment room, further strengthening the early fault detection capability of critical areas; thereby improving the audio line status identification effect. Attached Figure Description

[0015] Figure 1 A flowchart illustrating an audio line status identification method for communication equipment rooms provided by the present invention; Figure 2 This is a schematic diagram of an audio line status identification device for communication equipment rooms provided by the present invention.

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0019] Example 1, see Figure 1 This invention provides a method for identifying the status of audio lines in communication equipment rooms, the method comprising the following steps: Step S1: Audio data acquisition, acquiring audio data under different line conditions, and constructing an audio spatiotemporal coordinate dataset after preprocessing; Step S2: Acoustic physical constraint modeling. The acoustic wave propagation mechanism in the computer room is described based on the three-dimensional wave equation, and edge absorption conditions are established to complete the acoustic physical constraint modeling. Step S3: Divide the importance of acoustic features in multiple regions, divide the regions into different weight zones, assign the number of monitoring points to each zone and calculate the weight of the acoustic features in each region; Step S4: Audio line status recognition model design. A dual-channel feature fusion network is adopted, which combines feature extraction branch, physical inference branch and fusion decision head to design an audio line status recognition model. Step S5: Dynamic balance optimization of constraint loss. The baseline magnitudes of various losses are determined and baseline weights are assigned through forward propagation. A noise correction term and a dynamic weight relaxation coefficient are introduced to achieve dynamic balance optimization of constraint loss. Step S6: Construct the acoustic feature loss function. Construct three types of losses: basic acoustic features, acoustic physical constraints, and edge conditions. Combine multi-region weighting and regional acoustic feature weights to establish the total loss function and complete the training of the audio line state recognition model. Step S7: Audio line status identification, real-time audio line status identification is performed based on the trained audio line status identification model.

[0020] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, audio data acquisition is performed by acquiring audio data under different line conditions through sensors: microphone arrays are arranged in different locations in the equipment room to capture spatial audio distribution characteristics and collect audio data under normal line conditions, poor contact conditions, insulation aging conditions, and partial discharge conditions; the collected audio data is preprocessed: the time domain signal is converted into a time-frequency domain representation using short-time Fourier transform, and the frequency axis is nonlinearly perceptually weighted using Mel filter banks to highlight the frequency bands sensitive to human ears and discharge, suppress high-frequency noise and low-frequency interference, the audio sampling rate is set to 48kHz, the frame length is 25ms, the frame shift is 10ms, and the number of Mel filter banks is 128; finally, an audio spatiotemporal coordinate dataset is obtained.

[0021] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the acoustic physical constraint modeling is based on the three-dimensional wave equation to describe the propagation mechanism of sound waves in the computer room. Combined with the acoustic impedance characteristics of different materials such as the computer room wall (concrete), cabinet (metal), and sound-absorbing cotton, edge absorption conditions are established to ensure that the neural network, when predicting the sound field, not only fits the data but also satisfies the physical conservation laws, thereby having a stronger generalization ability in low signal-to-noise ratio environments. The three-dimensional acoustic wave equation is expressed as: Establish the edge absorption conditions for the computer room, represented as: Where p is the sound pressure field and c is the sound velocity in the computer room; It is the Laplace operator, which characterizes the diffusion effect in space; It is the acoustic disturbance source term located at the spatiotemporal coordinate point (x,y,z,t), which characterizes the sound source intensity of the sound waves excited by partial discharge, arc, and poor contact faults in the computer room at the corresponding spatiotemporal location. x,y,z are three-dimensional coordinates, and t is time. α is the sound pressure gradient in the normal direction of the edge; α is the edge absorption coefficient; metal surface ≈ 0.001m -1 Sound-absorbing material ≈ 0.1~0.3m -1 .

[0022] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, the importance division of multi-region acoustic features is based on the layout of the equipment room and audio characteristics. The entire monitoring area is divided into zones of different importance, with higher monitoring weights assigned to critical equipment areas and areas with dense lines. The number of monitoring points is allocated to different areas based on equipment importance, line density, and historical failure rate, as shown below: ; Calculate the regional acoustic feature weights, expressed as: ; ; used to weight the acoustic feature loss of different regions, highlighting the recognition accuracy of key regions; among them, This represents the number of monitoring points assigned to the k-th partition; K is the number of partitions. and These are the risk weight coefficients for the k-th partition and the i-th partition, respectively; and These are the physical areas of the k-th partition and the i-th partition, respectively; This is the total number of monitoring points; It is the weight of the k-th partition in the total loss function, used to calculate the identification error of the weighted region; This is the sensitivity coefficient, with a value ranging from 0.5 to 1; The length of cables related to partition faults (only counting those that have experienced faults) needs to be normalized.

[0023] By performing the above operations, this solution addresses the problems of general audio line status identification methods failing to distinguish between critical equipment areas, densely populated line areas, and historically high-fault areas, resulting in insufficient identification accuracy for high-value areas, wasted computing power in non-critical areas, and low efficiency in fault location and early warning. This solution models the sound wave propagation in the equipment room based on the three-dimensional acoustic wave equation, adapting to the complex reverberation and multi-reflection environment of the equipment room, suppressing non-fault noise interference, and maintaining stable feature expression even under low signal-to-noise ratio. The solution divides the equipment room into zones, quantitatively allocates the number of monitoring points and regional loss weights, focuses on the core areas of the communication equipment room that are high-value, high-fault, and densely populated with lines, achieves precise allocation of monitoring resources, makes early faults of small signals easier to identify, and significantly improves the fault detection rate of critical lines.

[0024] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the audio line state recognition model is designed to simultaneously satisfy the high accuracy of audio classification and the physical consistency of sound field prediction. It adopts a dual-channel feature fusion network as the basic architecture, including: a feature extraction branch, which processes the input Mel spectrum and extracts the sound texture features of partial discharge and arc anomalies through convolutional layers; and a physical inference branch, which takes the spatiotemporal coordinates as input and learns the implicit representation of the sound pressure field through a multilayer perceptron. The decision fusion head concatenates the features from the two branches and outputs the final line state probability through a fully connected layer; the output layer uses Softmax classification, and the final prediction is represented as: ;in, It is the predicted probability distribution output by the model, the predicted probability of each line state; It is a reference sound pressure level; It is the softmax activation function; W and b are the weight matrix and bias vector, respectively; These are the features extracted by the feature extraction branch.

[0025] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, the dynamic balance optimization of constraint loss is designed based on the dynamic weighting strategy of the benchmark loss calibration for different constraints such as acoustic wave equation constraints, benchmark sound field conditions, and edge conditions. Before training, a forward propagation is performed to calculate the benchmark magnitude of each type of loss, and the reverse benchmark weights are assigned accordingly to ensure that each constraint has equal influence at the starting point of training. Then, a relaxation coefficient is introduced to gradually weaken the weight of the edge conditions in the later stage of training to prevent the model from overfitting edge noise. A computer room noise correction term is also introduced to improve the anti-interference ability of physical constraints under low signal-to-noise ratio. The average sound pressure level of the background noise is obtained through actual measurement using a computer room microphone array. Calculate the noise correction factor , is represented as: ; It is the reference sound pressure level; the reference constraint weight is expressed as: ; ; ;in, , and These are the baseline weights for acoustic physical constraints, basic acoustic features, and edge conditions, respectively. , and These are the baseline loss values ​​for acoustic physical constraints, basic acoustic features, and edge conditions, respectively; the dynamic weight relaxation coefficient is calculated, and the edge constraint weights are reduced in the later stages of training to avoid overfitting the edge conditions, expressed as: ;in, is the marginal relaxation coefficient during training, and n is the current training step number; It is the baseline relaxation coefficient; is the relaxation rate, which takes a value of 3 to 5, controlling the relaxation slope; T is the total number of training steps. It is a smooth term, a very small positive number.

[0026] Example 7, see Figure 1 This embodiment is based on the above embodiment. In step S6, the acoustic feature loss function is constructed by building a loss function that integrates acoustic physical constraints, multi-region weighting, and dynamic balance to improve the accuracy of audio line state recognition; and a basic acoustic feature loss function is established. , is represented as: ;in, These are model parameters; This refers to the batch size of the audio training samples; It is the state predicted by the model. This is the u-th audio sample data. This represents the true state of the audio sample data; an acoustic physical constraint loss is established. , is represented as: ;in, It refers to the number of physical constraint monitoring points, which is the number of coordinate points randomly sampled in the spatiotemporal domain; It is the m-th spatiotemporal coordinate point; It is a physical operator based on the wave equation. Establish the edge condition loss, expressed as: ;in, It represents the number of edge monitoring points, specifically the number of coordinate points sampled on the edge surface. It is the gradient of sound pressure on the edge normal n; finally, the total loss function is constructed. , is represented as: ;in, It is the local prediction error within the k-th physical region, and the mean square error is selected. , and These are the basic acoustic features, acoustic physical constraints, and edge condition weights for the nth training iteration, obtained by multiplying the baseline weights by the edge relaxation coefficient. , and These are the normalized basic acoustic features, acoustic physical constraints, and edge condition loss, respectively. Audio line state recognition model training: Based on the audio spatiotemporal coordinate dataset, the basic classification loss, physical constraint loss, edge loss and multi-region weighted loss are calculated respectively, and the total loss is obtained by weighting according to the dynamic weight strategy. The network parameters are updated by backpropagation. During the training process, the edge constraint weights are dynamically relaxed to avoid overfitting. The training is completed when both the physical constraint error and the classification accuracy converge.

[0027] By performing the above operations, this solution addresses the problem that general audio line status identification methods do not consider the characteristics of strong background noise and large differences in loss magnitude in communication equipment rooms, making it impossible to distinguish between critical line areas and ordinary areas. Key faults are easily smoothed and weakened, leading to poor audio line status identification. This solution introduces a noise correction coefficient based on the measured background noise sound pressure level in the equipment room, dynamically adjusting the proportion of physical constraints and edge constraints. Under high background noise in the communication equipment room, it automatically strengthens physical constraints and suppresses noise interference, significantly improving the robustness of identifying line discharge and poor contact faults at low signal-to-noise ratios. An exponential dynamic relaxation coefficient is designed, combined with baseline weight allocation, to improve generalization ability and fault sound feature purity. Higher loss weights are assigned to densely populated and high-fault areas in the equipment room, further strengthening the early fault detection capability in critical areas; thus improving the audio line status identification effect.

[0028] Example 8, see Figure 1This embodiment is based on the above embodiment. In step S7, the audio line status recognition is based on the trained audio line status recognition model. Audio data is acquired in real time, preprocessed, and then input into the audio line status recognition model. The model outputs the Softmax probability, and the category with the highest probability is taken as the current line status. If the current line status is a fault type, an early warning is given to the management personnel, and the location coordinate information is reported.

[0029] Example 9, see Figure 2 Based on the above embodiments, this invention provides an audio line status identification device for communication equipment rooms, including an audio data acquisition module, an acoustic physical constraint modeling module, a multi-region acoustic feature importance division module, an audio line status identification model design module, a constraint loss dynamic balance optimization module, an acoustic feature loss function construction module, and an audio line status identification module. The audio data acquisition module acquires audio data under different line conditions, and after preprocessing, constructs an audio spatiotemporal coordinate dataset. The acoustic physical constraint modeling module describes the sound wave propagation mechanism in the computer room based on the three-dimensional wave equation and establishes edge absorption conditions to complete the acoustic physical constraint modeling. The multi-region acoustic feature importance division module divides the region into different weighted partitions, assigns monitoring points to each partition, and calculates the regional acoustic feature weights. The audio line status recognition model design module adopts a dual-channel feature fusion network, combining feature extraction branch, physical inference branch and fusion decision head to design the audio line status recognition model; The constraint loss dynamic balance optimization module determines the benchmark magnitude of various losses and assigns benchmark weights through forward propagation, and introduces noise correction terms and dynamic weight relaxation coefficients to achieve constraint loss dynamic balance optimization. The acoustic feature loss function construction module constructs three types of losses: basic acoustic features, acoustic physical constraints, and edge conditions. It combines multi-region weighting and regional acoustic feature weights to establish a total loss function and complete the training of the audio line state recognition model. The audio line status recognition module performs real-time audio line status recognition based on the trained audio line status recognition model.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 process, method, article, or apparatus.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0032] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for identifying the status of audio lines in a communication equipment room, characterized in that: The method includes the following steps: Step S1: Audio data acquisition, acquiring audio data under different line conditions, and constructing an audio spatiotemporal coordinate dataset after preprocessing; Step S2: Acoustic physical constraint modeling. The acoustic wave propagation mechanism in the computer room is described based on the three-dimensional wave equation, and edge absorption conditions are established to complete the acoustic physical constraint modeling. Step S3: Divide the importance of acoustic features in multiple regions, divide the regions into different weight zones, assign the number of monitoring points to each zone and calculate the weight of the acoustic features in each region; Step S4: Audio line status recognition model design. A dual-channel feature fusion network is adopted, which combines feature extraction branch, physical inference branch and fusion decision head to design an audio line status recognition model. Step S5: Dynamic balance optimization of constraint loss. The baseline magnitudes of various losses are determined and baseline weights are assigned through forward propagation. A noise correction term and a dynamic weight relaxation coefficient are introduced to achieve dynamic balance optimization of constraint loss. Step S6: Construct the acoustic feature loss function. Construct three types of losses: basic acoustic features, acoustic physical constraints, and edge conditions. Combine multi-region weighting and regional acoustic feature weights to establish the total loss function and complete the training of the audio line state recognition model. Step S7: Audio line status identification, real-time audio line status identification is performed based on the trained audio line status identification model.

2. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S3, the multi-region acoustic feature importance division is based on the equipment room layout and audio characteristics, dividing the entire monitoring area into zones of different importance; After dividing the region, the number of monitoring points is allocated to different regions, and the regional acoustic feature weights are calculated.

3. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S5, the constraint loss dynamic balance optimization is to design a dynamic weighting strategy based on the benchmark loss calibration: before training, a forward propagation is performed to calculate the benchmark magnitude of various losses and assign reverse benchmark weights. Then, a relaxation coefficient is introduced to gradually weaken the weight of edge conditions in the later stage of training, and a computer room noise correction term is introduced to improve the anti-interference ability of physical constraints under low signal-to-noise ratio.

4. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S6, the acoustic feature loss function construction involves constructing a loss function that integrates acoustic physical constraints, multi-region weighting, and dynamic balance, and completing the training of the audio line state recognition model.

5. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S2, the acoustic physical constraint modeling is based on the three-dimensional wave equation to describe the propagation mechanism of sound waves in the computer room, and the edge absorption conditions are established by combining the acoustic impedance characteristics of different materials in the computer room.

6. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S4, the audio line state recognition model is designed using a dual-channel feature fusion network as its basic architecture, including: a feature extraction branch that extracts sound texture features through convolutional layers; a physical reasoning branch that learns the implicit representation of the sound pressure field through a multilayer perceptron; a fusion decision head that concatenates the features of the two branches and outputs the final line state probability through a fully connected layer; and an output layer that uses Softmax classification.

7. The method for identifying the status of audio lines in a communication equipment room according to claim 1, characterized in that: In step S7, the audio line status recognition is based on the trained audio line status recognition model. Audio data is acquired in real time, preprocessed, and then input into the audio line status recognition model. The model outputs a Softmax probability, and the category with the highest probability is taken as the current line status. If the current line status is a fault type, an early warning is issued to the management personnel, and the location coordinate information is reported.

8. An audio line status identification device for a communication equipment room, used to implement the audio line status identification method for a communication equipment room as described in any one of claims 1-7, characterized in that: It includes an audio data acquisition module, an acoustic physical constraint modeling module, a multi-region acoustic feature importance classification module, an audio line state recognition model design module, a constraint loss dynamic balance optimization module, an acoustic feature loss function construction module, and an audio line state recognition module; The audio data acquisition module acquires audio data under different line conditions, and after preprocessing, constructs an audio spatiotemporal coordinate dataset. The acoustic physical constraint modeling module describes the sound wave propagation mechanism in the computer room based on the three-dimensional wave equation and establishes edge absorption conditions to complete the acoustic physical constraint modeling. The multi-region acoustic feature importance division module divides the region into different weighted partitions, assigns monitoring points to each partition, and calculates the regional acoustic feature weights. The audio line status recognition model design module adopts a dual-channel feature fusion network, combining feature extraction branch, physical inference branch and fusion decision head to design the audio line status recognition model; The constraint loss dynamic balance optimization module determines the benchmark magnitude of various losses and assigns benchmark weights through forward propagation, and introduces noise correction terms and dynamic weight relaxation coefficients to achieve constraint loss dynamic balance optimization. The acoustic feature loss function construction module constructs three types of losses: basic acoustic features, acoustic physical constraints, and edge conditions. It combines multi-region weighting and regional acoustic feature weights to establish a total loss function and complete the training of the audio line state recognition model. The audio line status recognition module performs real-time audio line status recognition based on the trained audio line status recognition model.

Images