Sound wave monitoring system based on optical fiber sensing

Abstract

The invention discloses a sound wave monitoring system based on optical fiber sensing, and relates to the technical field of optical fiber sensing sound wave monitoring, the sound wave monitoring system comprises a plurality of modules, the optical fiber sensing module adopts a novel microstructure optical fiber with a unique porous structure, the surface of the optical fiber sensing module is coated with an intelligent composite material, and the optical fiber sensing module is filled with nanofluid to enhance response to sound waves; the signal conversion module combines various photoelectric conversion advantages and dynamically switches modes according to light intensity, the signal processing module adopts a hybrid filtering and adaptive quantification technology, the data analysis module is a network based on deep learning and integrates multi-scale features, the early warning module dynamically warns early according to risk assessment and integrates multiple factors to adjust thresholds and levels, and all the modules cooperate to achieve the purpose of early warning. And efficient monitoring, analysis and early warning of sound wave signals are realized. The sensor is excellent in performance, high in sensitivity and wide in dynamic range; intelligent and accurate analysis and early warning can be realized, sound wave characteristics are judged, and timely warning is realized; data storage is safe and efficient, and an advanced and reliable scheme is provided for sound wave monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of optical fiber sensing sound wave monitoring, and in particular to a sound wave monitoring system based on optical fiber sensing. Background Art

[0002] As industries evolve, acoustic wave monitoring is becoming increasingly important as a means of obtaining environmental and equipment information. However, traditional acoustic wave monitoring methods have numerous limitations and are unable to meet the complex and diverse needs of today's applications.

[0003] Traditional acoustic wave monitoring equipment is mostly based on conventional technologies such as piezoelectric sensors. These devices have poor sensitivity and limited ability to capture weak acoustic signals. For example, in the early stages of earthquake monitoring, tiny seismic wave signals are difficult to accurately detect, resulting in delayed early warnings and the potential loss of optimal response opportunities. Furthermore, their narrow dynamic range makes them prone to saturation distortion when faced with strong acoustic signals, preventing them from fully recording acoustic information. Furthermore, traditional equipment has weak anti-interference capabilities. In complex environments, such as industrial sites with extensive electromagnetic interference, the accuracy of monitoring data can be severely affected, making it difficult for monitoring results to truly reflect actual conditions.

[0004] Furthermore, with the advent of the Internet of Things and big data, various industries are increasingly demanding the integration and intelligence of acoustic wave monitoring systems. However, existing acoustic wave monitoring systems often operate in isolation, making it difficult to effectively integrate with other systems. In industrial production, equipment management systems and acoustic wave monitoring systems lack data exchange, making it impossible to timely adjust production plans and maintenance strategies based on the acoustic wave signals generated by equipment during operation, resulting in low production efficiency and increased equipment maintenance costs. Furthermore, traditional systems lack intelligent analysis capabilities and simply record acoustic wave data, failing to extract valuable information from massive amounts of data. This makes it difficult to accurately determine the source, nature, and development trends of acoustic wave signals, and thus fails to provide strong support for decision-making.

[0005] Specialized fields, such as ocean acoustic monitoring and underground resource exploration, place higher demands on the performance of acoustic wave monitoring systems. Traditional systems, due to technical bottlenecks, are unable to adapt to complex environmental conditions and diverse monitoring needs. Complex ocean hydrological conditions and the heterogeneity of underground media pose significant challenges to acoustic wave monitoring. Therefore, there is an urgent need for innovative acoustic wave monitoring systems that can overcome the limitations of traditional technologies and achieve high sensitivity, a wide dynamic range, strong anti-interference capabilities, and intelligent monitoring and analysis to meet the growing application needs of various fields. Summary of the Invention

[0006] The present invention proposes an acoustic wave monitoring system based on optical fiber sensing to solve the problems mentioned in the above-mentioned prior art.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] An acoustic wave monitoring system based on optical fiber sensing includes the following modules:

[0009] Fiber optic sensing module: It uses microstructured optical fiber with a porous structure inside, and coats the surface of the optical fiber with intelligent composite materials. The material properties are adjusted according to the formula: E adjust =E0+k1·A+k2·f, where E adjust is the adjusted elastic modulus, E0 is the initial elastic modulus, A is the acoustic amplitude, f is the acoustic frequency, k1 and k2 are the material property adjustment coefficients;

[0010] Signal conversion module: Using multi-modal photoelectric conversion technology, combined with PIN photodiodes and avalanche photodiodes (APDs), the conversion mode is dynamically switched according to the intensity of the optical signal. The formula is: Among them G switch is the gain after switching, G APD is the APD gain, G PIN is the PIN photodiode gain, P is the input optical power, P threshold is the power threshold;

[0011] Signal processing module: A hybrid filtering algorithm is used to combine the traditional bandpass filter with the wavelet transform filter. The decomposition and reconstruction process of the wavelet transform is based on the improved Mallat algorithm, and its iterative formula is: c j+1 [n]=∑ k h[k-2n]c j [k], d j+1 [n]=∑ k g[k-2n]c j [k], where c j [n] is the approximate coefficient, d j [n] is the detail coefficient, h[k] and g[k] are the wavelet filter coefficients;

[0012] Data analysis module: A multi-scale feature fusion network based on deep learning is introduced to analyze acoustic wave signals. This network combines the advantages of convolutional neural networks and long short-term memory networks, and fuses features of different scales through a multi-scale feature fusion layer. The network loss function adopts an improved cross-entropy loss function: where y ij is the true label, p ij is the predicted probability, N is the number of samples, M is the number of categories, λ is the regularization coefficient, W l is the weight matrix;

[0013] Early warning module: Adopts a dynamic early warning strategy based on risk assessment and quantifies risks through fuzzy comprehensive evaluation method. The formula is: Where R is the risk assessment value, w i is the weight, μ i For membership, when the risk assessment value exceeds a certain threshold, an early warning is triggered and relevant personnel are notified.

[0014] Furthermore, a calibration module is included. This module adopts an adaptive calibration method based on a virtual reference signal. A virtual reference signal is generated during system operation. By comparing the difference between the monitoring signal and the virtual reference signal, the system parameters are automatically adjusted using an adaptive algorithm. The adaptive adjustment formula is: where θ n is the parameter vector of the nth iteration, μ is the learning rate, J(θ n ) is the loss function. At the same time, a calibration history database is established to record and analyze the parameters and results of each calibration.

[0015] Furthermore, it also includes a data storage module, which performs preliminary data processing on the edge device at the monitoring site, extracts key features, and then encrypts and stores the processed data on distributed storage nodes. Blockchain technology is used to achieve data access control and sharing through smart contracts. The redundancy of data storage is dynamically adjusted according to the importance and real-time nature of the data. The formula is: R redundancy =α·I+β·T, where P redundancy is the redundancy, I is the data importance index, T is the data real-time index, and α and β are adjustment coefficients.

[0016] Furthermore, in the optical fiber sensing module, nanofluid with acoustic amplification function is filled in the porous structure of the optical fiber. When the sound wave acts on the optical fiber, the nanoparticles in the nanofluid vibrate and scatter, enhancing the light-acoustic interaction inside the optical fiber. The relationship between the acoustic amplification factor, the nanoparticle concentration and the sound wave frequency follows the formula: A amplify =A0·(1+k·C·f), where A amplify is the amplified sound wave amplitude, A0 is the initial sound wave amplitude, C is the nanoparticle concentration, f is the sound wave frequency, and k is the amplification factor.

[0017] Furthermore, in the signal conversion module, a light-to-electric conversion enhancement structure based on photonic crystals is adopted. This structure is composed of photonic crystals and photoelectric conversion materials. By optimizing the structural parameters of the photonic crystals, the conversion of light signals of different frequencies is achieved. The relationship between light absorption efficiency and photonic crystal structural parameters is fitted through numerical simulation and experimental data, and the empirical formula is obtained: η absorb =a·L+b·r+c, where η absorbis the light absorption efficiency, L is the lattice constant, r is the filling ratio, and a, b, and c are fitting coefficients.

[0018] Furthermore, in the signal processing module, an adaptive noise cancellation technology based on genetic algorithm optimization is adopted. This technology searches for the optimal filter coefficient through the genetic algorithm. The fitness function of the genetic algorithm is defined as the inverse of the sum of squares of the errors between the monitoring signal and the noise estimation signal. The population is evolved through selection, crossover and mutation operations. The fitness function formula is: Where F is the fitness value, s i To monitor the signal, is the noise estimation signal, and N is the signal length.

[0019] Furthermore, in the data analysis module, a knowledge fusion method based on transfer learning is adopted. When faced with a new monitoring scenario, the existing relevant model knowledge is used for transfer learning, and the model parameters are fine-tuned to adapt to the new task. At the same time, the knowledge of different data sources is integrated, and the weight distribution of knowledge fusion is dynamically adjusted according to the reliability and relevance of the data source. The formula is: where w source is the data source weight, R source is the reliability index of the data source, C source is the correlation index between the data source and the current task, and m is the number of data sources.

[0020] Furthermore, in the early warning module, a collaborative early warning mechanism based on multi-sensor fusion is adopted to process and analyze multi-sensor data through a data fusion algorithm. The data fusion algorithm adopts the Kalman filter algorithm, and its state prediction equation is: The update equation is: in is the state estimate, F is the state transfer matrix, B is the control input matrix, u is the control input, z is the measurement value, H is the measurement matrix, and K is the Kalman gain.

[0021] Furthermore, in the calibration module, a real-time calibration method based on online learning is adopted. During the operation of the system, new calibration data is collected and the calibration model is updated in real time using the online learning algorithm. The online learning algorithm adopts the stochastic gradient descent algorithm, and its parameter update formula is: where θ n is the parameter vector of the nth iteration, μ is the learning rate, L is the loss function, x n is the nth calibration data sample, y n is the corresponding true value.

[0022] Furthermore, in the data storage module, an intelligent cache is set up on the edge device to store data in a hierarchical manner according to the access frequency and importance of the data. The cache replacement strategy adopts the least recently used (LRU) algorithm to improve the access efficiency of the data. The relationship between the cache hit rate, cache size and data access pattern is obtained through simulation experiments. The empirical formula: H cache =α·S cache +b·M access +c, where H cache is the cache hit rate, S cache is the cache size, M access is the data access pattern indicator, and a, b, and c are fitting coefficients.

[0023] Compared with the existing technology, the beneficial effects of the present invention are:

[0024] In terms of monitoring performance, the system demonstrates exceptional sensitivity and a wide dynamic range. The combination of novel microstructured optical fibers and intelligent composite materials, along with the acoustic amplification capabilities of nanofluids, significantly enhances the ability to capture weak acoustic signals, enabling precise detection of even the slightest acoustic changes. Furthermore, the system operates stably even in the presence of strong acoustic signals, without saturation or distortion, and is able to fully record acoustic information. Furthermore, its anti-interference capabilities are significantly enhanced. The unique optical fiber structure and signal processing technology effectively resist external electromagnetic interference, ensuring accurate and reliable monitoring data even in complex environments.

[0025] The system achieves intelligent and precise data analysis and early warning capabilities. A multi-scale feature fusion network based on deep learning can deeply analyze acoustic signals, accurately identify signal characteristics, and quickly determine the source, nature, and development trend of the sound waves. A dynamic early warning strategy based on risk assessment and a collaborative early warning mechanism using multi-sensor fusion comprehensively consider multiple factors to issue timely and accurate warning information, enabling relevant personnel to prepare for response and effectively reducing losses caused by disasters and accidents.

[0026] In terms of system integration and data management, the architecture combining distributed encrypted storage with edge computing enables efficient data storage and secure management. The application of blockchain technology ensures data immutability and security, and intelligent caching strategies improve data access efficiency. Furthermore, the system boasts excellent integration, enabling data exchange and collaboration with other systems to meet the diverse needs of various industries. The innovative approach of the calibration module ensures the accuracy and stability of the system's long-term operation, providing strong support for the reliability of monitoring results. In short, this system provides an advanced and reliable solution for the acoustic wave monitoring needs of various industries, promoting the development and application of acoustic wave monitoring technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic block diagram of the acoustic wave monitoring system based on optical fiber sensing proposed in the present invention;

[0028] Figure 2 The bar graph is a comparison of the sensitivity of the traditional monitoring system and this system;

[0029] Figure 3 This is a line graph showing the changes in the early warning accuracy of the system during five consecutive operation periods;

[0030] Figure 4 A pie chart showing the proportion of different types of data (raw monitoring data, processed data, calibration data, warning data, and other data) stored in this system within a certain period of time. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0033] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0034] Reference Figures 1 to 4 :A sound wave monitoring system based on optical fiber sensing, including the following modules:

[0035] Fiber optic sensing module: It uses an innovative microstructured optical fiber with a unique and sophisticated porous structure system built inside. These porous structures are not randomly distributed, and their pore sizes range from nanometers to micrometers. They are manufactured through precise photolithography and etching processes, forming a highly ordered array distribution. Based on advanced fluid-solid coupling theory, the size, shape and spacing of the porous structure are repeatedly optimized and simulated. By combining numerical calculations and experimental verification, the structural parameters are precisely adjusted to ensure that the interaction between the optical fiber and the sound waves can be maximized under different environments, significantly improving the sensitivity to sound waves of various frequencies. At the same time, the smart composite material coated on the surface of the optical fiber is prepared by a special blending process of a variety of functional nanoparticles and polymers. These nanoparticles include piezoelectric nanoparticles, magnetostrictive nanoparticles, etc., which are uniformly dispersed in a polymer matrix. The composite material has adaptive acoustic properties. When external sound waves act, the internal nanoparticles will interact with the sound waves. Its elastic modulus and damping coefficient can be dynamically adjusted according to the intensity and frequency of the external sound waves, following the precisely derived formula E adjust =E0+k1·A+k2·f, further enhancing the optical fiber's ability to respond to sound waves and achieving efficient capture and accurate perception of sound wave signals, where E adjust is the adjusted elastic modulus, E0 is the initial elastic modulus, A is the acoustic wave amplitude, f is the acoustic wave frequency, and k1 and k2 are material property adjustment coefficients.

[0036] Signal conversion module: It uses cutting-edge multi-modal photoelectric conversion technology and deeply integrates the unique advantages of PIN photodiodes and avalanche photodiodes (APDs). This technology monitors the intensity of optical signals in real time through high-precision signal detection circuits. In hardware design, PIN photodiodes with ultra-low noise characteristics are carefully selected. They are made of high-purity semiconductor materials. Through advanced epitaxial growth processes, the impurity concentration and crystal structure of the materials are precisely controlled. When strong light signals are input, they can achieve high-precision linear conversion by virtue of their own good linearity. For avalanche photodiodes (APDs), a special semiconductor structure design is adopted, and a multi-layer heterojunction is constructed inside them. By precisely controlling the thickness and doping concentration of each layer, it can utilize the avalanche multiplication effect to achieve high-gain conversion under weak light signal conditions. The system monitors the input optical power P in real time and compares it with the pre-set power threshold P threshold The gain adjustment during the conversion process follows the following formula: Among them G switch is the gain after switching, G APDis the gain of the APD, G PIN is the gain of the PIN photodiode. <P threshold , that is, when the optical signal is weak, it automatically switches to the APD working mode, using its high gain characteristic G APD Amplify and convert the optical signal; when P≥P threshold When the optical signal is strong, it switches to the PIN photodiode working mode, using its linear conversion characteristic G PIN , ensuring high-precision conversion and effectively expanding the dynamic range of signal conversion.

[0037] Signal processing module: It uses an innovative hybrid filtering algorithm that organically combines the advantages of traditional bandpass filters and wavelet transform filters. The bandpass filter is constructed using high-quality electronic components. The center frequency and bandwidth of the filter are precisely set according to the expected frequency range of the signal. It can effectively filter out interference signals outside the frequency band and achieve preliminary frequency screening. The decomposition and reconstruction process of the wavelet transform is based on the improved Mallat algorithm, and its iterative formula is:

[0038] c j+1 [n]=∑ k h[k-2n]c j [k],

[0039] d j+1 [n]=∑ k g[k-2n]c j [k], where c j [n] is the approximate coefficient, d j [n] is the detail coefficient, h[k] and g[k] are the wavelet filter coefficients. After multi-stage amplification, the signal is quantized using adaptive quantization technology, which dynamically adjusts the quantization step size based on the statistical characteristics of the signal to reduce quantization error.

[0040] The data analysis module introduces a deep learning-based Multi-Scale Feature Fusion Network (MSFF-Net) to analyze acoustic signals. This network combines the advantages of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). The CNN employs a carefully customized multi-layer convolutional architecture. The convolution kernel sizes range from 3×3 to 5×5, optimally configured based on the frequency characteristics and local details of the acoustic signal. Through convolution operations at different levels, it can keenly capture local characteristics of acoustic signals in the time and frequency domains, such as signal intensity variations within specific frequency bands and subtle fluctuations in the waveform. Batch normalization is applied after each convolution layer to accelerate network convergence and improve training stability. The LSTM is designed to accurately capture the temporal characteristics of acoustic signals. Its internal memory units are meticulously tuned, and the weight parameters of the input, forget, and output gates are continuously optimized through training on large amounts of acoustic data. It can effectively process long-term temporal dependencies of signals and accurately grasp information such as the onset, duration, and periodicity of the signal. The multi-scale feature fusion layer is used to fuse features of different scales to improve the classification and recognition capabilities of sound wave signals. The network loss function uses the improved cross entropy loss function: where y ij is the true label, p ij is the predicted probability, N is the number of samples, M is the number of categories, λ is the regularization coefficient, W l is the weight matrix of the network layer l.

[0041] Early warning module: Adopting a dynamic early warning strategy based on risk assessment, a risk assessment model is constructed by comprehensively and meticulously considering multiple factors. For the intensity of the sound wave signal, high-precision sensors are used for real-time monitoring, and different intervals are divided according to the decibel level; in terms of frequency, algorithms such as fast Fourier transform are used for precise analysis to determine whether it is in the low frequency, medium frequency or high frequency range. The duration is accurately calculated by recording the start and end time with a timestamp. The time and location information is obtained by the supporting time synchronization system and positioning device. When using the fuzzy comprehensive evaluation method to quantify the risk, according to the formula Calculate the risk assessment value R,w i is the weight of the i-th factor, μ i The membership degree of the i-th factor is dynamically adjusted according to the risk assessment value. When the risk assessment value exceeds a certain threshold, an advanced warning is triggered and relevant personnel are notified in a timely manner using various methods (such as SMS, voice, image, etc.).

[0042] The present invention also includes a calibration module. This module uses an adaptive calibration method based on a virtual reference signal. During system operation, a virtual reference signal is generated in real time, which simulates the characteristics of a standard acoustic signal. By comparing the difference between the monitoring signal and the virtual reference signal, an adaptive algorithm is used to automatically adjust system parameters such as gain and filter coefficient. The adaptive adjustment formula is: where θ n is the parameter vector of the nth iteration, μ is the learning rate, J(θ n ) is a loss function used to measure the difference between the monitoring signal and the virtual reference signal. At the same time, a calibration history database is established to record and analyze the parameters and results of each calibration in order to optimize the calibration strategy.

[0043] The present invention also includes a data storage module. This module adopts an architecture that combines distributed encrypted storage with edge computing. Preliminary data processing and analysis are performed on the edge devices at the monitoring site to extract key features, and then the processed data is encrypted and stored on the distributed storage nodes. Blockchain technology is used to ensure the security and non-tamperability of data, and data access control and sharing are achieved through smart contracts. The redundancy of data storage is dynamically adjusted according to the importance and real-time nature of the data, and the formula is: R redundancy =α·I+β·T, where R redundancy is the redundancy, I is the data importance index, T is the data real-time index, and α and β are adjustment coefficients.

[0044] In the present invention, a nanofluid with acoustic amplification function is filled into the porous structure of the optical fiber in the optical fiber sensing module. The nanofluid is composed of nanoparticles and a base liquid. The type and concentration of the nanoparticles are optimized to enhance the amplification effect of the sound waves. When the sound waves act on the optical fiber, the nanoparticles in the nanofluid vibrate and scatter, enhancing the light-acoustic interaction within the optical fiber. The relationship between the acoustic amplification factor, the nanoparticle concentration and the sound wave frequency follows the formula: amplify =A0·(1+k·C·f), where A amplify is the amplified sound wave amplitude, A0 is the initial sound wave amplitude, C is the nanoparticle concentration, f is the sound wave frequency, and k is the amplification factor.

[0045] In the present invention, a photonic crystal-based light-to-electric conversion enhancement structure is used in the signal conversion module. This structure is composed of a photonic crystal and a photoelectric conversion material. The photonic crystal has a special photonic bandgap characteristic that can localize light of a specific frequency in the photoelectric conversion material, improving the light absorption efficiency and conversion efficiency. By optimizing the structural parameters of the photonic crystal, such as the lattice constant and fill ratio, efficient conversion of light signals of different frequencies is achieved. The relationship between light absorption efficiency and photonic crystal structural parameters can be fitted through numerical simulation and experimental data to obtain the empirical formula: η absorb =a·L+b·r+c, where η absorb is the light absorption efficiency, L is the lattice constant, r is the filling ratio, and a, b, and c are fitting coefficients.

[0046] In the present invention, an adaptive noise cancellation technique based on genetic algorithm optimization is used in the signal processing module. This technique starts with the random generation of an initial filter coefficient population, which determines the filter's ability to suppress noise of different frequencies. The core of the genetic algorithm lies in the fitness function, which is defined as Acquire monitoring signals through high-precision sensors and advanced signal acquisition circuits i , using the noise estimation model based on statistical analysis and machine learning to obtain the noise estimation signal N is determined by both the duration of signal acquisition and the sampling frequency. During the algorithm's iterations, the selection operation uses strategies such as roulette wheel selection to select high-performing individuals based on their fitness, increasing the probability that filter coefficients with high fitness will be adopted into the next generation. The crossover operation, through single-point and multi-point crossover methods, exchanges gene fragments between individuals, generating new, potentially superior combinations. The mutation operation randomly alters some of an individual's genes with a low probability to maintain population diversity. By repeatedly repeating these operations, the population evolves until the optimal filter coefficients are found, effectively canceling out noise.

[0047] In the present invention, a knowledge fusion method based on transfer learning is adopted in the data analysis module. When faced with new monitoring scenarios or data types, existing relevant model knowledge is used for transfer learning, and the model parameters are fine-tuned to quickly adapt to new tasks. At the same time, the knowledge of different data sources is integrated, such as historical monitoring data, expert experience knowledge, etc., to improve the analysis and understanding of acoustic signals. The weight distribution of knowledge fusion is dynamically adjusted according to the reliability and relevance of the data source, and the formula is: where w source is the weight of the data source, R source is the reliability index of the data source, C source is the correlation index between the data source and the current task, and m is the number of data sources.

[0048] In this invention, the early warning module adopts a collaborative early warning mechanism based on multi-sensor fusion. In addition to the acoustic wave signals monitored by the fiber optic sensing module, data from other types of sensors, such as vibration sensors and pressure sensors, are also integrated. The multi-sensor data is processed and analyzed through a data fusion algorithm to improve the accuracy and reliability of the early warning. The data fusion algorithm adopts the Kalman filter algorithm, and its state prediction equation is: The update equation is: in is the state estimate, F is the state transfer matrix, B is the control input matrix, u is the control input, z is the measurement value, H is the measurement matrix, and K is the Kalman gain.

[0049] In the present invention, the calibration module adopts a real-time calibration method based on online learning to ensure the long-term stable and accurate operation of the system. When the system is running, the data acquisition unit uses a high-precision sensor as the front end and cooperates with a high-speed data transmission line to continuously collect various new calibration data. These data cover a variety of parameters in the operation of the system, such as the impact of ambient temperature and humidity on the measurement results. The online learning algorithm adopted is the stochastic gradient descent algorithm, which is targeted and optimized. The learning rate μ is not a fixed value, but is dynamically adjusted according to the change of the loss function of each iteration through an adaptive adjustment strategy. The loss function L is constructed by comprehensively considering multiple error factors, such as the absolute error and relative error between the measured value and the true value. In each iteration, according to the nth calibration data sample x n and its corresponding true value y n , according to the parameter update formula The parameter vector θ for the nth iteration n Updates are made to correct the calibration model in real time, effectively reducing errors and ensuring the accuracy and stability of the system in complex and changing environments.

[0050] In the present invention, an efficient data storage strategy based on intelligent caching is adopted in the data storage module. An intelligent cache is set on the edge device to store data in a hierarchical manner according to the access frequency and importance of the data. Frequently accessed and important data are preferentially stored in the cache; infrequently used data are stored on distributed storage nodes. The cache replacement strategy adopts the least recently used (LRU) algorithm to improve data access efficiency. The relationship between cache hit rate, cache size and data access pattern can be obtained through simulation experiments. The empirical formula: H cache =a·S cache +b·M access +c, where H cache is the cache hit rate, S cache is the cache size, M access is the data access pattern indicator, and a, b, and c are fitting coefficients.

[0051] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An acoustic wave monitoring system based on optical fiber sensing, characterized in that: Includes the following modules: Fiber optic sensing module: It uses microstructured optical fiber with a porous structure inside, and coats the surface of the optical fiber with intelligent composite materials. The material properties are adjusted according to the formula: E adjust =E0+k1·A+k2·f, where E adjust is the adjusted elastic modulus, E0 is the initial elastic modulus, A is the acoustic amplitude, f is the acoustic frequency, k1 and k2 are the material property adjustment coefficients; Signal conversion module: Using multi-modal photoelectric conversion technology, combined with PIN photodiodes and avalanche photodiodes (APDs), the conversion mode is dynamically switched according to the intensity of the optical signal. The formula is: Among them G switch is the gain after switching, G APD is the APD gain, G PIN is the PIN photodiode gain, P is the input optical power, P threshold is the power threshold; Signal processing module: A hybrid filtering algorithm is used to combine the traditional bandpass filter with the wavelet transform filter. The decomposition and reconstruction process of the wavelet transform is based on the improved Mallat algorithm, and its iterative formula is: c j+1 [n]=∑ k h[k-2n]c j [k], d j+1 [n]=∑ k g[k-2n]c j [k], where c j [n] is the approximate coefficient, d j [n] is the detail coefficient, h[k] and g[k] are the wavelet filter coefficients; Data analysis module: A multi-scale feature fusion network based on deep learning is introduced to analyze acoustic wave signals. This network combines the advantages of convolutional neural networks and long short-term memory networks, and fuses features of different scales through a multi-scale feature fusion layer. The network loss function adopts an improved cross-entropy loss function: where y ij is the true label, p ij is the predicted probability, N is the number of samples, M is the number of categories, λ is the regularization coefficient, W l is the weight matrix; Early warning module: Adopts a dynamic early warning strategy based on risk assessment and quantifies risks through fuzzy comprehensive evaluation method. The formula is: Where R is the risk assessment value, w i is the weight, μ i For membership, when the risk assessment value exceeds a certain threshold, an early warning is triggered and relevant personnel are notified.

2. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: The system also includes a calibration module, which uses an adaptive calibration method based on a virtual reference signal. A virtual reference signal is generated during system operation. By comparing the difference between the monitoring signal and the virtual reference signal, the system parameters are automatically adjusted using an adaptive algorithm. The adaptive adjustment formula is: where θ n is the parameter vector of the nth iteration, μ is the learning rate, J(θ n ) is the loss function. At the same time, a calibration history database is established to record and analyze the parameters and results of each calibration.

3. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: It also includes a data storage module, which performs preliminary data processing on the edge device at the monitoring site, extracts key features, and then encrypts and stores the processed data on distributed storage nodes. Blockchain technology is used to achieve data access control and sharing through smart contracts. The redundancy of data storage is dynamically adjusted according to the importance and real-time nature of the data. The formula is: R redundancy =α·I+β·T, where R redundancy is the redundancy, I is the data importance index, T is the data real-time index, and α and β are adjustment coefficients.

4. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: In the optical fiber sensing module, nanofluid with acoustic amplification function is filled in the porous structure of the optical fiber. When the sound wave acts on the optical fiber, the nanoparticles in the nanofluid vibrate and scatter, enhancing the light-acoustic interaction inside the optical fiber. The relationship between the acoustic amplification factor, the nanoparticle concentration and the sound wave frequency follows the formula: A amplify =A0·(1+k·C·f), where A amplify is the amplified sound wave amplitude, A0 is the initial sound wave amplitude, C is the nanoparticle concentration, f is the sound wave frequency, and k is the amplification factor.

5. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: In the signal conversion module, a light-to-electric conversion enhancement structure based on photonic crystals is adopted. This structure is composed of photonic crystals and photoelectric conversion materials. By optimizing the structural parameters of the photonic crystals, the conversion of light signals of different frequencies is realized. The relationship between light absorption efficiency and photonic crystal structural parameters is fitted through numerical simulation and experimental data, and the empirical formula is obtained: η absorb =a·L+b·r+c, where η absorb is the light absorption efficiency, L is the lattice constant, r is the filling ratio, and a, b, and c are fitting coefficients.

6. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: In the signal processing module, an adaptive noise cancellation technology based on genetic algorithm optimization is used. This technology searches for the optimal filter coefficients through the genetic algorithm. The fitness function of the genetic algorithm is defined as the inverse of the sum of squares of the errors between the monitoring signal and the noise estimation signal. The population is evolved through selection, crossover, and mutation operations. The fitness function formula is: Where F is the fitness value, s i To monitor the signal, is the noise estimation signal, and N is the signal length.

7. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: In the data analysis module, a knowledge fusion method based on transfer learning is adopted. When faced with a new monitoring scenario, the existing relevant model knowledge is used for transfer learning, and the model parameters are fine-tuned to adapt to the new task. At the same time, the knowledge of different data sources is integrated. The weight distribution of knowledge fusion is dynamically adjusted according to the reliability and relevance of the data source. The formula is: where w source is the data source weight, R source is the reliability index of the data source, C source is the correlation index between the data source and the current task, and m is the number of data sources.

8. The acoustic wave monitoring system based on optical fiber sensing according to claim 1, characterized in that: In the early warning module, a collaborative early warning mechanism based on multi-sensor fusion is adopted to process and analyze multi-sensor data through a data fusion algorithm. The data fusion algorithm adopts the Kalman filter algorithm, and its state prediction equation is: The update equation is: in is the state estimate, F is the state transfer matrix, B is the control input matrix, u is the control input, z is the measurement value, H is the measurement matrix, and K is the Kalman gain.

9. The acoustic wave monitoring system based on optical fiber sensing according to claim 2, characterized in that: In the calibration module, a real-time calibration method based on online learning is adopted. During the operation of the system, new calibration data is collected and the calibration model is updated in real time using the online learning algorithm. The online learning algorithm adopts the stochastic gradient descent algorithm, and its parameter update formula is: where θ n is the parameter vector of the nth iteration, μ is the learning rate, L is the loss function, x n is the nth calibration data sample, y n is the corresponding true value.

10. The acoustic wave monitoring system based on optical fiber sensing according to claim 3, characterized in that: In the data storage module, an intelligent cache is set up on the edge device to store data in a hierarchical manner according to the access frequency and importance of the data. The cache replacement strategy adopts the least recently used (LRU) algorithm to improve the access efficiency of the data. The relationship between the cache hit rate, cache size and data access pattern is obtained through simulation experiments. The empirical formula is: H cache =a·S cache +b·M access +c, where H cache is the cache hit rate, S cache is the cache size, M access is the data access pattern indicator, and a, b, and c are fitting coefficients.

Images