A three-dimensional temperature field online reconstruction system based on transceiving integrated acoustic wave sensor

Abstract

The application discloses a kind of three-dimensional temperature field online reconstruction systems based on transceiving integrated acoustic wave sensing, it is related to boiler monitoring field, including: information acquisition module, launch control module, reconstruction sensing module, path optimization module, acoustic wave receiving module, limiting amplitude filter module, decoding separation module, feature extraction module, physical simulation module, multi-source fusion module, three-dimensional reconstruction module, condition monitoring module and online calibration module;The application can make transmitting signal still maintain high distinguishability in high noise, strong interference background, improve the reliability of subsequent inversion and data assimilation, make sensing network can self-optimization, improve observation accuracy and spatial coverage capability, realize wide dynamic range and high linearity, make simulation result more close to real working condition, significantly improve the reliability of temperature field inversion.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of boiler monitoring, in particular to a three-dimensional temperature field online reconstruction system based on transceiving integrated acoustic wave sensing. BACKGROUND

[0002] In industrial processes, the three-dimensional temperature distribution inside the room or cavity directly affects the heat transfer efficiency, combustion completeness, equipment life and safety. Traditional point or fixed array temperature sensing is often limited in high temperature, high noise, blocked or strong radiation environment: sparse point leads to low spatial resolution, fixed layout cannot cover blind area, and signal quality decreases in strong noise background. Acoustic temperature detection uses the physical relationship that sound speed changes with temperature, and can obtain volume temperature information through volume distribution measurement, becoming a complementary or alternative means. However, in actual industrial boilers and other complex media, the propagation of acoustic waves is seriously affected by non-uniform media, strong turbulence, multipath and high-intensity background noise, and traditional linear inversion and fixed hardware are difficult to meet the engineering needs in real-time and high-resolution scenarios. At the same time, the development of digital twin, data assimilation (such as ensemble Kalman filter) and deep learning provides a feasible path for real-time fusion of high-dimensional physical models and sparse / noisy observations. Reconfigurable micro-sensor array and advanced anti-noise receiving chain (including quantum limit amplitude idea or superconducting low-noise front end) make it possible to obtain cleaner and more informative acoustic data in extreme environments. Systematizing and engineering these technologies together can realize online and high-resolution reconstruction of three-dimensional temperature field in complex industrial cavities, thereby supporting process control, fault warning and energy efficiency optimization. Therefore, the present application proposes a three-dimensional temperature field online reconstruction system based on transceiving integrated acoustic wave sensing.

[0003] The existing three-dimensional temperature field online reconstruction system cannot maintain high distinguishability of the transmitted signal in a high noise and strong interference background, reduces the credibility of subsequent inversion and data assimilation, the sensing network cannot be self-optimized, reduces the observation accuracy and spatial coverage ability, the simulation result deviates too much from the real working condition, and reduces the reliability of temperature field inversion. Therefore, we propose a three-dimensional temperature field online reconstruction system based on transceiving integrated acoustic wave sensing. SUMMARY

[0004] The purpose of the present application is to solve the defects in the prior art and provide a three-dimensional temperature field online reconstruction system based on transceiving integrated acoustic wave sensing.

[0005] The application provides a three-dimensional temperature field online reconstruction system based on a transceiving integrated acoustic wave sensor, which comprises an information acquisition module, a transmission control module, a reconstruction sensing module, a path optimization module, an acoustic wave receiving module, an amplitude limiting filter module, a decoding separation module, a feature extraction module, a physical simulation module, a multi-source fusion module, a three-dimensional reconstruction module, a state monitoring module and an online calibration module.

[0006] The information acquisition module is used for acquiring the operation parameters of the inside and periphery of the boiler in real time.

[0007] The transmission control module is used for generating acoustic wave coding signals suitable for a high-noise environment.

[0008] The reconstruction sensing module adjusts each sensing node in a three-dimensional space through a micro-actuator to form a variable topology array.

[0009] The path optimization module designs an optimal acoustic wave propagation path and array configuration strategy according to real-time acoustic wave propagation data and the working condition of the boiler.

[0010] The acoustic wave receiving module is used for synchronously receiving multi-path acoustic wave signals from different directions, different codes and different frequency bands, and performing preliminary synchronization and buffering processing.

[0011] The amplitude limiting filter module is used for performing amplitude limiting filter processing on the received multi-path acoustic wave signals.

[0012] The decoding separation module is used for decoding the processed acoustic wave signals to separate out the propagation time delay and amplitude change information corresponding to the temperature.

[0013] The feature extraction module is used for processing the decoded acoustic wave signals and extracting the propagation time delay, attenuation coefficient, multi-path feature and phase offset information.

[0014] The physical simulation module is used for constructing a CFD-based real-time operation simulation model of the boiler.

[0015] The multi-source fusion module is used for fusing the volume temperature information measured by the acoustic wave, the prediction result of the digital twin and the sensing data.

[0016] The three-dimensional reconstruction module generates a real-time temperature field distribution according to the fused multi-source data.

[0017] The state monitoring module is used for displaying each item of data in a graphical manner and providing an alarm prompt and an abnormal area positioning.

[0018] The online calibration module is used for online correcting the position deviation of the sensing array, the error of the acoustic velocity model, the parameter drift of the digital twin and other data deviations, and continuously comparing the measured data and the predicted data.

[0019] Preferably, the specific steps of the transmission control module in generating the acoustic wave coded signal suitable for high-noise environments are as follows:

[0020] S1.1: Collect the background noise sample sequence measured by the receiver within the preset time interval before the controller sends the signal, then perform window function segmentation on the background noise sample sequence, and calculate the power spectrum estimate of each frame to obtain the noise power spectral density profile. Based on the spectrum recognition, identify the high-energy interference band in the noise power spectrum, and calculate the proportion of each frequency band in the total noise energy.

[0021] S1.2: Based on the proportion of each frequency band in the total noise energy, select multiple candidate sub-bands with a noise proportion lower than a preset threshold as priority transmission frequency bands, determine the center frequency and bandwidth in each candidate sub-band, calculate the linear frequency modulation slope of the spread spectrum chirp of each candidate sub-band, then count the duration of the linear frequency modulation signal to match the time domain processing window of the receiver, and then superimpose a pseudo-random phase sequence on each linear frequency modulation signal. At the same time, select the switching period and seed of the PRN according to the preset control strategy.

[0022] S1.3: When there are multiple transmitting units, the controller allocates the orthogonal coding or different PRN seeds for each transmitting unit according to the control strategy, and calculates the transmission weight of each transmitting unit to form the desired sound field directivity. Based on the transmission weight of each transmitting unit, the controller obtains the baseband signal transmitted by each transmitting unit in each sub-band. Based on the premise of meeting the upper limit of total transmission power, the controller allocates power to different sub-bands and different transmitting paths according to the control strategy, and adds corresponding controllable frequency hopping or phase hopping modes to each baseband signal according to the PRN seed of each transmitting unit.

[0023] S1.4: Select the window type according to the control strategy, and perform bandpass filtering and amplitude and phase pre-distortion compensation on each baseband signal after the window. Then, embed the PRN seed, packet number, current subband index table and other control metadata into the header of each baseband signal, and set the verification and confirmation mechanism for each control metadata.

[0024] Preferably, the specific calculation formula for the power spectrum estimation in S1.1 is as follows:

[0025] ;

[0026] In the formula, The noise power spectral density function obtained from the estimation is represented by this function. Representative moment Time Received noise time-domain sample of the frame; The Fourier transform operator representing the time domain to the frequency domain returns the frequency. Complex value spectrum at; The number of frames used for averaging.

[0027] Preferably, the path optimization module designs the optimal sound wave propagation path and the specific steps of the array configuration strategy as follows:

[0028] S2.1: Real-time acquisition of original sound wave measurement data from the transmitting or receiving end, including time-domain sampling sequence and sampling time stamp of each receiving channel, while obtaining current combustion rate, flue gas flow rate, and overall pressure from the working condition sensor, and interpolating various data according to a unified time reference to obtain time-synchronized observation data;

[0029] S2.2: Time delay estimation, angle of arrival estimation, spectral density estimation, and instantaneous signal-to-noise ratio calculation are performed on the observation data to generate physically interpretable observation features, which are then standardized and concatenated into candidate observation sub-vectors according to the receiving channel, and the state features are combined according to a fixed order, and the current working condition and array topology description are added, and principal component analysis is used to reduce the dimensionality of the features;

[0030] S2.3: Design the action space for each movable sensor node, and map each action vector in the action space to the instruction set executable by each movable sensor node, and add corresponding physical constraints to each action vector in the instruction set according to known physical rules, and then calculate the signal quality gain, path propagation attenuation reduction, and coverage index at the next time after the execution of each action vector in the instruction set, and then weighted sum the calculated indexes to establish a reward function, calculate the immediate reward corresponding to the current action vector, and dynamically update the index weight value;

[0031] S2.4: Input the reduced state features into the pre-trained policy network, and output the action distribution at the current time through forward reasoning of the policy network, and analyze the execution of each action, convert the action distribution output by the policy network into specific driving commands, and issue displacement commands to the corresponding movable sensor nodes, and send frequency band, phase, or power allocation instructions to the transmitting unit;

[0032] S2.5: After starting the action, collect a new round of observations at the next time, calculate the immediate reward of the current action, and store the current state, action, immediate reward, and next state as an experience set in the experience pool, randomly select multiple samples from the experience pool at regular intervals, calculate the TD target value of each sample according to the selected DRL algorithm, obtain the policy network loss value based on each TD target value, and update the network parameters using the gradient descent algorithm.

[0033] Preferably, the specific calculation formula of the reward function in S2.3 is as follows:

[0034] ;

[0035] wherein: ;

[0036] wherein, represents the instant reward at time ; represents the change of average SNR before and after the action; represents the average attenuation change; represents the coverage increment; represents the action cost term, wherein, represents the cost weight, represents the action resource consumption; represents the weight of the first term index at time ; represents the time window length; represents the number of receiving channels; represents the SNR value of the first receiving channel at time ; represents the SNR value of the first receiving channel at time ; represents the number of rays being evaluated; represents the propagation loss of the first ray at time ; represents the propagation loss of the first ray at time ; represents the set of spatial grid points considered to be detectable at time ; represents the set of spatial grid points considered to be detectable at time ; represents the total number of grid points; represents the cardinality of the set.

[0037] Preferably, the current working condition quantity described in S2.2 specifically includes a temperature probe scalar and a combustion rate, etc.; and the array topology description specifically includes node positions and orientations.

[0038] The action space described in S2.3 specifically includes displacement of a movable sensing node and adjustment of an orientation, and main pointing angle and frequency band selection of a transmitting beam, transmitting power distribution ratio, etc.

[0039] Preferably, the specific steps of the limiting filter module for limiting filter processing of the received multiple sound wave signals are as follows:

[0040] S3.1: Put the external superconducting limiter into the cryogenic cold stage, start cooling and continuously monitor the critical temperature margin of the device, after the temperature reaches the preset target working point, wait for thermal equilibrium and measure the static transmission characteristics of the device to confirm the superconducting state, then apply a direct current bias current or magnetic flux, adjust the working point to make the nonlinear threshold of the superconducting limiter located in the required instantaneous input amplitude interval, and record the bias-response curve at the same time;

[0041] S3.2: Establish an adjustable matching network by fine-tuning inductance, capacitance or programmable matching, and connect it with the input end of the superconducting limiter, match the source impedance to the input impedance of the superconducting limiter to the preset standing wave ratio range through the adjustable matching network, record the rise time and saturation level of the transient response of the superconducting limiter when strong interference occurs, and monitor the recovery time of the superconducting limiter to the linear working area in real time after the strong interference;

[0042] S3.3: Quantitatively process the compression effect of the superconducting limiter on the input noise power spectrum of various data, and obtain the equivalent noise compression ratio of the superconducting limiter in different input power intervals by measurement, then calculate the equivalent input reference noise after the superconducting limiter using the equivalent noise compression factor, and evaluate the net benefit to the back-end signal processing;

[0043] S3.4: Restore the limited data of various types to the preset dynamic range and phase response through the linearization corrector, measure and calibrate the noise baseline of the output under different biases and environments, then filter the corrected data of various types through the adaptive Wiener filter to suppress residual background noise, and based on the adaptive notch filter and combined with phase compensation, deeply suppress the residual interference after the filtering is completed;

[0044] S3.5: Real-time calculation of the instantaneous variance, residual noise band energy and interference residual index based on discrimination of the adaptive Wiener filter output data, and based on the indicators, adjust the bias or bypass shunt factor of the superconducting limiter and update the noise spectrum estimation of the digital filter.

[0045] Preferably, the specific steps of the physical simulation module constructing the CFD-based real-time operation simulation model of the boiler are as follows:

[0046] S4.1: Obtain the three-dimensional geometric information inside the boiler by field measurement, preprocess the collected three-dimensional geometric information, and use finer grids in the wall neighborhood, nozzle inlet and outlet, flue corner and expected sound wave path crossing area, and use coarser grids in the remaining areas to generate a mixed grid and evaluate the grid quality, and output the grid file based on the evaluation results and the mixed grid structure;

[0047] S4.2: Collect working condition sensing data, and construct boundary condition distribution through spatial interpolation method, then input the boundary field obtained by interpolation into CFD boundary file, set as time-dependent boundary at the same time, and establish uncertainty estimation for mapping error, then convert each acoustic sensing data into integral constraint on sound speed field or temperature field, and apply each integral constraint as soft constraint on corresponding grid point of CFD, to establish a complete boiler operation simulation model;

[0048] S4.3: According to the relationship of gas thermodynamics, the sound speed and temperature of each local area of the operation simulation model are associated, and the data constraint term of temperature or thermodynamic quantity is established, then according to the discrete temperature, speed and pressure field on the current grid, the corresponding state set is established, and the numerical prediction is carried out for each state, when the new observation data is input, the gain is calculated and each state is updated by using the analysis step of EnKF, and the analyzed state set is generated, to obtain the real-time prediction value of the boiler temperature.

[0049] The beneficial effects of the present application are:

[0050] The application collects background noise before emission, frames and power spectrum estimates the noise sequence, identifies high-energy interference frequency bands, and selects multiple groups of low-noise subbands according to the energy proportion, then determines the center frequency, bandwidth and linear frequency modulation slope for each subband, sets the frequency modulation time length to match the receiving end processing window, embeds a pseudo-random phase sequence in the chirp signal of each subband, and allocates orthogonal codes, PRN seeds and transmission weights to each transmission unit according to the strategy, completes the joint configuration of frequency band, phase and power under the premise of meeting the total power constraint, then performs window function processing, band pass filtering and amplitude and phase pre-distortion compensation on each baseband signal, and writes control metadata such as PRN seed and subband index at the head of the signal, synchronously collects time domain data and working condition parameters of each receiving channel, and completes interpolation and synchronization on the unified time axis, then extracts time delay, angle of arrival, spectral density and instantaneous signal-to-noise ratio, standardizes them, and combines working condition parameters and array topology to construct state characteristics, establishes the action space for the movable sensing node after PCA dimensionality reduction, applies physical constraints to the action, calculates the improvement of each action on signal quality, attenuation and coverage, and constructs a reward function, inputs the state into the strategy network to obtain the action distribution, converts it into specific array movement and transmission strategy, forms a closed loop sampling-decision-execution process, and all states, actions and rewards enter the experience pool for TD update and network training of DRL, in the receiving link, the superconducting limiter is adjusted to the target working point through low-temperature debugging, the input impedance is adjusted through the matching network, and the nonlinear threshold, transient response, recovery time and noise compression ratio are measured, then the net signal-to-noise ratio is improved by using the compression effect of the limiter on the input noise, and the residual noise is suppressed through linearization correction, adaptive Wiener filtering and adaptive notch filtering, while the bias and filter parameters are updated in real time, the three-dimensional geometry of the boiler is constructed by field measurement and hybrid mesh is generated, the time-dependent boundary conditions are set by using the working condition data, and the acoustic observation is converted into integral soft constraints on sound speed or temperature and applied to the CFD model, the numerical prediction is performed according to the discrete temperature, pressure and velocity fields, when new observations are input, the state set is updated by using the EnKF analysis step to form a real-time coupled temperature field prediction, which can make the transmission signal still maintain high recognizability in the background of high noise and strong interference, improve the reliability of subsequent inversion and data assimilation, make the sensing network self-optimized, improve the observation accuracy and spatial coverage capability, realize wide dynamic range and high linearity, and make the simulation result closer to the real working condition, thereby significantly improving the reliability of temperature field inversion. BRIEF DESCRIPTION OF DRAWINGS

[0051] The application will be further described below in combination with the drawings.

[0052] Figure 1 It is a three-dimensional temperature field online reconstruction system framework based on transceiving integrated acoustic wave sensing. DETAILED DESCRIPTION

[0053] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0054] Based on the embodiments in the present application, all the other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0055] Embodiment 1 provides a three-dimensional temperature field online reconstruction system based on a transceiving integrated acoustic wave sensor. Figure 1 , Figure 1 A framework diagram of the three-dimensional temperature field online reconstruction system based on the transceiving integrated acoustic wave sensor is provided in the embodiments of the present application. The system comprises an information acquisition module, a transmission control module, a reconstruction sensing module, a path optimization module, an acoustic wave receiving module, a limiting amplitude filter module, a decoding separation module, a feature extraction module, a physical simulation module, a multi-source fusion module, a three-dimensional reconstruction module, a state monitoring module and an online calibration module.

[0056] The information acquisition module is used to acquire the running parameters of the inside and periphery of the boiler in real time; and the transmission control module is used to generate acoustic wave coding signals suitable for a high-noise environment.

[0057] Specifically, within a preset time interval before the controller emits a signal, a background noise sample sequence measured by the receiver is acquired. This sequence is then framed using a window function, and the power spectrum estimate for each frame is calculated to obtain the noise power spectral density profile. Based on spectral identification, high-energy interference bands in the noise power spectrum are identified, and the proportion of each band in the total noise energy is calculated. Based on the proportion of each band in the total noise energy, multiple candidate sub-bands with noise proportions below a preset threshold are selected as priority transmission frequency bands. Within each candidate sub-band, the center frequency and bandwidth are determined, and the linear frequency modulation slope of the spread spectrum chirp for each candidate sub-band is calculated. The duration of the linear frequency modulation signal is then statistically analyzed to match the time-domain processing window at the receiver. A pseudo-random phase sequence is then superimposed on each linear frequency modulation signal. Simultaneously, the switching period and seed of the PRN are selected according to a preset control strategy. When there are multiple transmitting units, the controller allocates orthogonal coding or different PRN seeds to each transmitting unit according to the control strategy. At the same time, it calculates the transmission weight of each transmitting unit to form the desired sound field directivity. Based on the transmission weight of each transmitting unit, it obtains the baseband signal transmitted by each transmitting unit in each sub-band. Under the premise of meeting the upper limit of total transmission power, it allocates power to different sub-bands and different transmitting paths according to the control strategy. According to the PRN seed of each transmitting unit, it adds corresponding controllable frequency hopping or phase hopping modes to each baseband signal. It selects the window type according to the control strategy and performs bandpass filtering and amplitude and phase pre-distortion compensation on each baseband signal after the window. Then, it embeds the PRN seed, packet number, current sub-band index table and other control metadata into the header of each baseband signal, and sets the verification and confirmation mechanism for each control metadata.

[0058] It should be further explained that the specific calculation formula for power spectrum estimation is as follows:

[0059] ;

[0060] In the formula, The noise power spectral density function obtained from the estimation is represented by this function. Representative moment Time Received noise time-domain sample of the frame; The Fourier transform operator representing the time domain to the frequency domain returns the frequency. Complex value spectrum at; The number of frames used for averaging.

[0061] The reconfiguration sensing module adjusts each sensing node in three-dimensional space through micro-actuators to form a variable topology array; the path optimization module designs the optimal sound wave propagation path and array configuration strategy based on real-time sound wave propagation data and boiler operating conditions.

[0062] Specifically, the original sound wave measurement data is acquired in real time from the transmitting or receiving end, including the time domain sampling sequence and sampling time stamp of each receiving channel, and the current combustion rate, flue gas flow rate, and overall pressure are obtained from the working condition sensor. The various types of data are interpolated according to a unified time reference to obtain time-synchronized observation data. Time delay estimation, angle of arrival estimation, spectral density estimation, and instantaneous signal-to-noise ratio calculation are performed on the observation data to generate physically interpretable observation features. The observation features are then standardized and spliced into candidate observation sub-vectors according to the receiving channel. The candidate observation sub-vectors are combined into state features in a fixed order, and the current working condition quantities and array topology descriptions are added. Principal component analysis is used to reduce the dimensionality of the state features. Action spaces are designed for each movable sensing node. The action vectors in the action space are mapped to the instruction set executable by each movable sensing node. According to known physical rules, physical constraints are added to each action vector in the instruction set. After calculating the execution of each action vector in the instruction set, the signal quality gain, path propagation attenuation reduction, and coverage index at the next time are calculated. The calculated indices are weighted and summed to establish a reward function. The immediate reward corresponding to the current action vector is calculated, and the index weight values are dynamically updated. The reduced state features are input into the pre-trained policy network. The policy network outputs the action distribution at the current time through forward reasoning. The action distribution output by the policy network is converted into specific driving commands, and displacement commands are issued to the corresponding movable sensing nodes. The transmitting unit sends frequency band, phase, or power allocation instructions. After the action is started, a new round of observations is collected at the next time, and the immediate reward of the current action is calculated. The current state, action, immediate reward, and next state are stored in the experience pool as experience sets. Multiple samples are randomly selected from the experience pool at regular intervals. According to the selected DRL algorithm, the TD target value of each sample is calculated. The policy network loss value is obtained based on the TD target value. The network parameters are updated using the gradient descent algorithm.

[0063] In this embodiment, the current working condition quantities specifically include temperature probe scalars, combustion rates, etc. The array topology description specifically includes node positions and orientations. The action space specifically includes displacement and orientation adjustment of the movable sensing node, as well as main pointing angle and frequency band selection of the transmitting beam, and transmitting power allocation ratio, etc.

[0064] The specific calculation formula of the reward function is as follows:

[0065] ;

[0066] Where: ;

[0067] In the formula, represents the time Instant rewards below; This represents the change in average SNR before and after the action; This represents the average change in decay. Represents the increase in coverage; Represents the action cost term, where, Represents cost weight, Represents the consumption of action resources; Representative moment The following is for the first The weight of each indicator; Represents the length of the time window; Represents the number of receive channels; Representative moment Next The SNR value of each receive channel; Representative moment Next The SNR value of each receive channel; Represents the number of rays being evaluated; Representative moment Next The propagation loss of each ray; Representative moment Next The propagation loss of each ray; Representative moment The following is considered to be the set of detectable spatial grid points; Representative moment The following is considered to be the set of detectable spatial grid points; Represents the total number of grid points; Represents the cardinality of the set.

[0068] Example 2: This embodiment of the invention provides a three-dimensional temperature field online reconstruction system based on an integrated transceiver acoustic wave sensor. See also... Figure 1 , Figure 1 This is a framework diagram of an online reconstruction system for a three-dimensional temperature field based on an integrated transceiver acoustic wave sensor, provided in an embodiment of the present invention. The system includes: an information acquisition module, a transmission control module, a reconstruction sensing module, a path optimization module, an acoustic wave receiving module, an amplitude limiting and filtering module, a decoding and separation module, a feature extraction module, a physical simulation module, a multi-source fusion module, a three-dimensional reconstruction module, a status monitoring module, and an online calibration module.

[0069] The acoustic wave receiving module is used to synchronously receive multiple acoustic wave signals from different directions, different codes and different frequency bands, and to perform preliminary synchronization and buffering processing; the amplitude limiting and filtering module is used to perform amplitude limiting and filtering processing on the received multiple acoustic wave signals.

[0070] Specifically, the external superconducting limiter is placed in a cryogenic cold stage, the cooling is started and the critical temperature margin of the device is continuously monitored, after the temperature reaches the preset target working point, the thermal equilibrium is waited for and the static transmission characteristics of the device are measured to confirm the superconducting state, then the direct current bias current or magnetic flux is applied, the working point is adjusted to make the nonlinear threshold of the superconducting limiter located in the required instantaneous input amplitude interval, the bias-response curve is recorded, the adjustable matching network is established by fine-tuning inductance, capacitance or programmable matching, and the input end of the superconducting limiter is matched with the source impedance and the input impedance of the superconducting limiter to the preset standing wave ratio range through the adjustable matching network, the rise time and saturation level of the transient response of the superconducting limiter when strong interference occurs are recorded, and the recovery time of the superconducting limiter to return to the linear working area after the strong interference is monitored in real time, the compression effect of the superconducting limiter on the input noise power spectrum of various data is quantitatively processed, the equivalent noise compression ratio of the superconducting limiter in different input power intervals is obtained by measurement, then the equivalent input reference noise after the superconducting limiter is calculated by using the equivalent noise compression factor, and the net benefit of the signal processing of the rear end is evaluated, the dynamic range and phase response of the various data after the limiting are restored to the preset through the linearization corrector, the noise baseline of the output under different bias and environment is measured and calibrated, then the corrected various data are filtered through the adaptive Wiener filter to suppress the residual background noise, after the filtering is completed, the residual interference is deeply suppressed based on the adaptive notch filter and combined with the phase compensation, the instantaneous variance, residual noise band energy and interference residual index based on discrimination of the adaptive Wiener filter output data are calculated in real time, and based on the indexes, the bias or bypass shunt factor of the superconducting limiter is adjusted and the noise spectrum estimation of the digital filter is updated.

[0071] The decoding separation module is used for decoding the processed acoustic wave signal to separate the temperature corresponding propagation time delay and amplitude change information; the feature extraction module is used for processing the decoded acoustic wave signal and extracting the propagation time delay, attenuation coefficient, multipath feature, phase offset and other information; the physical simulation module is used for constructing a CFD-based real-time operation simulation model of the boiler.

[0072] Specifically, the three-dimensional geometric information inside the boiler is obtained by field measurement, and the collected three-dimensional geometric information is preprocessed, at the same time, finer grids are used in the wall neighborhood, nozzle entrance and exit, flue corner and the expected crossing region of the sound wave path, and the rest of the region uses coarse grid, to generate a mixed grid and evaluate the grid quality, and based on the evaluation results and the mixed grid structure, output the grid file, collect the working condition sensing data, and construct the boundary condition distribution by spatial interpolation method, then input the boundary field obtained by interpolation into the CFD boundary file, at the same time, set it as a time-dependent boundary, and establish uncertainty estimation for the mapping error, then convert each acoustic sensing data into integral constraint on the sound speed field or temperature field, and apply each integral constraint as a soft constraint on the corresponding grid point of CFD, to establish a complete boiler operation simulation model, according to the gas thermodynamic relationship, the sound speed and temperature of each local area of the operation simulation model are associated, and the data constraint term of temperature or thermodynamic quantity is established, then according to the discrete temperature, velocity and pressure field on the current grid, the corresponding state set is established, and the numerical prediction is carried out for each state, when new observation data is input, the gain is calculated by using the analysis step of EnKF and each state is updated, and the state set after analysis is generated, to obtain the real-time prediction value of the boiler temperature.

[0073] The multi-source fusion module is used for fusing the body temperature information measured by the sound wave, the digital twin prediction result and the sensing data; the three-dimensional reconstruction module generates the real-time temperature field distribution according to the fused multi-source data; the state monitoring module is used for displaying each data in a graphical manner, and providing alarm prompt and abnormal area positioning; the online calibration module is used for online correcting the position deviation of the sensing array, the sound speed model error, the parameter drift of the digital twin and each data deviation, and continuously comparing the measured data and the predicted data.

[0074] The above describes one embodiment of the present application in detail, but the content described is only the preferred embodiment of the present application, and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made according to the scope of the present application should still belong to the patent coverage range of the present application.

Claims

1. A three-dimensional temperature field online reconstruction system based on a transceiving integrated acoustic wave sensor, characterized in that, The application relates to a boiler temperature field real-time monitoring system. The information acquisition module is used for acquiring the operation parameters of the inside and periphery of a boiler in real time. The transmission control module is used for generating an acoustic wave coding signal suitable for a high-noise environment. The reconstruction sensing module adjusts each sensing node in a three-dimensional space through a micro actuator to form a variable topology array. The path optimization module designs an optimal acoustic wave propagation path and array configuration strategy according to real-time acoustic wave propagation data and boiler working conditions. The sound wave receiving module is used for synchronously receiving multi-path sound wave signals from different directions, different codes and different frequency bands, and performing preliminary synchronization and buffering processing. The amplitude limiting filter module is used for performing amplitude limiting filter processing on the received multi-path sound wave signals. The decoding separation module is used for decoding the processed sound wave signals to separate the propagation time delay and amplitude change information corresponding to the temperature. The feature extraction module is used for processing the decoded sound wave signals and extracting propagation time delay, attenuation coefficient, multi-path feature, phase offset and other information. The physical simulation module is used for constructing a CFD-based real-time operation simulation model of the boiler. The multi-source fusion module is used for fusing the body temperature information measured by the sound wave, the digital twin prediction result and the sensing data. The three-dimensional reconstruction module generates a real-time temperature field distribution according to the fused multi-source data. The state monitoring module is used for displaying each item of data in a graphical manner, and providing alarm prompts and abnormal area positioning. The online calibration module is used for online correcting the position deviation of the sensing array, the sound velocity model error, the digital twin parameter drift and other data deviations, and continuously comparing the measured data and the predicted data. The specific steps of the transmission control module for generating an acoustic wave coding signal suitable for a high-noise environment are as follows: 2.The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor of claim 1, wherein, S1.1: collecting a background noise sample sequence measured by the receiver in a preset time interval before the controller sends a signal, performing window function framing on the background noise sample sequence, calculating the power spectrum estimation of each frame to obtain a noise power spectrum density profile, identifying high-energy interference bands in the noise power spectrum based on the spectrum diagram, and calculating the proportion of each frequency band in the total noise energy. ​ S1.2: According to the proportion of each frequency band in the total noise energy, select multiple groups of candidate subbands with noise proportion below the preset threshold as the priority transmission frequency band, and determine the center frequency and bandwidth in each candidate subband, calculate the chirp linear frequency modulation slope of each candidate subband, then calculate the linear frequency modulation signal length to match the time domain processing window of the receiving end, and superimpose the pseudo-random phase sequence on each linear frequency modulation signal, and select the switching period and seed of PRN according to the preset control strategy; S1.3: When there are multiple transmitting units, the controller assigns the orthogonal code or different PRN seeds of each transmitting unit according to the control strategy, and calculates the transmission weight of each transmitting unit to form the expected sound field directivity, and based on the transmission weight of each transmitting unit, obtains the baseband signal transmitted by each transmitting unit in each subband, and based on the premise of satisfying the upper limit of total transmission power, power is allocated to different subbands and different transmission paths according to the control strategy, and according to the PRN seed of each transmitting unit, a controllable frequency hopping or phase hopping mode is added to each baseband signal; S1.4: According to the control strategy, select the window type, and perform band-pass filtering and amplitude-phase pre-distortion compensation on each baseband signal after the window, then embed the PRN seed, packet number, and current subband index table control metadata in the header of each baseband signal, and set the verification and confirmation mechanism for each control metadata. 3.The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor of claim 2, wherein, The specific calculation formula of the power spectrum estimation in S1.1 is as follows: ； wherein represents an estimated acquired noise power spectral density function; represents the time instant the time instant the received noise time domain samples of the frame; represents the Fourier transform operator from time domain to frequency domain, returning the complex valued spectrum at frequencies represents the Fourier transform operator from time domain to frequency domain, returning the complex valued spectrum at frequencies is the number of frames used for averaging. 4.The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor of claim 2, wherein, The specific steps of the path optimization module to design the optimal sound wave propagation path and array configuration strategy are as follows: S2.1: Real-time acquisition of original sound wave measurement data from the transmitting or receiving end, including time domain sampling sequence and sampling time stamp of each receiving channel, and acquisition of current combustion rate, flue gas flow rate, and overall pressure from working condition sensors, and interpolation of various data according to a unified time reference to obtain time-synchronized observation data; S2.2: Time delay estimation, angle of arrival estimation, spectral density estimation, and instantaneous signal-to-noise ratio calculation are performed on the observation data to generate physically interpretable observation features, which are then standardized and concatenated into candidate observation sub-vectors according to the receiving channel, and then combined into state features according to a fixed order, and the current working condition and array topology are described, and principal component analysis is used to reduce the dimensionality of the state features; S2.3: Design corresponding action space for each movable sensing node, and map each action vector in the action space to a set of executable instructions for each movable sensing node, and add corresponding physical constraints to each action vector in the instruction set according to known physical rules, then calculate the signal quality gain, path propagation attenuation reduction, and coverage index at the next time after the execution of each action vector in the instruction set, and then weight and sum the calculated indices to establish a reward function, calculate the immediate reward corresponding to the current action vector, and dynamically update the index weight values. S2.4: input the state features after dimensionality reduction into the pre-trained policy network, output the action distribution at the current time through forward inference of the policy network, and analyze the action distribution output by the policy network after each action is performed, and convert the action distribution output by the policy network into specific drive commands, and issue displacement commands to the corresponding movable sensor nodes, and send frequency band, phase or power allocation instructions to the transmitting unit; S2.5: after starting the action, a new round of observations is collected at the next time, the immediate reward of the current action is calculated, the current state, action, immediate reward and next state are taken as experience sets and stored in the experience pool, a plurality of samples are randomly extracted from the experience pool at regular intervals, and the TD target value of each sample is calculated according to the selected DRL algorithm, the policy network loss value is obtained based on each TD target value, and the network parameters are updated by using the gradient descent algorithm.

5. The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor according to claim 4, characterized in that, The specific calculation formula of the reward function in S2.3 is as follows: ； wherein: ; wherein, representing the instant reward at time ; representing the average SNR change before and after the action; representing the average attenuation change amount; representing the coverage increment; representing the action cost term, wherein, representing the cost weight, representing the action resource consumption; representing the weight of the first term index at time ; representing the time window length; representing the number of receiving channels; representing the SNR value of the first receiving channel at time ; representing the SNR value of the first receiving channel at time ; representing the number of rays being evaluated; representing the propagation loss of the first ray at time ; representing the propagation loss of the first ray at time ; representing the set of spatial grid points considered detectable at time ; representing the set of spatial grid points considered detectable at time ; representing the total number of grid points; representing the cardinality of the set. 6.The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor of claim 4, wherein, The specific steps of the amplitude limiting filter module for performing amplitude limiting filtering on the received multiple sound wave signals are as follows: S3.1: Put the external superconducting limiter into the low-temperature cold table, start cooling and continuously monitor the critical temperature margin of the device, wait for thermal equilibrium after the temperature reaches the preset target working point, and measure the static transmission characteristics of the device to confirm the superconducting state, then apply a direct current bias current or magnetic flux, adjust the working point to make the non-linear threshold of the superconducting limiter located in the required instantaneous input amplitude interval, and record the bias-response curve at the same time; S3.2: Establish an adjustable matching network by fine-tuning inductance, capacitance or programmable matching, and match the source impedance and the input impedance of the superconducting limiter to the preset standing wave ratio range through the adjustable matching network, and record the rise time and saturation level of the transient response of the superconducting limiter when strong interference occurs, and monitor the recovery time of the superconducting limiter to the linear working area in real time after the strong interference; S3.3: Quantitatively process the compression effect of the superconducting limiter on the input noise power spectrum of various data, and obtain the equivalent noise compression ratio of the superconducting limiter in different input power intervals by measurement, then calculate the equivalent input reference noise after the superconducting limiter using the equivalent noise compression factor, and evaluate the net benefit of the back-end signal processing; S3.4: Restore the amplitude-limited data to the preset dynamic range and phase response through the linearization corrector, measure and calibrate the noise baseline of the output under different biases and environments, then filter the corrected data through an adaptive Wiener filter to suppress residual background noise, and based on an adaptive notch filter and combined with phase compensation, deeply suppress the residual interference; S3.5: Real-time calculation of the instantaneous variance, residual noise band energy and interference residual index based on discrimination of the adaptive Wiener filter output data, and based on the indicators, adjust the bias or bypass shunt factor of the superconducting limiter and update the noise spectrum estimation of the digital filter.

7. The three-dimensional temperature field online reconstruction system based on the integrated transceiving acoustic wave sensor according to claim 6, characterized in that, The specific steps of the physical simulation module for constructing a CFD-based real-time operation simulation model of the boiler are as follows: S4.1: Obtain the three-dimensional geometric information inside the boiler through field measurement, preprocess the collected three-dimensional geometric information, adopt finer grid in the wall neighborhood, nozzle inlet and outlet, flue corner and the expected crossing region of the sound wave path, adopt coarse grid in the remaining region, generate a mixed grid and evaluate the grid quality, and output the grid file based on the evaluation results and the mixed grid structure; S4.2: Collect working condition sensing data, construct boundary condition distribution through spatial interpolation method, then input the boundary field obtained by interpolation into the CFD boundary file, set it as a time-dependent boundary, establish uncertainty estimation for the mapping error, then convert each acoustic sensing data into integral constraint on the sound speed field or temperature field, and apply each integral constraint as a soft constraint on the corresponding grid point of CFD, to establish a complete boiler operation simulation model; S4.3: According to the gas thermodynamic relationship, correlate the sound speed and temperature of each local region of the operation simulation model, and establish a data constraint term of temperature or thermodynamic quantity, then according to the discrete temperature, velocity and pressure field on the current grid, establish the corresponding state set, and perform numerical prediction for each state, when new observation data is input, calculate the gain and update each state using the analysis step of EnKF, and generate the analyzed state set to obtain the real-time prediction value of the boiler temperature.

Images