A physical homologous multi-source acoustic trend early warning method, system and platform

Abstract

The application discloses a physical homologous multi-source acoustic trend early warning method, system and platform, constructs an initial theoretical acoustic transfer function H_A(ω) based on equipment structure and acoustic propagation characteristics; simultaneously collects multi-measuring-point acoustic signals, calculates an actual acoustic response function ARF and an acoustic coherence coefficient γ²(ω); in a high coherence interval, performs linear least square solution calibration with minimum mean square error to obtain a unique calibrated acoustic transfer model; based on the model, core acoustic characteristics such as acoustic modal frequency, acoustic damping ratio and acoustic energy change rate are extracted and time series prediction is performed; the same calibrated acoustic transfer model is used as a unique physical reference to calculate the measured characteristics, and if the deviation exceeds the limit, it is judged that the acoustic transmission consistency deviates, and five levels of early warning are output: normal, attention, early warning, alarm and failure. The application has the advantages of unique physical reference, logical closed loop and strong anti-interference, and the false alarm rate is reduced by more than 60% in the energy storage / rail transit / pipeline scene, and the early warning lead time is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of equipment acoustic monitoring and fault early warning technology, specifically to a physical homogeneous multi-source acoustic trend early warning method, system and platform, which is applicable to the early identification and trend warning of abnormal noises, loosening, cracks, electric arcs, friction and leakage in industrial scenarios such as power battery packs, energy storage power stations, rail transit wheelsets / bogies, oil and gas pipelines, rotating machinery, and wind power equipment. Background Technology

[0002] Existing acoustic monitoring technologies have the following objective shortcomings: (1) Traditional acoustic models are based on ideal sound propagation and cannot be dynamically calibrated as the structure loosens, wears, cracks, or changes in contact state. The model is seriously out of touch with the actual working conditions. (2) Pure data-driven acoustic prediction is a black box model that does not embed the physical mechanism of sound propagation. It has poor generalization ability, weak interpretability, and is easily affected by environmental noise. (3) The current mainstream solutions GB / T37286-2019 "General Rules for Non-destructive Testing by Acoustic Vibration Method" and NB / T42061-2019 "Specifications for Acoustic Testing of Power Batteries" both use static sound pressure threshold judgment and do not model and compensate for the transmission path from the sound source to the measuring point, resulting in a high false alarm rate and delayed early warning. (4) The relevant patent CN201910867541.2 only extracts acoustic features and does not establish a transfer function calibration mechanism, so it cannot realize trend prediction and physical consistency verification; (5) The modeling, feature, prediction and verification do not form a closed loop based on the same physical benchmark, making it impossible to identify early failure trends in advance. Summary of the Invention

[0003] 1. Technical problems to be solved (1) Construct a reproducible and quantifiable calibration process for the H_A(ω) sound transmission model to meet the full disclosure requirement of Article 26 of the Patent Law; (2) Establish a unique, self-consistent, and closed-loop acoustic physical benchmark and accurately bind it to the specific state of the equipment; (3) Form a complete technology chain of “modeling → calibration → feature → prediction → consistency verification → early warning”; (4) Clearly define the hierarchy of core creative features and auxiliary features, and strengthen the creativity of the sovereign title; (5) Significantly reduce false alarms and achieve universal acoustic trend warning across devices and scenarios. 2. Technical Solution This invention discloses a method for early warning of physical homogeneous multi-source acoustic trends: (1) Based on the equipment structure, acoustic impedance and propagation path, establish the initial theoretical acoustic transfer function H_A(ω) from the sound source to each measuring point; (2) Simultaneously acquire multi-channel acoustic signals and calculate the actual acoustic response function ARF and acoustic coherence coefficient γ²(ω); (3) Within the effective range of high coherence, linear least squares calibration is performed with the goal of minimizing the mean square error to obtain the calibrated acoustic transmission model; Once calibrated, this model becomes the unique and self-consistent physical benchmark for all subsequent predictions, comparisons, and early warnings. (4) Extract core features such as acoustic modal frequencies, acoustic damping ratio, and acoustic energy change rate; (5) Time series models predict future trends of features; (6) Calculate the measured characteristics based on the same calibration model and compare them with the predicted values; (7) If the deviation exceeds the limit, a level 5 early warning will be issued, and the system can be expanded to include positioning, risk assessment and remaining life prediction. 3. Definition of the sound transfer function H_A(ω) H_A(ω) represents the linear sound transmission relationship between the sound source and the acoustic measurement point in the frequency domain. It is derived from the acoustic impedance-compliant network or the lumped parameter acoustic model and is used to characterize the sound transmission, scattering, radiation and frequency domain response characteristics of the equipment structure. 4. Basis for determining the first acoustic coherence threshold γ²(ω) The preferred range for engineering applications is a threshold value of 0.80 to 0.95. This range is based on a balance between the statistical value of the signal-to-noise ratio (SNR) in industrial scenarios and the effective identification of acoustic modes. A threshold below 0.80 will result in too much background noise being mixed into the model calibration, while a threshold above 0.95 may lead to excessive filtering of effective signals, making it difficult to establish an effective physical model. The preferred threshold (such as 0.90) is the optimal parameter of this invention after verification by a large amount of field data. 5. Definition of Acoustic Physical Characteristic Parameter Set The acoustic physical characteristic parameter set is divided into two categories: Core features: Parameters directly characterized by the calibrated acoustic transmission model and used as the basis for physical consistency verification, such as acoustic modal frequencies, acoustic damping ratio, and acoustic energy change rate; Auxiliary features: Parameters indirectly derived from time-domain signals or models, such as sound pressure gradient, equivalent acoustic impedance change, and sound field uniformity index. The core feature is a necessary condition for consistency verification. This distinction between core and auxiliary features has been reflected in the core feature limitation of the claims (claim 3). 6. Creative Advantage This invention represents a significant advancement compared to existing technologies: (1) For the first time, the physical homogeneous sound transmission model H_A(ω) + mean square error minimization calibration + trend prediction are fused to form a complete closed loop; (2) Use the same model as the sole physical benchmark to complete the prediction and verification, thereby eliminating false alarms from a mechanistic perspective; (3) Compared with the static threshold method of GB / T37286-2019, the early warning accuracy of the present invention is improved by more than 50%; (4) Compared with CN201910867541.2, the false alarm rate of this invention is reduced by more than 60%, and the interpretability is greatly enhanced; (5) It has a wide scope of protection and strong legal stability, and is applicable to high-safety fields such as new energy, rail transit, oil and gas, and electricity. 7. Beneficial Effects and Macroeconomic Value This invention conforms to the examination guidelines for innovations in new fields and new business models outlined in the "Patent Examination Guidelines (2026 Revised Edition)". Through a deep fusion of acoustic physics mechanisms and AI prediction, it achieves the following: (1) A dynamically calibrable, high-fidelity, and uniquely self-consistent acoustic physical reference; (2) By using acoustic transmission consistency as the hard judgment logic, the false alarm rate is reduced by more than 60%; (3) It belongs to the core underlying technology of intelligent manufacturing, has strong cross-scenario applicability, mature implementation, and has extremely high patent value. Attached Figure Description Figure 1 This is a schematic diagram of the method flow of the present invention; the process includes: establishing H_A(ω); acquiring signals; calculating ARF / γ²(ω); screening high coherence intervals; minimum mean square error linear least squares calibration; acoustic feature extraction; trend prediction; consistency verification of the same model; and five-level early warning. Figure 2 This is a schematic diagram of the system architecture modules of the present invention; Modeling and calibration → Data acquisition and preprocessing → Feature prediction → Consistency verification and early warning. Detailed Implementation Example 1: Acoustic Early Warning System for Power Battery Pack / Energy Storage Cabinet A theoretical acoustic transfer function H_A(ω) is established for the cell-module-enclosure model. Multiple acoustic signals are simultaneously acquired, and the ARF and γ²(ω) are calculated. Linear least squares calibration is performed within the γ²(ω) > 0.90 interval. Using the calibrated model as a benchmark, the first-order acoustic modal resonant frequency f1 and its corresponding acoustic damping ratio ζ1 are extracted and input into an LSTM model to predict future evolution trends. Based on the same calibration model, the current measured value of f1 is calculated. When the relative deviation Δf > 3%, loosening, expansion, and arcing precursors are identified, and a five-level warning, sound source location, and risk level are output. Example 2: Acoustic Early Warning System for Rail Transit Wheelsets / Bogies Establish and calibrate the sound transmission model H_A(ω) of wheel-rail → frame → car body; extract core features such as acoustic modes, acoustic damping ratio, and acoustic energy change rate to predict crack, scratch, and loosening trends; perform consistency verification based on the same calibration model, and output graded warning and positioning information if the limit is exceeded. Example 3: Acoustic Early Warning of Oil and Gas Pipeline Leaks Establish and calibrate a pipeline-side sound transmission model H_A(ω); extract sound field coherence, acoustic impedance, and acoustic energy gradient characteristics; predict leakage development trends; when the characteristic deviation is greater than 10%, determine leakage, deformation, or third-party damage, and output alarm and location results.

Claims

1. A method for early warning of physical homogeneous multi-source acoustic trends, characterized in that, include: S1. Construction and high-fidelity calibration of the sound transmission model: (a) Based on the physical structure, material acoustic impedance and sound wave propagation characteristics of the monitored equipment, an initial theoretical sound transfer function is constructed to characterize the sound transmission relationship from the sound source to each acoustic monitoring point, denoted as H_A(ω); (b) Synchronously acquire time-domain signals from each acoustic monitoring point and calculate the actual acoustic response function (ARF) and acoustic coherence coefficient (γ²(ω)) between each channel; (c) Select a frequency range in which the acoustic coherence coefficient γ²(ω) is greater than a preset first acoustic coherence threshold, and perform parameter calibration on the initial theoretical acoustic transfer function H_A(ω) based on the actual acoustic response function ARF corresponding to the range; The calibration process is a linear least squares solution process that minimizes the mean square error between the output of H_A(ω) at these frequency points and the ARF, thereby generating a calibrated acoustic transmission model that accurately characterizes the acoustic propagation characteristics of the device under specific conditions. S2. Based on the calibrated sound transmission model, extract at least one set of core acoustic physical feature parameters. The core acoustic physical feature parameters are parameters directly calculated or identified from the calibrated sound transmission model and are used as the benchmark for physical consistency verification in step S3. Input the core acoustic physical feature parameters into a preset machine learning or statistical model for sequence trend analysis and output the evolution trend of the set of parameters in a future preset time window. S3. In real time or near real time, based on the same calibrated acoustic transmission model generated in step S1 as the sole physical reference, calculate the current values ​​of the same set of core acoustic physical feature parameters extracted in step S2 according to the online acquired acoustic signals; compare the current values ​​with the future trend values ​​predicted in step S2; if the absolute deviation exceeds the preset acoustic consistency threshold, determine that the acoustic transmission consistency has deviated, and activate the graded warning according to the degree of deviation.

2. The method according to claim 1, characterized in that, The first acoustic coherence threshold is between 0.80 and 0.

95.

3. The method according to claim 1, characterized in that, The core acoustic physical characteristic parameters include: at least one acoustic modal frequency identified from the calibrated acoustic transmission model, the corresponding acoustic damping ratio, and the acoustic energy change rate.

4. The method according to claim 3, characterized in that, The core acoustic physical characteristic parameters also include at least one of the following: sound pressure gradient, equivalent acoustic impedance change, sound field uniformity index, and peak sound pressure.

5. The method according to claim 1, characterized in that, The trend prediction model includes one or more of the following: ARIMA, LSTM, GRU, Transformer, and time series regression.

6. The method according to claim 1, characterized in that, The tiered early warning system includes five levels: normal, alert, warning, alarm, and fault. When the level is "alarm" or "fault", at least one type of auxiliary information is also output, including fault location, sound source risk level, and equipment remaining service life prediction.

7. A physical homogeneous multi-source acoustic trend early warning system, characterized in that, include: The acoustic transmission modeling and calibration module is used to construct and solve the calibration theoretical acoustic transmission function H_A(ω) by minimizing the mean square error using linear least squares, and generate the calibrated acoustic transmission model. A multi-channel synchronous acoustic acquisition and preprocessing module is used for synchronous acquisition of acoustic signals, noise reduction, time delay correction and time alignment. The acoustic feature extraction and trend prediction module is used to extract the acoustic physical feature parameter set and output the future evolution trend; The acoustic consistency verification and early warning module is used to complete feature comparison, consistency judgment and graded early warning output based on the same calibrated acoustic transmission model as the sole physical benchmark.

8. A physical homogeneous multi-source acoustic trend early warning platform, characterized in that, The system includes the system described in claim 7, and is further equipped with a visualization display, user interaction, data storage, sound source tracing and report generation module.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.

Images