A self-locking wave method and system based on laser photoacoustic spectroscopy gas detection

Abstract

The application provides a self-locking wave method and system based on laser photoacoustic spectrum gas detection, relates to the technical field of optical gas sensing, and solves a first state vector representing the real performance of the current system in real time by fusing multi-physical field sensing data such as photoacoustic signals, sound field distribution and thermal deformation, simultaneously optimizes the signal-to-noise ratio, linearity and control energy consumption as criteria by using a multi-objective optimization function, rolls to generate a cooperative control instruction, accurately adjusts the excitation light modulation parameter and the cavity acoustic mode parameter, realizes the double dynamic tracking and locking of the laser modulation frequency and the cavity acoustic resonance frequency, can automatically and continuously maintain the detection sensitivity close to the theoretical limit and the signal stability at all times when the external conditions change, uses the calculated system nonlinearity index as an internal criterion, intelligently switches the concentration inversion mode to calculate the gas concentration value, and realizes seamless, continuous and wide dynamic range measurement from the ppb level to the percentage level.

Description

Technical Field

[0001] This invention relates to the field of optical gas sensing technology, specifically to a self-locking wave method and system based on laser photoacoustic spectroscopy gas detection. Background Technology

[0002] Laser photoacoustic spectroscopy has become an important method for trace gas detection due to its high sensitivity, wide dynamic range and good gas selectivity. The basic principle is to use a modulated laser of a specific wavelength to excite gas molecules, generate periodic thermal disturbances through non-radiative relaxation, and then excite detectable acoustic signals. In use, a theoretically optimal operating point is determined by precisely machining a fixed cavity structure, selecting a quartz tuning fork with a specific resonance frequency or preset laser modulation parameters, and performing static optimization.

[0003] In practical applications, firstly, the acoustic resonant frequency of the photoacoustic cavity drifts with changes in ambient temperature, pressure, and gas composition, while the laser modulation frequency is usually fixed, making the system highly prone to detuning and causing severe signal amplitude attenuation. Secondly, when the gas concentration spans multiple orders of magnitude, the system response exhibits nonlinearity; at high concentrations, traditional photoacoustic signals saturate or even attenuate, and current technologies rely on manual range switching or sample dilution, making continuous monitoring impossible. Finally, during the detection process, there are invalid photothermal signals generated by absorption by non-target gases such as window adsorption and dust scattering, which mix with the target signal, reducing the selectivity and accuracy of detection. Existing improvement schemes mostly focus on optimizing single aspects, such as improving light source stability, designing higher-order resonant cavities, or employing differential noise suppression, lacking an intelligent closed-loop control mechanism that can perceive the global state of the system in real time and coordinate the regulation of multiple degrees of freedom, resulting in insufficient stability, dynamic range, and anti-interference capability of the system under complex operating conditions.

[0004] In summary, the existing technology has the following technical problems when used:

[0005] Problem 1: Existing laser photoacoustic spectroscopy systems based on static operating point optimization cannot adapt to the system state drift caused by environmental fluctuations and gas concentration changes in actual detection. This leads to a continuous mismatch between laser excitation efficiency, acoustic resonance and signal detection, and the system cannot maintain peak sensitivity and optimal signal-to-noise ratio in dynamic changes.

[0006] Problem 2: Traditional photoacoustic spectroscopy faces bottlenecks of signal saturation and nonlinearity when detecting high-concentration gases. At the same time, it lacks effective means to distinguish between target gas absorption and non-target interference, such as window heating and water vapor cross-absorption. The resulting background signal limits its dynamic range and reduces the accuracy and selectivity of measurements in complex gas mixtures or harsh environments. Summary of the Invention

[0007] To achieve the above objectives, the present invention provides the following technical solution: a self-locking wave method and system based on laser photoacoustic spectroscopy for gas detection, wherein the method includes:

[0008] The target gas is introduced into the multi-resonant coupled smart photoacoustic cavity integrated with a piezoelectric actuator array to obtain the first photoacoustic signal, the first sound field signal and the first deformation signal generated by the interaction of modulated excitation light and the target gas.

[0009] Based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, and combined with the preset digital twin model, a first state vector containing the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index is calculated.

[0010] Using the first state vector as input, a solution is obtained based on a preset multi-objective optimization function to generate a first subset of control instructions for adjusting the excitation light modulation parameters and a second subset of control instructions for adjusting the cavity acoustic mode parameters.

[0011] The excitation light modulation parameters are adjusted according to the first control command subset, the cavity acoustic mode parameters are adjusted according to the second control command subset, the current concentration inversion mode is determined based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector, and the output gas concentration value is calculated.

[0012] Furthermore, the construction of the digital twin model includes:

[0013] During the initialization phase, a standard gas is introduced into the multi-resonant coupled smart photoacoustic cavity to perform parameter scanning, causing the excitation light modulation parameters to vary in the first range and the cavity acoustic mode parameters to vary in the second range.

[0014] During the parameter scanning process, multiple sets of scanning parameter combinations and the corresponding photoacoustic signals, sound field signals and deformation signals under each set of parameter combinations are collected and recorded simultaneously to form training data pairs.

[0015] Using training data pairs, a mathematical model that can map the relationship between input parameters, system state, and output signal is obtained through machine learning algorithms, serving as a digital twin model.

[0016] Furthermore, the calculation of the first state vector, which includes the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity exponent, includes:

[0017] The background interference intensity value is obtained by performing digital-to-analog conversion on the first deformation signal;

[0018] The first photoacoustic signal, the first sound field signal, and the first deformation signal are currently acquired. Feature extraction is performed to obtain the first feature vector.

[0019] Obtain the currently used excitation light modulation parameters and cavity acoustic mode parameters, and input them together with the first feature vector as input data into the trained digital twin model;

[0020] The input data is processed based on the digital twin model, and the output is an estimate of the hidden state of the current system as the first state vector.

[0021] Furthermore, the generation of a first subset of control commands for adjusting the excitation light modulation parameters and a second subset of control commands for adjusting the cavity acoustic mode parameters includes:

[0022] The sub-terms of the multi-objective optimization function include a concentration prediction error term, a signal quality reciprocal term, a system nonlinearity exponent term, and a control energy consumption term.

[0023] Using a predictive control algorithm with a digital twin model as the internal predictive model, and aiming to minimize the multi-objective optimization function within a future time window, rolling optimization calculations are performed.

[0024] By solving the rolling optimization calculation, a set of future control sequences is obtained. The instruction of the first control cycle in the future control sequence is selected as the first control instruction subset and the second control instruction subset at the current moment.

[0025] Furthermore, determining the current concentration inversion mode and calculating the output gas concentration value includes:

[0026] Set a first preset threshold and a second preset threshold, with the second preset threshold being greater than the first preset threshold. Obtain the system nonlinearity index in the first state vector and compare it with the preset threshold.

[0027] When the system nonlinearity index is lower than the first preset threshold, it is determined to be in linear mode. The main frequency amplitude of the second photoacoustic signal after phase-locked amplification is used as the main observation value to calculate the gas concentration value.

[0028] When the system nonlinearity index is higher than the first preset threshold but lower than the second preset threshold, it is determined to be a nonlinear transition mode. The resonant frequency offset of the multi-resonant coupled smart photoacoustic cavity is obtained and fused with the main frequency amplitude of the second photoacoustic signal as the main observation value to calculate the gas concentration value.

[0029] When the system nonlinearity index is higher than the second preset threshold, it is determined to be a deep nonlinear mode. The core frequency parameters used to maintain the cavity resonance tracking state are obtained from the second control instruction subset. The core frequency parameters are used as the main observation values ​​for concentration inversion to calculate the gas concentration value.

[0030] Furthermore, the calculation of the effective absorption coefficient includes:

[0031] After performing digital-to-analog conversion on the first deformation signal, a scalar value characterizing the gradient of the non-uniform thermal deformation distribution on the cavity surface is obtained, which is denoted as the background interference intensity value.

[0032] The background interference intensity value is converted into a corresponding equivalent background absorption signal amplitude;

[0033] The net signal amplitude obtained by subtracting the equivalent background absorption signal amplitude from the total absorption signal amplitude obtained by spectral analysis of the first photoacoustic signal is used to finally calculate the effective absorption coefficient.

[0034] Furthermore, adjusting the excitation light modulation parameters according to the first subset of control instructions includes:

[0035] The target laser wavelength offset value, target laser modulation frequency value, and target laser modulation depth value are parsed from the first subset of control commands.

[0036] The center wavelength of the output laser is adjusted according to the target laser wavelength offset value, and the modulation frequency and modulation depth of the laser are adjusted according to the target laser modulation frequency value and the target laser modulation depth value.

[0037] Furthermore, adjusting the cavity acoustic mode parameters according to the second subset of control commands includes:

[0038] The second control instruction subset includes a set of multiple control signal parameters for driving multiple actuation units on the cavity, each control signal parameter including drive amplitude, drive phase and drive frequency;

[0039] Based on the multi-channel control signal parameters, corresponding multi-channel high-voltage drive signals are generated and applied to the corresponding actuation units respectively;

[0040] By setting different driving phase combinations, the cavity is excited to generate axial, radial, and hybrid acoustic resonance modes. By setting the driving frequency, the resonant frequency of the selected acoustic resonance mode is tuned.

[0041] Furthermore, the method also includes:

[0042] During continuous system operation, the first photoacoustic signal, the second photoacoustic signal, the first deformation signal, the first control command subset, the second control command subset, the excitation light modulation parameters, the cavity acoustic mode parameters, and the first state vector are periodically acquired and stored as data samples in the rolling historical database.

[0043] Periodically calculate the average error between the predicted signal output by the digital twin model and the real signal across all recent data samples;

[0044] When the average error exceeds the preset error threshold, an update process is triggered, using the latest data samples from the rolling historical database to perform an incremental learning and fine-tuning of the parameters of the digital twin model.

[0045] A self-locking wave system based on laser photoacoustic spectroscopy gas detection, the system comprising:

[0046] A tunable laser excitation module is used to generate a probe laser with tunable wavelength and composite modulation.

[0047] Multi-resonant coupled intelligent photoacoustic cavity module, used for target gas absorption and sound wave generation;

[0048] A multi-physics field fusion sensing module is used to acquire a first photoacoustic signal, a first acoustic field signal, and a first deformation signal generated by the interaction of modulated excitation light with the target gas.

[0049] The signal processing and state observation module is used to calculate the first state vector based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, combined with a preset digital twin model.

[0050] The multi-objective optimization decision and control module is used to take the first state vector as input, solve based on the preset multi-objective optimization function, generate a first control command subset for adjusting the tunable laser excitation module and a second control command subset for adjusting the multi-resonant coupled intelligent photoacoustic cavity module; and determine the current concentration inversion mode and calculate the output gas concentration value based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector.

[0051] This invention provides a self-locking wave method and system based on laser photoacoustic spectroscopy for gas detection. It has the following beneficial effects:

[0052] 1. This invention fundamentally solves the static operating point mismatch problem through a dynamic dual-closed-loop feedback mechanism based on a digital twin model and nonlinear state observation. It constructs a digital twin model that can reflect the optical, thermal, and acoustic conversion links in real time, integrates multi-physics field sensing data such as photoacoustic signals, sound field distribution, and thermal deformation, and calculates the first state vector characterizing the current system's true performance in real time through a nonlinear state observer. This vector includes the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index. Using a multi-objective optimization function to simultaneously optimize the signal-to-noise ratio, linearity, and control energy consumption, it generates collaborative control commands in a rolling manner, precisely adjusting the excitation light modulation parameters and cavity acoustic mode parameters. This achieves dual dynamic tracking and locking of the laser modulation frequency and the cavity acoustic resonance frequency, giving the system self-locking capability. When external conditions change, the system's operating point can be automatically and continuously pulled to the Pareto optimal state under the current conditions. Thus, it maintains near-theoretical detection sensitivity and signal stability over a wide range of environmental changes and concentrations, improving adaptability and robustness.

[0053] 2. The system of this invention utilizes the calculated system nonlinearity index as an intrinsic criterion to intelligently switch concentration inversion modes. In the low-concentration linear region, a high-sensitivity signal amplitude detection is adopted. When the index increases, indicating nonlinearity, it smoothly transitions to a fusion algorithm that integrates the signal amplitude and the resonant frequency offset. In the deep nonlinear region, it switches to a detection mode that uses the core parameters of the control commands required to maintain the cavity resonance tracking state as the main observation quantity, perfectly avoiding the signal saturation problem. This achieves seamless, continuous, and wide dynamic range measurement from the ppb level to the percentage level. At the same time, by independently acquiring and analyzing the first deformation signal reflecting the local cavity, it can directly quantify the ineffective thermal disturbances caused by non-target absorption or adsorption effects. In the nonlinear state observer, the first deformation signal is used as a key input to model the mixed photoacoustic signal and subtract the background interference component in real time, thereby accurately extracting the effective absorption coefficient that is only related to the target gas. This greatly enhances the gas selectivity of the system in complex background gases or polluted environments, making the detection results more accurate and reliable. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the steps of a self-locking wave method for gas detection based on laser photoacoustic spectroscopy according to the present invention.

[0055] Figure 2 This is a data transmission diagram of a self-locking wave method for gas detection based on laser photoacoustic spectroscopy according to the present invention;

[0056] Figure 3 This is an architectural diagram of a self-locking wave system based on laser photoacoustic spectroscopy gas detection according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] like Figures 1 to 2 As shown, a self-locking wave method based on laser photoacoustic spectroscopy for gas detection is described, the method comprising:

[0059] Step S100: The target gas is introduced into the multi-resonant coupled smart photoacoustic cavity integrated with a piezoelectric actuator array to obtain the first photoacoustic signal generated by the interaction of the modulated excitation light and the target gas, the first sound field signal reflecting the spatial distribution of the sound field in the cavity, and the first deformation signal reflecting the local thermal deformation of the cavity.

[0060] Among them, the piezoelectric actuator array is a set of independently controllable actuator units embedded in the cavity wall and end face of the multi-resonant coupled smart photoacoustic cavity. By applying electrical signals of specific frequency, phase and amplitude to it, the acoustic resonant modes in the multi-resonant coupled smart photoacoustic cavity are actively excited, shaped and dynamically adjusted. The piezoelectric actuator array generates precise mechanical vibrations to change the boundary conditions of the cavity, thereby realizing the active matching and locking of the sound field and the modulation frequency of the excitation light. It is the key physical basis for realizing the self-locking wave capability.

[0061] Target gas typically refers to trace or high-concentration gases that need to be detected with high sensitivity and selectivity, such as methane and carbon monoxide in environmental monitoring, ammonia and hydrogen sulfide in industrial process gases, or acetone in medical diagnostic gases. The target gas is introduced into the core reaction region of the main resonant cavity through a microchannel integrated in the inner layer of the multi-resonant coupled intelligent photoacoustic cavity at a controlled flow rate and pressure, ensuring that the gas interacts fully with the modulated excitation laser.

[0062] The multi-resonant coupled intelligent photoacoustic cavity includes a main resonant cavity for gas absorption and sound wave generation, an outer anti-interference buffer cavity for isolating external mechanical noise and temperature fluctuations, and microchannels and optical waveguides integrated in the inner layer for guiding excitation light and realizing the entry and exit of target gas. The multi-resonant coupled intelligent photoacoustic cavity provides a closed environment with high acoustic quality factor (Q value) and modal tunability, which efficiently converts the weak thermal energy generated by the target gas absorbing light energy into a strong acoustic resonance signal of a specific mode, and can be dynamically tuned to adapt to different working conditions.

[0063] The modulated excitation light is generated by a tunable laser excitation module. The output wavelength of the tunable semiconductor laser in the tunable laser excitation module is aligned with the characteristic absorption line of the target gas. It is then modulated by an integrated electro-optic modulator and subjected to intensity modulation of a specific frequency and waveform, thus becoming a periodically pulsating modulated excitation light. After being collimated by an optical waveguide, the modulated excitation light passes perpendicularly or coaxially through the main resonant cavity filled with the target gas. After the target gas molecules selectively absorb the light energy of the modulated excitation light, they convert the light energy into periodic thermal expansion and contraction through non-radiative relaxation, thereby exciting sound waves.

[0064] Step S101: Before the target gas is introduced, a digital twin model is constructed. The construction process includes:

[0065] First, during the initialization phase, a standard gas is introduced into the multi-resonant coupled intelligent photoacoustic cavity to perform parameter scanning. According to the preset step size and sequence, the modulation frequency, modulation depth, and wavelength bias of the modulation excitation light are changed sequentially within the first range. At the same time, for each excitation light modulation parameter, the piezoelectric actuator array is driven to work sequentially at different frequencies, phases, and amplitudes within the second range, thereby exciting different acoustic resonance modes. This causes the excitation light modulation parameters to change within the first range, and the acoustic mode parameters of the cavity to change within the second range.

[0066] Then, during the parameter scanning process, multiple sets of scanning parameter combinations and the corresponding photoacoustic signals, sound field signals, and deformation signals under each parameter combination are simultaneously acquired and recorded to form training data pairs. The multiple sets of scanning parameter combinations include the specific excitation light modulation parameters and cavity acoustic mode parameters used at each moment. The excitation light modulation parameters include the current wavelength bias, modulation frequency, and modulation depth. The cavity acoustic mode parameters include the amplitude D1, phase P1, and frequency F1 of actuator 1, ..., and the amplitude D of actuator N. n Phase P n Frequency F n ;

[0067] Finally, using the training data pairs, a mathematical model that maps the relationship between input parameters, system state, and output signal is obtained through machine learning algorithms, serving as a digital twin model. Input parameters refer to the excitation light modulation parameters and cavity acoustic mode parameters. System state specifically refers to physical quantities that describe the key performance of the system and are difficult to measure directly, namely the effective absorption coefficient of the first state vector, the acoustic-optical coupling efficiency factor, and the system nonlinearity index. When using the digital twin model, the input parameters and system state are used together as input data. Output signal specifically refers to the original physical signal directly measured by the system sensors, namely the first photoacoustic signal, the first sound field signal, and the first deformation signal. The digital twin model learns the dynamic and nonlinear relationship between these three.

[0068] The digital twin model includes a machine learning algorithm that encapsulates the physical laws governing the coupling of light, heat, sound, and electricity. This algorithm is a deep, fully connected neural network, comprising an input layer (the number of nodes equals the sum of the input parameters and the dimension of the first feature vector), multiple hidden layers (each containing several neurons and employing nonlinear activation functions such as ReLU for high-order feature abstraction and complex relation mapping), and an output layer (the number of nodes equals the three dimensions of the first state vector, corresponding to the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity exponent, typically using linear activation functions). During training... In step S101, the excitation light modulation parameters, cavity acoustic mode parameters, and the first feature vector are collected as input parameters, and the real state vector obtained by theoretical calculation or indirect calibration of the standard gas signal is used as the target label. The backpropagation algorithm is used, with the mean square error between the network output and the target label as the loss function. A gradient descent optimizer (such as Adam) is used to iteratively adjust all weights and bias parameters in the network. After sufficient iteration, the deep fully connected neural network learns a robust mapping from the input to the state vector. This trained deep fully connected neural network can then be used as a usable digital twin model.

[0069] Standard gases are pre-prepared target gases whose concentrations have been calibrated by authoritative metrology institutions, such as 10ppm methane / nitrogen mixtures. In the initial stage of building a digital twin model, a set of known and real input-output response relationships are provided, so that machine learning algorithms can learn what signals and states will be generated by various combinations of system parameters under ideal calibration conditions, providing a data basis for subsequent state estimation of gases with unknown concentrations.

[0070] The excitation light modulation parameters are a set of instructions that control the output characteristics of the modulated excitation light, including the wavelength offset of the laser center wavelength relative to the absorption peak of the target gas, the modulation frequency of the laser intensity, and the modulation depth; used to drive the laser and modulator to control and adjust the parameters of the modulated excitation light; the cavity acoustic mode parameters are a set of instructions that control the operation of the piezoelectric actuator array, including the amplitude, phase, and frequency of the driving voltage signal applied to each actuator unit, used to drive the piezoelectric actuator array, thereby actively exciting, selecting, and locking the specific acoustic resonance modes of the multi-resonant coupled smart photoacoustic cavity, and determining the efficiency of heat energy conversion into sound energy;

[0071] The first range is the variation range of the excitation light modulation parameters, which is set based on the physical tuning capability of the laser, the linewidth and line shape of the target gas absorption line, and empirical values ​​to avoid signal saturation and excessive nonlinearity. For example, the wavelength offset range is set to ±2 times the full width at half maximum (FWHM) of the absorption line center frequency; the modulation frequency range is set to 0.5 to 2 times the theoretical value of the first-order resonant frequency of the multi-resonant coupled smart photoacoustic cavity. The second range is the variation range of the cavity acoustic mode parameters, which is set based on the finite element simulation results of the acoustic resonant modes of the multi-resonant coupled smart photoacoustic cavity, the effective driving frequency band and voltage amplitude limit of the piezoelectric actuator array, and the desired coupling mode order. For example, the frequency range covers the first to third order resonant frequencies of the cavity axis, and the phase combinations include typical configurations such as the 0° / 180° excitation symmetric mode and the 90° / 270° excitation antisymmetric mode.

[0072] Step S102: Obtain the first photoacoustic signal, the first sound field signal, and the first deformation signal generated by the interaction of the target gas with the modulated excitation light after the target gas is introduced into the multi-resonant coupled smart photoacoustic cavity.

[0073] The first photoacoustic signal is acquired by a high-Q acoustic sensor (such as a miniature microphone or piezoelectric ultrasonic sensor) placed close to the wall of the main resonant cavity (e.g., connected by an acoustic coupling agent) or inside the cavity. The target gas absorbs and modulates the excitation light to generate periodic thermal expansion, which excites sound pressure fluctuations in the main resonant cavity. These sound pressure fluctuations directly act on the sensitive element of the high-Q acoustic sensor, converting it into a corresponding analog voltage signal. This voltage signal, after pre-amplification and filtering, serves as the first photoacoustic signal characterizing the gas absorption intensity. The first photoacoustic signal mainly includes the main frequency amplitude, which is related to the gas concentration, and the corresponding harmonic components, serving as direct observations for gas concentration inversion.

[0074] For the first sound field signal, a distributed optical fiber acoustic wave sensor array wound around the outer wall of the cavity or embedded in the cavity structure is used to acquire it. The sound field excited inside the cavity will cause micro-strain in the optical fiber, thereby changing the phase of the probe light transmitted in it. By demodulating this phase change (such as using φ-OTDR or the principle of interferometer), the sound pressure information of each point spatially distributed along the optical fiber can be reconstructed to form a sound field image reflecting the spatial distribution of sound pressure inside the cavity (such as the position of antinodes and nodes of standing waves), which is the first sound field signal. The first sound field signal is a set of data vectors or images corresponding to the sound pressure amplitude at spatial positions, containing spatial distribution information of different acoustic resonance modes, which is used to evaluate the acoustic-optical coupling efficiency, that is, to determine whether the thermal energy generated by the excitation light is efficiently coupled to the dominant resonance mode of the current cavity.

[0075] The first deformation signal is acquired using a differential photothermal deformation sensor. This sensor contains two probe lasers focused on adjacent points on the cavity surface. The heat generated by the target gas absorbing the light energy not only excites sound waves but also causes micron-level deformation due to localized heating of the cavity. After the two probe lasers are reflected from the deformed surface, their optical path difference changes. By measuring the phase difference between the two reflected beams using interferometry, a voltage signal proportional to the thermal deformation gradient between the two points is obtained. This signal is the first deformation signal. The first deformation signal is a low-frequency, slowly varying voltage signal proportional to the local thermal deformation gradient on the cavity surface. It reflects the intensity of ineffective thermal disturbances caused by absorption by non-target gases (such as window adsorption or cavity wall absorption) or fluctuations in ambient temperature. It is used for real-time quantification and subtraction of background interference, thereby improving the selectivity and accuracy of detection.

[0076] Step S200: Based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, and combined with the preset digital twin model, calculate the first state vector, which includes the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index.

[0077] Step S201: Extract features from the currently acquired first photoacoustic signal, first sound field signal, and first deformation signal to obtain a first feature vector; the first feature vector includes the fundamental frequency amplitude, signal-to-noise ratio, and ratio of specific harmonic amplitude to fundamental frequency amplitude obtained from the first photoacoustic signal through spectral analysis; the proportion of the main resonant mode energy to the total energy extracted from the first sound field signal, the Euclidean distance between the maximum sound pressure point and the center of the optical path as the positional deviation, the sound field spatial symmetry index; the average thermal deformation gradient amplitude and the DC offset of the deformation signal calculated from the first deformation signal;

[0078] Step S202: Obtain the currently used excitation light modulation parameters and cavity acoustic mode parameters, and input them together with the first feature vector as input data into the trained digital twin model; process the input data based on the digital twin model, and output the estimated value of the current hidden state of the system as the first state vector;

[0079] First, the input layer receives input data composed of the current excitation light modulation parameters, cavity acoustic mode parameters, and the first feature vector;

[0080] Then, the input data is passed to the first hidden layer. Each neuron performs a weighted summation of the input and adds a bias, and then generates an output through a non-linear activation function such as ReLU. This process realizes the non-linear combination and transformation of primary features. The output of the previous hidden layer is used as the input of the next hidden layer, and the above steps of weighting, summation and activation are repeated.

[0081] Finally, through a series of hidden layers of abstraction, the digital twin model captures the complex, high-order mapping relationship between the input parameters and the internal state of the system. In the output layer, the output of the last hidden layer is linearly weighted and combined to directly generate three scalar values, which correspond to the effective absorption coefficient of the first state vector, the acoustic-optical coupling efficiency factor, and the system nonlinearity index, respectively.

[0082] Step S2021: The effective absorption coefficient is a quantitative indicator of the effective absorption intensity produced solely by the absorption of excitation light energy by target gas molecules after deducting background ineffective thermal effects. It can be used for concentration inversion and is directly proportional to the concentration of the target gas. It is the core physical quantity for concentration calculation. The calculation of the effective absorption coefficient includes:

[0083] First, after performing digital-to-analog conversion on the first deformation signal, a scalar value characterizing the gradient of the non-uniform thermal deformation distribution on the cavity surface is obtained, which is denoted as the background interference intensity value.

[0084] Secondly, the background interference intensity value is converted into a corresponding equivalent background absorption signal amplitude. A linear scaling factor K is experimentally established between the background interference intensity value and the background sound signal amplitude. The background sound signal amplitude is data measured on a high-Q acoustic sensor, generated by a known background interference source (such as the heated cavity wall). In real-time measurement, the calculated background interference intensity value is multiplied by the linear scaling factor K to obtain the equivalent background absorption signal amplitude A. bg ;

[0085] Then, the total absorption signal amplitude A obtained from the spectral analysis of the first photoacoustic signal is... total In the middle, subtract the equivalent background absorbed signal amplitude A bg The obtained net signal amplitude A net Used for the final calculation of the effective absorption coefficient, i.e., A net =A total -A bg ;

[0086] Finally, the effective absorption coefficient α eff Through the relation, α eff =(A net *G) / (P0*B); where P0 is the average power of the incident laser, B is a constant of the calibrated multi-resonant coupled smart photoacoustic cavity, determined by the cavity geometry and gas properties; G is the gain factor, which is a function of the acousto-optic coupling efficiency factor, with the relationship G=f(η) / (P0*B). coupling The gain factor is determined during system calibration. The effective absorption coefficient is proportional to the net acoustic signal amplitude and inversely proportional to the incident light power, and is determined by B and η. coupling Perform corrections.

[0087] Step S2022, Acousto-optic coupling efficiency factor ηcoupling The acoustic-optic coupling efficiency factor is an index characterizing the efficiency of the thermal energy absorbed by the target gas being coupled into the specific acoustic resonance mode currently excited in the multi-resonant coupled smart photoacoustic cavity. Its value ranges from 0 to 1; a higher value indicates a stronger acoustic field resonance enhancement effect and a larger signal. The calculation of the acoustic-optic coupling efficiency factor includes:

[0088] First, the dominant acoustic mode currently excited is determined from the first sound field signal through modal recognition;

[0089] Then, the sound pressure energy of the dominant mode is obtained by summing the squares of the sound pressure amplitudes in the spatial region corresponding to that mode. The ratio of the sound pressure energy of the dominant mode to the total sound pressure energy of all modes in the entire cavity is used to obtain the mode purity ratio η. mode ;

[0090] Next, the spatial intensity distribution of the excitation laser spot is analyzed to determine the degree of spatial overlap between the laser beam and the acoustic pressure antinode region of the current dominant acoustic mode. A spatial overlap integral factor λ is calculated, while considering the current laser modulation frequency f. mod The resonant frequency f of the current dominant cavity mode res The detuning amount Δf, Δf=|f mod -f res The frequency matching factor is calculated using a pre-stored Lorentz linear function L(Δf) that characterizes the sharpness of the cavity's acoustic frequency response.

[0091] Finally, the acoustic-optical coupling efficiency factor η coupling It consists of the product of these three factors, η coupling =η mode *λ*L(Δf), the acoustic-optical coupling efficiency factor η coupling It reflects in real time the efficiency of converting thermal energy into useful resonant sound energy.

[0092] Step S2023: The system nonlinearity index NI is a dimensionless index that quantitatively describes the degree to which the entire photoacoustic sensing link (including gas absorption itself, sound wave excitation and propagation, cavity response, etc.) deviates from the ideal linear behavior. The higher the index, the more severe the nonlinearity (such as absorption saturation, waveform distortion caused by excessive sound pressure). It is calculated in one go by using a digital twin model as a nonlinear state observer and fusing multi-dimensional input data. The calculation of the system nonlinearity index NI includes:

[0093] First, a spectral analysis is performed on the first photoacoustic signal to accurately extract its fundamental frequency amplitude A1 and second harmonic amplitude A2. The harmonic distortion HD2 = A2 / A1 is then calculated. Simultaneously, the current excitation light modulation parameters, cavity acoustic mode parameters, and the first eigenvector are input into the digital twin model, allowing the digital twin model to predict the fundamental frequency amplitude A1 of the photoacoustic signal that should be generated under the ideal linear assumption. predCalculate the model prediction error ratio ER, ER=|A1-A1 pred | / A1 pred ;

[0094] Then, the first sound field signal is analyzed to detect whether there is spatial distribution distortion of sound pressure amplitude, such as the position of antinode shift or the appearance of asymmetric components, and a sound field distortion index DI is calculated.

[0095] Finally, the system nonlinearity index NI is calculated through a weighted fusion function, with the formula NI = w1*HD2 + w2*ER + w3*DI, where w1, w2, and w3 are pre-set weighting coefficients used to comprehensively reflect the nonlinear performance from three dimensions: signal harmonics, model bias, and sound field morphology. The system nonlinearity index NI comprehensively quantifies the overall nonlinearity of the system.

[0096] Step S300: Using the first state vector as input, solve the problem based on a preset multi-objective optimization function to generate a first subset of control instructions for adjusting the excitation light modulation parameters and a second subset of control instructions for adjusting the cavity acoustic mode parameters.

[0097] Step S301: The sub-terms of the multi-objective optimization function include a concentration prediction error term, a signal quality reciprocal term, a system nonlinearity exponent term, and a control energy consumption term. A predictive control algorithm, using a digital twin model as the internal prediction model, performs rolling optimization calculations with the objective of minimizing the multi-objective optimization function within a future time window. The preset multi-objective optimization function is defined as F = α·E1 + β·E2 + γ·E3 + δ·E4, where E1 is the concentration prediction error term, E2 is the signal quality reciprocal term, E3 is the system nonlinearity exponent term, E4 is the control energy consumption term, and α, β, γ, and δ are weighting coefficients.

[0098] Among them, the concentration prediction error term E1 comes from the digital twin model. By inputting the current first state vector into the model, the predicted signal in the next time window is predicted if the parameters remain unchanged in the current state. Then, the predicted value is converted into a concentration prediction value through a preset concentration inversion algorithm. The root mean square error between the concentration prediction value and the short-term estimate of the real concentration obtained based on the real signal is calculated.

[0099] The reciprocal of the signal quality term E2 is derived from the first photoacoustic signal and the first sound field signal. It is obtained by calculating the reciprocal of the signal-to-noise ratio of the current photoacoustic signal and combining it with the purity of the dominant mode in the sound field signal (the lower the purity, the larger the value of this term).

[0100] The system nonlinearity index term E3 is directly derived from the system nonlinearity index itself in the first state vector calculated in step S202;

[0101] The control energy consumption term E4 is derived from the first and second control instruction subsets to be generated. It is obtained by calculating the weighted sum of the square of the change in the laser wavelength tuning current to be applied, the square of the change in the modulation depth voltage, and the total power of the multiple high-voltage signals required to drive the piezoelectric actuator array.

[0102] Step S302: Solve the rolling optimization calculation to obtain a set of future control sequences. Select the instruction of the first control cycle in the future control sequence as the first control instruction subset and the second control instruction subset at the current moment.

[0103] First, taking the first state vector at the current moment as the initial system state, in the digital twin model, taking the next N control cycles as the prediction time domain, and taking the candidate excitation light modulation parameters and cavity acoustic mode parameter sequence as input, the evolution of the system state and the generation of the output signal are simulated in a rolling manner.

[0104] Then, an optimization algorithm (such as gradient descent or sequential quadratic programming) is used to search within the feasible region of the parameters to find the specific set of control parameters that minimizes the cumulative multi-objective optimization function F value over the next N control cycles.

[0105] Finally, the specific control parameter sequences are summarized in chronological order to generate the optimal future control sequence containing the next N time points, which is the parameter combination of the excitation light modulation parameters and the parameter combination of the cavity acoustic mode parameters corresponding to each time point.

[0106] The first set of control instructions consists of immediate instructions for adjusting the tunable laser excitation module, including the target laser wavelength bias value, the target laser modulation frequency value, and the target laser modulation depth value. The second set of control instructions consists of immediate instructions for adjusting the multi-resonant coupled intelligent photoacoustic cavity module, including a set of multiple control signal parameters for driving each actuation unit in the array, each containing the driving amplitude, driving phase, and driving frequency.

[0107] Step S400: Adjust the excitation light modulation parameters according to the first control instruction subset, adjust the cavity acoustic mode parameters according to the second control instruction subset, determine the current concentration inversion mode based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector, and calculate the output gas concentration value.

[0108] Step S401, adjusting the excitation light modulation parameters according to the first control command subset, including:

[0109] First, the target laser wavelength offset value, target laser modulation frequency value, and target laser modulation depth value are parsed from the first subset of control commands. The target laser wavelength offset value represents the specific amount by which the laser center wavelength needs to be shifted from the center of the target gas absorption line towards a longer or shorter wavelength direction. This is used for fine alignment or slight deviation from the absorption line to optimize the signal-to-noise ratio or avoid saturation. The target laser modulation frequency value represents the sinusoidal (or other waveform) modulation frequency applied to the laser intensity. This is used to match the acoustic resonance frequency of the multi-resonant coupled smart photoacoustic cavity to achieve signal enhancement. The target laser modulation depth value is a dimensionless proportionality coefficient (e.g., between 0 and 1), representing the amplitude of the laser intensity fluctuating around its average value, affecting the intensity and linearity of the photoacoustic signal.

[0110] Then, based on the target laser wavelength bias value, the center wavelength of the output laser is adjusted by the laser driver circuit. Based on the target laser modulation frequency value and the target laser modulation depth value, the modulation frequency and modulation depth of the laser are adjusted by the laser modulator driver circuit. The laser driver circuit is a precision current and temperature control circuit that, based on the target laser wavelength bias value, precisely and stably controls the output center wavelength of the laser by adjusting the injection current of the laser diode and / or the current of the thermistor. The laser modulator driver circuit is a high-speed voltage or current drive amplifier that, based on the target laser modulation frequency value and the target laser modulation depth value, generates a high-quality AC signal with corresponding frequency and amplitude to drive the electro-optic modulator or directly modulate the laser current, thereby applying the required intensity modulation to the laser output.

[0111] Step S402, adjusting the cavity acoustic mode parameters according to the second control command subset, including:

[0112] First, the second control command subset contains a set of multi-channel control signal parameters for driving multiple actuators on the cavity. Each control signal parameter includes drive amplitude, drive phase, and drive frequency. Among them, the drive amplitude determines the magnitude of the vibration intensity of each actuator; the drive phase determines the temporal relationship between the vibration of this actuator and the vibration of other actuators, which is the key to forming specific modes (such as the position of antinodes and nodes of standing waves); the drive frequency determines the speed of vibration and needs to match or be close to the natural resonant frequency of the cavity.

[0113] Then, based on the parameters of the multiple control signals, the signals are converted by a multi-channel digital-to-analog converter to generate multiple low-voltage analog sine wave signals. After being synchronously amplified by an amplifier, the signals are generated into corresponding multiple high-voltage drive signals, which are then applied to the corresponding actuators.

[0114] Finally, by setting different driving phase combinations, the cavity is excited to generate axial, radial, and hybrid acoustic resonance modes. By setting the driving frequency, the resonant frequency of the selected acoustic resonance mode is tuned.

[0115] Among them, the acoustic pressure antinodes of the axial mode are distributed along the cavity axis and overlap with the coaxial optical path, making it the most commonly used and efficient mode; the acoustic pressure of the radial mode is distributed radially and is used for special cavity shapes or to realize multimodal sensing; the hybrid mode combines the characteristics of the axial mode and the radial mode and is used to extend the tuning range or suppress specific interference.

[0116] When tuning the resonant frequency of the selected acoustic resonance mode, the target acoustic resonance mode that needs to be maintained or switched is determined through feedback from the first sound field signal. The driving frequency of the actuator corresponding to the target acoustic resonance mode in the second control command subset is set to a value that is close to but may slightly deviate from the natural frequency of the mode. When the high-voltage driving signal is applied, the cavity is forced to vibrate. The amplitude and phase of the actual sound field response are continuously monitored by a distributed fiber optic acoustic wave sensor array. The driving frequency is finely adjusted by using a phase-locked loop or peak search algorithm so that the driving frequency of the actuator always tracks and locks to the frequency that maximizes the sound pressure response amplitude of the target acoustic resonance mode in the cavity (i.e., the resonant point). This dynamic tracking process realizes the active tuning and locking of the resonant frequency.

[0117] Step S403: Determine the current concentration inversion mode and calculate the output gas concentration value, including:

[0118] Set a first preset threshold θ1 and a second preset threshold θ2, where θ2 > θ1. Obtain the system nonlinearity index NI in the first state vector and compare it with the preset threshold to determine the concentration inversion mode. There are three concentration inversion modes: linear mode, nonlinear transition mode, and deep nonlinear mode. By setting the concentration inversion mode, the system can adaptively span a wide dynamic range from trace amounts (ppb level) to high concentrations (percentage level). Under different nonlinearities, the optimal observation and inversion methods are adopted to ensure measurement accuracy and continuity throughout the entire range.

[0119] When NI≤θ1, it is determined to be in linear mode. The main frequency amplitude of the second photoacoustic signal after lock-in amplification is used as the main observation value. The gas concentration value C is directly calculated by querying the pre-stored linear calibration curve. At this time, C is calculated by the formula C=K*A, where A is the main frequency amplitude of the second photoacoustic signal after lock-in amplification, and K is the linear calibration coefficient, which is obtained by pre-calibration using a low-concentration standard gas. This formula is based on the linear theory of photoacoustic effect, which assumes that the signal amplitude is proportional to the gas concentration and absorption coefficient at low concentrations.

[0120] When θ1 < NI ≤ θ2, it is determined to be a nonlinear transition mode. The resonant frequency offset of the multi-resonant coupled intelligent photoacoustic cavity is obtained and fused with the main frequency amplitude of the second photoacoustic signal. The gas concentration value C is calculated by the nonlinear mapping function stored in the digital twin model. At this time, C is calculated by the formula C = f(A, Δf) = g1(A)*(1-ω) + g2(Δf)*ω. Where A is the main frequency amplitude, Δf is the resonant frequency offset of the multi-resonant coupled intelligent photoacoustic cavity in Hz, which is the difference between the current resonant frequency and the cavity or low-concentration reference frequency. g1(A) and g2(Δf) are nonlinear mapping functions stored in the digital twin model, which are learned by neural network or polynomial fitting and describe the nonlinear relationship between amplitude and concentration, and frequency offset and concentration, respectively. ω is a weighting factor, which is dynamically calculated by the current system nonlinearity index NI, ω = (NI-θ2) / (θ2-θ1), so that the stronger the nonlinearity, the higher the weight of the frequency offset.

[0121] When η > θ2, it is determined to be a deep nonlinear mode. The core frequency parameters used to maintain the cavity resonance tracking state in the second control command subset are obtained. The core frequency parameters are used as the master observation value for concentration inversion. The gas concentration value C is calculated based on the high-concentration nonlinear function learned by the digital twin model. At this time, C is obtained by the formula C = h(f core ) calculate, where f core It is the core frequency parameter used to maintain the cavity resonance tracking state in the second control instruction subset. h is a high-concentration nonlinear function learned by the digital twin model. It is usually a monotonic function or a lookup table. This function establishes the correspondence between the driving frequency and the gas concentration required to compensate for the sound speed change caused by the gas and maintain resonance lock under strong absorption.

[0122] The first preset threshold θ1 and the second preset threshold θ2 are two critical criteria values ​​used to distinguish the three concentration inversion modes. θ1 and θ2 are set based on the relationship curve between the system nonlinearity index and the gas concentration obtained in the calibration experiment. θ1 is usually set at the index value corresponding to the curve starting to deviate significantly from linearity (e.g., harmonic distortion reaches 2%), while θ2 is set at the index value corresponding to the signal amplitude approaching saturation (e.g., amplitude increase is less than 10% of concentration increase). θ1 is generally taken as 0.05 to 0.15 (dimensionless), and θ2 is generally taken as 0.25 to 0.4. The specific values ​​are determined through experimental calibration for different gases and chambers.

[0123] The linear mode represents the system operating in the linear region, where the photoacoustic signal amplitude is highly linearly proportional to the gas concentration; the nonlinear transition mode represents the system entering the weak nonlinear region, where the relationship between signal amplitude and concentration shows a predictable and smooth nonlinear deviation; the deep nonlinear mode represents the system entering the strong nonlinear or saturation region, where traditional amplitude detection fails.

[0124] The master observation refers to one or more physical quantities that are the most reliable core inputs for concentration calculation under the current concentration inversion mode. In linear mode, the signal amplitude has the highest signal-to-noise ratio and good linearity, so it is used as the master observation. In nonlinear transition mode, the linearity of the amplitude decreases, while the resonant frequency offset begins to show a monotonic relationship with the concentration, so the two are combined as the master observation. In deep nonlinear mode, the amplitude may saturate, and the frequency offset may also become unstable. The driving frequency (control quantity) required to maintain cavity resonance becomes the only reliable observation that is strongly correlated with the concentration and changes monotonically, so it is used as the master observation. This switching is to always use the most robust and sensitive physical quantity under the current state for inversion.

[0125] The core frequency parameter of the cavity resonance tracking state refers to the drive frequency setpoint calculated by the multi-objective optimization decision and control module. This setpoint is adjusted and output in real time to precisely lock the drive frequency of the piezoelectric actuator array at the resonant peak of the current optimal acoustic resonance mode. It is included in the second control instruction subset and serves as the frequency reference for generating multiple high-voltage drive signals. In the deep nonlinear mode, when direct photoacoustic signal observation fails, this active control parameter becomes the main basis for concentration inversion due to its definite monotonic dependence on gas concentration.

[0126] Step S404: During the continuous operation of the system, the first photoacoustic signal, the second photoacoustic signal, the first deformation signal, the first control command subset, the second control command subset, the excitation light modulation parameters, the cavity acoustic mode parameters, and the first state vector are periodically acquired and stored as data samples in the rolling historical database.

[0127] The average error between the predicted signal output by the digital twin model and the real signal is calculated periodically across all recent data samples. When the average error exceeds a preset error threshold, an update process is triggered, using the latest data samples from the rolling historical database to perform an incremental learning and fine-tuning of the parameters of the digital twin model.

[0128] The predicted signal output by the digital twin model refers to the estimated value of the sensor signal that the system should produce, calculated by the model based on the input data. This includes the predicted photoacoustic signal characteristics (such as the dominant frequency amplitude), the predicted sound field distribution vector, and the predicted deformation gradient value. It is used to compare with the real sensor signal collected at the same time. The error between the two is calculated to evaluate the prediction accuracy and reliability of the digital twin model, providing triggering conditions and training targets for online model updates. The real signal refers to the actual measurement data collected in real time from each sensor of the multi-physics fusion sensing module and preprocessed (such as filtering and feature extraction). This includes the feature representations of the actual first photoacoustic signal, first sound field signal, and first deformation signal.

[0129] The process of incrementally learning and fine-tuning the parameters of a digital twin model includes:

[0130] First, when the average error exceeds a preset threshold, the most recent batch (e.g., the past 1000 sets) of data samples is extracted from the rolling historical database, including the input parameters, the first feature vector and the corresponding real signal.

[0131] Next, this batch of data is divided into a fine-tuning training set and a validation set. The current digital twin model is loaded, and the parameters of most of its hidden layers are frozen. Only the parameters of the last 1-2 hidden layers and the output layer are unfrozen to prevent catastrophic forgetting. A small learning rate (e.g., 1 / 10 of the initial learning rate) is used to fine-tune the training set data. The loss function is to minimize the mean square error between the model's predicted signal and the real signal. One or a few rounds of iterative training are performed.

[0132] Finally, after training is complete, the performance of the fine-tuned model is evaluated using the validation set. If the average error drops below the threshold, the original model parameters are replaced with the updated parameters; otherwise, the learning rate is adjusted or more layers are unfrozen for re-fine-tuning until the requirements are met or the maximum number of iterations is reached.

[0133] In this embodiment, during actual use, the establishment of the digital twin model and the solution of the multi-objective optimization function can be carried out in computer software. NVIDIA Omniverse is used as the digital twin foundation and a unified data computing platform. In Omniverse, a deep learning model trained by PyTorch or a reduced-order physical model generated by ANSYS TwinBuilder is integrated as the core state observer. Within the Omniverse platform, optimization libraries such as LibMOON or pymoo are called as the solution engine for the multi-objective optimization decision module. NVIDIA Omniverse is a collection of libraries and microservices specifically designed for developing physical AI simulation and industrial digital twin applications. It has powerful physical simulation rendering capabilities and the ability to carry and process complex multimodal data, and can simulate the coupling process of multiple physical fields such as light, heat, and sound with high fidelity.

[0134] like Figure 3 As shown, a self-locking wave system based on laser photoacoustic spectroscopy gas detection is disclosed. The system includes:

[0135] The tunable laser excitation module generates a wavelength-tunable and composite-modulated detection laser. Its optical path integrates a colloidal quantum dot film to enhance the photoacoustic conversion efficiency of the target gas. The colloidal quantum dot film contains quantum dots such as lead sulfide or cadmium selenide. When the excitation wavelength matches the size-tuned absorption peak of the quantum dots, the quantum dots generate a strong local optical field enhancement and convert the absorbed light energy into heat energy almost non-radiatively. This heat is rapidly transferred to adjacent target gas molecules, thereby amplifying the thermal expansion and contraction effect generated by the gas molecules themselves absorbing light energy, ultimately multiplying the generated photoacoustic signal. It is especially suitable for the detection of trace gases with small absorption cross-sections.

[0136] A multi-resonant coupled intelligent photoacoustic cavity module is used for target gas absorption and acoustic wave generation; it includes a main resonant cavity integrating a multi-zone independently controlled piezoelectric actuator array, an outer anti-interference buffer cavity, and an inner layer integrating microchannels and optical waveguides.

[0137] The multi-physics field fusion sensing module is used to acquire the first photoacoustic signal, the first sound field signal, and the first deformation signal generated by the interaction of modulated excitation light with the target gas; it is used to synchronously sense the multidimensional response of the system, including a distributed fiber optic acoustic wave sensor array for measuring the sound field distribution in the cavity, a differential photothermal deformation sensor for measuring the photothermal deformation gradient of the cavity, and a high-Q acoustic sensor for calibrating the sound pressure signal.

[0138] The signal processing and state observation module is used to calculate the first state vector based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, combined with a preset digital twin model.

[0139] The multi-objective optimization decision and control module is used to take the first state vector as input, solve based on the preset multi-objective optimization function, generate a first control command subset for adjusting the tunable laser excitation module and a second control command subset for adjusting the multi-resonant coupled intelligent photoacoustic cavity module; and determine the current concentration inversion mode and calculate the output gas concentration value based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector.

[0140] In this embodiment, the static operating point mismatch problem is fundamentally solved by a dynamic dual-closed-loop feedback mechanism based on a digital twin model and nonlinear state observation. By constructing a digital twin model that can reflect the light, heat, and sound conversion links in real time, and integrating multi-physics field sensing data such as photoacoustic signals, sound field distribution, and thermal deformation, the first state vector characterizing the current system's true performance is calculated in real time by a nonlinear state observer. This vector includes the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index. Using a multi-objective optimization function to simultaneously optimize the signal-to-noise ratio, linearity, and control energy consumption, collaborative control commands are generated in a rolling manner to precisely adjust the excitation light modulation parameters and cavity acoustic mode parameters. This achieves dual dynamic tracking and locking of the laser modulation frequency and the cavity acoustic resonance frequency, giving the system self-locking capability. When external conditions change, the system's operating point can be automatically and continuously pulled to the Pareto optimal state under the current conditions. Thus, it maintains detection sensitivity and signal stability close to the theoretical limit within a wide range of environmental changes and concentrations, improving adaptability and robustness.

[0141] Using the calculated system nonlinearity index as an intrinsic criterion, the concentration inversion mode is intelligently switched. In the low-concentration linear region, a high-sensitivity signal amplitude detection is adopted. When the index increases, indicating nonlinearity, it smoothly transitions to a fusion algorithm that integrates the signal amplitude and the resonant frequency offset. In the deep nonlinear region, the detection mode is switched to a detection mode that uses the core parameters of the control commands required to maintain the cavity resonance tracking state as the main observation quantity, perfectly avoiding the signal saturation problem. This achieves seamless, continuous, and wide dynamic range measurement from the ppb level to the percentage level. At the same time, by independently acquiring and analyzing the first deformation signal reflecting the local cavity, the ineffective thermal disturbance caused by non-target absorption or adsorption effects can be directly quantified. In the nonlinear state observer, the first deformation signal is used as a key input to model the mixed photoacoustic signal and subtract the background interference component in real time, thereby accurately separating the effective absorption coefficient that is only related to the target gas. This greatly enhances the gas selectivity of the system in complex background gases or polluted environments, making the detection results more accurate and reliable.

[0142] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the self-locking wave method and system based on laser photoacoustic spectroscopy gas detection as described above.

[0143] The methods and systems according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the self-locking wave method and system based on laser photoacoustic spectroscopy gas detection provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components in the electronic device shown in this application may be omitted according to actual needs.

[0144] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0145] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A self-locking wave method based on laser photoacoustic spectroscopy for gas detection, characterized in that, The method includes: The target gas is introduced into the multi-resonant coupled smart photoacoustic cavity integrated with a piezoelectric actuator array to obtain the first photoacoustic signal generated by the interaction of modulated excitation light and target gas, the first sound field signal reflecting the spatial distribution of the sound field in the cavity, and the first deformation signal reflecting the local thermal deformation of the cavity. Based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, and combined with a preset digital twin model, a first state vector containing the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index is calculated; the construction of the digital twin model includes: During the initialization phase, a standard gas is introduced into the multi-resonant coupled smart photoacoustic cavity to perform parameter scanning, causing the excitation light modulation parameters to vary in the first range and the cavity acoustic mode parameters to vary in the second range. During the parameter scanning process, multiple sets of scanning parameter combinations and the corresponding photoacoustic signals, sound field signals and deformation signals under each set of parameter combinations are collected and recorded simultaneously to form training data pairs. Using training data pairs, a mathematical model that can map the relationship between input parameters, system state, and output signal is obtained through machine learning algorithms as a digital twin model; the input parameters refer to the excitation light modulation parameters and cavity acoustic mode parameters, the system state refers to the effective absorption coefficient of the first state vector, the acoustic-optical coupling efficiency factor, and the system nonlinearity index, and the output signal refers to the first photoacoustic signal, the first sound field signal, and the first deformation signal; Using the first state vector as input, a solution is obtained based on a preset multi-objective optimization function to generate a first subset of control instructions for adjusting the excitation light modulation parameters and a second subset of control instructions for adjusting the cavity acoustic mode parameters. The excitation light modulation parameters are adjusted according to the first control command subset, and the cavity acoustic mode parameters are adjusted according to the second control command subset. Based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector, the current concentration inversion mode is determined. The concentration inversion mode includes a linear mode, a nonlinear transition mode, and a deep nonlinear mode. The optimal observation and inversion method is adopted under different nonlinearities, and the output gas concentration value is calculated.

2. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 1, characterized in that, The calculation of the first state vector, which includes the effective absorption coefficient, the acoustic-optical coupling efficiency factor, and the system nonlinearity index, includes: The background interference intensity value is obtained by performing digital-to-analog conversion on the first deformation signal; The first photoacoustic signal, the first sound field signal, and the first deformation signal are currently acquired. Feature extraction is performed to obtain the first feature vector. Obtain the currently used excitation light modulation parameters and cavity acoustic mode parameters, and input them together with the first feature vector as input data into the trained digital twin model; The input data is processed based on the digital twin model, and the output is an estimate of the hidden state of the current system as the first state vector.

3. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 1, characterized in that, The generation of a first subset of control commands for adjusting the excitation light modulation parameters and a second subset of control commands for adjusting the cavity acoustic mode parameters includes: The sub-terms of the multi-objective optimization function include a concentration prediction error term, a signal quality reciprocal term, a system nonlinearity exponent term, and a control energy consumption term. Using a predictive control algorithm with a digital twin model as the internal predictive model, and aiming to minimize the multi-objective optimization function within a future time window, rolling optimization calculations are performed. By solving the rolling optimization calculation, a set of future control sequences is obtained. The instruction of the first control cycle in the future control sequence is selected as the first control instruction subset and the second control instruction subset at the current moment.

4. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 1, characterized in that, The process of determining the current concentration inversion mode and calculating the output gas concentration value includes: Set a first preset threshold and a second preset threshold, with the second preset threshold being greater than the first preset threshold. Obtain the system nonlinearity index in the first state vector and compare it with the preset threshold. When the system nonlinearity index is lower than the first preset threshold, it is determined to be in linear mode. The main frequency amplitude of the second photoacoustic signal after phase-locked amplification is used as the main observation value to calculate the gas concentration value. When the system nonlinearity index is higher than the first preset threshold but lower than the second preset threshold, it is determined to be a nonlinear transition mode. The resonant frequency offset of the multi-resonant coupled smart photoacoustic cavity is obtained and fused with the main frequency amplitude of the second photoacoustic signal as the main observation value to calculate the gas concentration value. When the system nonlinearity index is higher than the second preset threshold, it is determined to be a deep nonlinear mode. The core frequency parameters used to maintain the cavity resonance tracking state are obtained from the second control instruction subset. The core frequency parameters are used as the main observation values ​​for concentration inversion to calculate the gas concentration value.

5. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 2, characterized in that, The calculation of the effective absorption coefficient includes: After performing digital-to-analog conversion on the first deformation signal, a scalar value characterizing the gradient of the non-uniform thermal deformation distribution on the cavity surface is obtained, which is denoted as the background interference intensity value. The background interference intensity value is converted into a corresponding equivalent background absorption signal amplitude; The net signal amplitude obtained by subtracting the equivalent background absorption signal amplitude from the total absorption signal amplitude obtained by spectral analysis of the first photoacoustic signal is used to finally calculate the effective absorption coefficient.

6. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 1, characterized in that, The step of adjusting the excitation light modulation parameters according to the first subset of control commands includes: The target laser wavelength offset value, target laser modulation frequency value, and target laser modulation depth value are parsed from the first subset of control commands. The center wavelength of the output laser is adjusted according to the target laser wavelength offset value, and the modulation frequency and modulation depth of the laser are adjusted according to the target laser modulation frequency value and the target laser modulation depth value.

7. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 6, characterized in that, The adjustment of cavity acoustic mode parameters according to the second set of control commands includes: The second control instruction subset includes a set of multiple control signal parameters for driving multiple actuation units on the cavity, each control signal parameter including drive amplitude, drive phase and drive frequency; Based on the multi-channel control signal parameters, corresponding multi-channel high-voltage drive signals are generated and applied to the corresponding actuation units respectively; By setting different driving phase combinations, the cavity is excited to generate axial, radial, and hybrid acoustic resonance modes. By setting the driving frequency, the resonant frequency of the selected acoustic resonance mode is tuned.

8. The self-locking wave method based on laser photoacoustic spectroscopy for gas detection according to claim 1, characterized in that, The method further includes: During continuous system operation, the first photoacoustic signal, the second photoacoustic signal, the first deformation signal, the first control command subset, the second control command subset, the excitation light modulation parameters, the cavity acoustic mode parameters, and the first state vector are periodically acquired and stored as data samples in the rolling historical database. Periodically calculate the average error between the predicted signal output by the digital twin model and the real signal across all recent data samples; When the average error exceeds the preset error threshold, an update process is triggered, using the latest data samples from the rolling historical database to perform an incremental learning and fine-tuning of the parameters of the digital twin model.

9. A self-locking wave system based on laser photoacoustic spectroscopy gas detection, used to implement the self-locking wave method based on laser photoacoustic spectroscopy gas detection as described in any one of claims 1-8, characterized in that, The system includes: A tunable laser excitation module is used to generate a probe laser with tunable wavelength and composite modulation. Multi-resonant coupled intelligent photoacoustic cavity module, used for target gas absorption and sound wave generation; A multi-physics field fusion sensing module is used to acquire a first photoacoustic signal, a first acoustic field signal, and a first deformation signal generated by the interaction of modulated excitation light with the target gas. The signal processing and state observation module is used to calculate the first state vector based on the first photoacoustic signal, the first sound field signal, and the first deformation signal, combined with a preset digital twin model. The multi-objective optimization decision and control module is used to take the first state vector as input, solve based on the preset multi-objective optimization function, generate a first control command subset for adjusting the tunable laser excitation module and a second control command subset for adjusting the multi-resonant coupled intelligent photoacoustic cavity module; and determine the current concentration inversion mode and calculate the output gas concentration value based on the second photoacoustic signal obtained after adjustment and the system nonlinearity index in the first state vector.

Images