Method and system for detecting trace organic pollutants in water body

Abstract

The invention discloses a method and system for detecting trace organic pollutants in a water body, and relates to the field of environmental water body detection.The method comprises the steps that environmental parameters of a target water body sample to be detected are collected, and baseline spectral data of the target water body sample are obtained; determining disturbance operation parameters matched with the target water body sample, and applying micro-disturbance to the target water body sample according to the disturbance operation parameters to obtain disturbance response spectral data; performing time sequence differential analysis on the disturbance response spectral data and the baseline spectral data to obtain a disturbance response residual error spectrum, and identifying an optimal response characteristic wavelength from a characteristic response wavelength interval of the target trace organic pollutants based on the disturbance response residual error spectrum; and solving the concentration correction model according to the optimal response characteristic wavelength, the disturbance operation parameters and the environmental parameters to obtain the concentration of the target trace organic pollutants. Therefore, the detection sensitivity and the anti-interference capability of trace organic pollutants in the water body sample are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental water detection, and particularly relates to a water trace organic pollutant detection method and system. BACKGROUND

[0002] With the accelerated industrialization and urbanization, various trace organic pollutants (such as phenol, polycyclic aromatic hydrocarbons, pesticide residues, etc.) continuously enter surface water and groundwater, which have the characteristics of strong biological toxicity, high environmental migration, and difficult natural degradation, and may pose a serious threat to the safety of the ecological system and human health even at trace levels. Therefore, how to detect trace organic pollutants in water with high sensitivity and high accuracy has become a technical problem to be solved in the fields of environmental science and public health.

[0003] At present, the composition of environmental water is extremely complex, and there are a large number of natural organic matter, colloid, inorganic ion and other coexisting components, which not only have a high overlap with the target pollutants in the optical and chemical signal levels, but also significantly affect the stability and specificity of the overall detection signal with dynamic changes in environmental conditions (such as temperature, pH, turbidity, etc.). In addition, the concentration of the target pollutant is often at an ultra-low level, and its signal is easily overwhelmed by complex background noise and coexisting components, resulting in a decrease in sensitivity, an increase in false positive and false negative rates in the actual detection process.

[0004] In addition, the spatial and temporal variation characteristics of environmental water are obvious, and the background conditions and interference sources of different sampling points and different sampling times are quite different, which makes the traditional detection method face a series of technical challenges such as difficulty in effectively correcting background interference, insufficient signal specificity, and limited detection accuracy and stability in actual application. SUMMARY

[0005] The present application provides a water trace organic pollutant detection method, system, storage medium, computer program product and electronic device, which at least solves the problem that the trace organic pollutant signal is easily overwhelmed by the water background, impurities and environmental noise in the related art, resulting in low detection sensitivity and poor anti-interference ability.

[0006] In a first aspect, an embodiment of the present application provides a water body trace organic pollutant detection method, the method comprising: collecting an environmental parameter of a target water body sample to be detected, and obtaining baseline spectral data of the target water body sample; determining a perturbation operation parameter matched with the target water body sample, and applying a micro-perturbation to the target water body sample according to the perturbation operation parameter to obtain perturbation response spectral data; the perturbation operation parameter comprises a perturbation operation type, a perturbation operation amplitude, and a perturbation operation duration; performing time series difference analysis on the perturbation response spectral data and the baseline spectral data to obtain a perturbation response residual spectrum, and identifying an optimal response characteristic wavelength from a characteristic response wavelength interval of a target trace organic pollutant based on the perturbation response residual spectrum; solving a concentration correction model according to the optimal response characteristic wavelength, the perturbation operation parameter, and the environmental parameter to obtain a concentration of the target trace organic pollutant; the concentration correction model is expressed by the following formula:

[0007] C target = f (△S (λ * ,t), A perturb ,E env ),

[0008] In the formula, C target is the concentration of the target trace organic pollutant, T is a perturbation time interval, △S (λ * ,t) is a response intensity of the perturbation response residual spectrum at the optimal response characteristic wavelength λ * and a perturbation time t, A perturb is the perturbation operation parameter, E env is the environmental parameter, and f (·) is a pre-trained concentration correction model function.

[0009] In a second aspect, an embodiment of the present application provides a water body trace organic pollutant detection system, the system comprising: a water body sample collection unit configured to collect an environmental parameter of a target water body sample to be detected, and obtain baseline spectral data of the target water body sample; a sample micro-perturbation unit configured to determine a perturbation operation parameter matched with the target water body sample, and apply a micro-perturbation to the target water body sample according to the perturbation operation parameter to obtain perturbation response spectral data; the perturbation operation parameter comprises a perturbation operation type, a perturbation operation amplitude, and a perturbation operation duration; an optimal wavelength identification unit configured to perform time series difference analysis on the perturbation response spectral data and the baseline spectral data to obtain a perturbation response residual spectrum, and identify an optimal response characteristic wavelength from a characteristic response wavelength interval of a target trace organic pollutant based on the perturbation response residual spectrum; a trace concentration identification unit configured to solve a concentration correction model according to the optimal response characteristic wavelength, the perturbation operation parameter, and the environmental parameter to obtain a concentration of the target trace organic pollutant; the concentration correction model is expressed by the following formula:

[0010] C target = f (△S (λ * ,t), A perturb ,E env ),

[0011] In the formula, C target is the concentration of the target trace organic pollutants, T is the time interval of the disturbance, △S (λ * ,t) is the response intensity of the disturbance response residual spectrum at the optimal response characteristic wavelength λ * and the disturbance time t, A perturb is the disturbance operation parameter, E env is the environmental parameter, and f (·) is the pre-trained concentration correction model function.

[0012] In a third aspect, an electronic device is provided, which includes at least one processor, and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the water body trace organic pollutant detection method of any embodiment of the present application.

[0013] In a fourth aspect, an embodiment of the present application provides a storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the steps of the water body trace organic pollutant detection method of any embodiment of the present application.

[0014] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the water body trace organic pollutant detection method of any embodiment of the present application.

[0015] The water body trace organic pollutant detection method and system provided by the present application can at least produce the following technical effects:

[0016] (1) By applying a precise controllable micro-disturbance to the water sample, the kinetic response of the target pollutant molecules is actively excited by setting the disturbance operation parameters (including the disturbance type, amplitude and duration), and the stable background signal is efficiently removed and the weak target signal is significantly amplified by combining the time sequence difference between the baseline spectrum and the disturbance response spectrum, thereby effectively improving the response sensitivity of the detection system to low-concentration pollutants.

[0017] (2) By time-difference between the baseline spectrum before disturbance and the response spectrum after disturbance, the net response characteristics of the pollutants after disturbance can be effectively extracted, and the influence of system background drift and irrelevant noise can be eliminated. Further, by comparing and analyzing each candidate wavelength in the characteristic response wavelength range of the target trace organic pollutants in the disturbance response residual spectrum, the best response wavelength most matched with the pollutant molecular characteristics can be automatically determined, thereby greatly enhancing the distinguishing ability between similar spectral lines, avoiding cross interference caused by wavelength overlap, and improving the detection specificity.

[0018] (3) Considering that the pollutant response is not only affected by its own concentration, but also jointly regulated by the disturbance mode and environmental conditions, an adaptive multi-factor coupling quantitative model is constructed with disturbance response intensity, disturbance operation parameters and environmental parameters as independent variables. Through pre-training strategy, the dynamic estimation and error compensation of the concentration of pollutants under different environmental backgrounds and disturbance operations can be realized, and the high stability and high accuracy of the detection results under various complex water body conditions are realized.

[0019] Through the technical scheme, by applying a micro-disturbance and analyzing the disturbance response residual spectrum, combined with the extraction of the optimal response characteristic wavelength, the concentration of trace organic pollutants in water can be accurately identified and quantitatively analyzed. At the same time, through the dynamic acquisition of environmental parameters and the adjustment of disturbance operation parameters, the influence of environmental condition changes on the detection signal can be corrected in real time, the false positive and false negative rates are reduced, and the stability and accuracy of the detection are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0021] Figure 1 A flow chart of an example of a water body trace organic pollutant detection method according to an embodiment of the present application is shown;

[0022] Figure 2 An operation flow chart of an example of determining disturbance operation parameters matched with a target water sample according to an embodiment of the present application is shown;

[0023] Figure 3 An example of a simulation effect schematic diagram of ultraviolet absorption spectrum response of phenolic pollutants under physical micro-oscillation is shown;

[0024] Figure 4A simulation effect schematic diagram showing an example of Raman spectrum response of organic amine trace pollutants under different pH conditions is shown.

[0025] Figure 5 A simulation effect schematic diagram showing an example of fluorescence spectrum response of polycyclic aromatic hydrocarbon trace organic pollutants under different ionic strength conditions is shown.

[0026] Figure 6 An operation flow chart showing an example of identifying an optimal response characteristic wavelength based on a perturbation response residual spectrum according to an embodiment of the present application is shown.

[0027] Figure 7 A prediction error distribution comparison simulation diagram showing an example of different kernel function models in trace organic pollutant concentration prediction experiments is shown.

[0028] Figure 8 A structural block diagram showing an example of a water body trace organic pollutant detection system according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0029] To make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0030] It should be noted that, with the continuous development of biological materials and intelligent technology, some effective water body trace organic pollutant detection directions have appeared in the current related technologies, but there are still some deficiencies.

[0031] Specifically, laboratory analysis technologies such as gas chromatography-mass spectrometry (GC-MS) and high-performance liquid chromatography-mass spectrometry (HPLC-MS) are widely used in qualitative and quantitative analysis of trace organic pollutants due to their high sensitivity and selectivity. Such technologies can achieve a detection limit of ppb (parts per billion) or even lower, but in actual application, they have the disadvantages of complicated sample pretreatment process, long analysis period, high equipment cost, and high operation technical requirements, which are difficult to meet the real-time, wide-range and high-throughput on-site detection requirements in water environment monitoring.

[0032] Spectroscopy-based non-destructive testing methods, such as absorption spectroscopy, fluorescence spectroscopy, and Raman spectroscopy, have been increasingly applied to rapid on-site screening and dynamic monitoring of water pollutants in recent years due to their advantages, including fast detection speed, ability to achieve online or in-situ automated monitoring, and high degree of information integration. However, in complex aquatic environments, these methods are often affected by factors such as multi-component background interference, high overlap of spectral features, instrument baseline drift, and fluctuations in environmental parameters. This leads to reduced detection sensitivity and increased false positive or false negative rates, making it difficult to achieve highly reliable identification of pollutants at extremely low concentrations.

[0033] With the development of big data and artificial intelligence technologies, some detection systems have begun to incorporate algorithms such as machine learning and pattern recognition to improve signal analysis capabilities and the accuracy of detection results. These methods can achieve efficient deconstruction of complex spectral signals and quantitative analysis of pollutants under specific sample conditions. However, their algorithm performance is highly dependent on the representativeness and quality of the training data, and in real-world environments, they are easily affected by factors such as fluctuations in the quality of acquired signals, insufficient sample labeling, and limited model generalization ability. Therefore, their practical application effectiveness and stability are still insufficient to meet the high demands of on-site environmental monitoring.

[0034] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0035] Figure 1 A flowchart illustrating an example of a method for detecting trace organic pollutants in water according to an embodiment of this application is shown.

[0036] Regarding the execution subject of the method in the embodiments of this application, it can be any controller or processor with computing or processing capabilities. By introducing a combination of perturbation operation and time-series differential analysis, the signal processing flow is optimized and an accurate calibration model is established, which can effectively improve the detection sensitivity and accuracy of trace organic pollutants in water.

[0037] In some examples, it can be integrated into an electronic device or terminal through software, hardware, or a combination of both, and the type of terminal or electronic device can be diverse, such as mobile phones, tablets, or desktop computers, etc.

[0038] like Figure 1 As shown, in step S110, environmental parameters of the target water sample to be detected are collected, and baseline spectral data of the target water sample are obtained.

[0039] Here, first, the environmental parameters of the target water sample need to be collected, including at least one of the following: water temperature, pH value, conductivity, turbidity, or dissolved oxygen concentration. Environmental parameters have different effects on the stability and characteristics of spectral signals. For example, changes in temperature can affect the concentration of dissolved oxygen in water, thereby changing the spectral characteristics of pollutants; changes in pH value can affect the molecular structure and chemical reactivity of organic pollutants. By obtaining these environmental parameters, the effects of different environmental factors can be effectively compensated.

[0040] In addition, baseline spectral data of the water sample are collected by a high-resolution spectrometer. The baseline spectrum reflects the spectral response of the water body without disturbance, representing the natural spectral response of the water body without external disturbance, mainly including the optical absorption characteristics of all components in the water body (such as water, dissolved organic matter, etc.). By analyzing these baseline data, the spectral characteristics of the water body without disturbance can be understood, and accurate reference standards are provided for subsequent disturbance response analysis and correction.

[0041] In step S120, the disturbance operation parameters matched with the target water sample are determined, and a micro-disturbance is applied to the target water sample according to the disturbance operation parameters to obtain the disturbance response spectral data.

[0042] Here, the appropriate disturbance operation parameters are selected according to the characteristics of the target pollutant, including disturbance operation type, disturbance operation amplitude, and disturbance operation duration. The disturbance operation type can be, for example, temperature change, pH adjustment, dissolved oxygen concentration change, etc., which can vary for different types of pollutants to be detected, aiming to selectively amplify the spectral response of the target pollutant to be detected.

[0043] In some examples of the embodiments of the present application, the disturbance operation type includes at least one of the following: physical micro-oscillation, pH micro-adjustment, ion strength micro-change, or light intensity periodic fluctuation.

[0044] Specifically, the physical micro-oscillation can adopt micro-magnetic stirring, ultrasonic micro-disturbance, low-frequency vibrator, etc., to excite the internal fluid dynamics response of the water sample, break the micro-environment static state, and strengthen the physical separation of pollutants and the background components of the water body. The pH micro-adjustment can be achieved by automatic titration or micro-acid-base injection to realize dynamic fluctuation of pH in a very small range, excite the characteristic response of pH-sensitive organic pollutants, and highlight their chemical intrinsic differences. The ion strength micro-change can be automatic addition of micro-amount of salt (such as NaCl, KCl, etc.) to adjust the ion strength of the water body, induce the dissolution, complexation or adsorption state transition of the target organic pollutant, and amplify its spectral characteristics. The light intensity periodic fluctuation can be the periodic intensity change of the control detection light source or the auxiliary excitation light source, which is suitable for Raman, fluorescence, etc. that need to be excited and responded, highlighting the excited state dynamics characteristics of the target substance.

[0045] It should be noted that the molecular structure, energy level distribution, polarity, and sensitivity to external physical and chemical environment of different types of trace pollutants (such as phenols, polycyclic aromatic hydrocarbons, organic amines, etc.) are essentially different. In addition, different types of molecules often have different disturbance types that are most strongly responsive to "excitation". For example: phenolic pollutants are highly sensitive to physical micro-vibration because they are easily broken by disturbance to break intermolecular hydrogen bonds and distribution state; polycyclic aromatic hydrocarbon pollutants (such as naphthalene, phenanthrene) are easily controlled by ion strength, and micro-salt disturbance is required; organic amine pollutants are extremely sensitive to pH environment, and pH fine-tuning can be selected as the main disturbance method. Therefore, the disturbance operation type can be customized according to the molecular recognition characteristics of the pollutants, and the "excitation channel" can be selected to maximize the signal selectivity of the target pollutants, while the main background signal remains basically unchanged, so as to improve the detection sensitivity.

[0046] In addition, the disturbance operation amplitude (i.e. the strength of the disturbance) and the disturbance operation duration (i.e. the application time of the disturbance) are aimed at exciting the change of the pollutant molecules in the spectral response, so that they can show characteristics that cannot be identified in the undisturbed case. For the same type of trace organic pollutants, they have similar response mechanisms to physical or chemical disturbance, and the same disturbance operation type can be used. However, the optimal disturbance operation amplitude and disturbance duration (i.e. disturbance strength, action time window) need to be optimized for different water samples, different backgrounds, and different instrument states, in order to effectively enhance the spectral response of the trace pollutants in the current water sample being detected, and effectively improve the accuracy and reliability of the detection.

[0047] In step S130, the disturbance response spectrum data and the baseline spectrum data are subjected to time series difference analysis to obtain a disturbance response residual spectrum, and an optimal response characteristic wavelength is identified from the characteristic response wavelength interval of the target trace organic pollutant based on the disturbance response residual spectrum.

[0048] Here, the difference between the disturbance response spectrum and the baseline spectrum reflects the response characteristics of the pollutants in the water body to the disturbance operation. Through time series difference analysis, the interference of environmental background noise and natural components (such as natural organic matter and inorganic ions) in the water body can be effectively removed, the background interference can be effectively removed, and the spectral characteristic response of the pollutants affected by the disturbance operation can be highlighted.

[0049] Further, based on the disturbance response residual spectrum, an optimal response characteristic wavelength is identified from the characteristic response wavelength interval of the target trace organic pollutant. Illustratively, based on the response peak in the residual spectrum, the most representative and sensitive response wavelength can be selected from a plurality of candidate wavelengths, which can provide the greatest degree of pollutant signal, ensuring that the target pollutant can be efficiently distinguished and its characteristics accurately identified in complex water quality, and the specificity and sensitivity of the signal are enhanced.

[0050] It should be noted that each kind of trace organic pollutant has one or more "characteristic response wavelength intervals" with high specificity and high sensitivity, which often corresponds to the molecular absorption peak, emission peak or scattering characteristics of the pollutant in the spectral domain of ultraviolet, visible, fluorescence or Raman, etc. For example, the main ultraviolet absorption peak of phenol is usually near 270 nm, polycyclic aromatic hydrocarbons have multiple absorption peaks in the range of 220-300 nm, certain pesticides will produce fluorescence emission at a specific wavelength under specific excitation, etc.

[0051] Although the above-mentioned characteristic response interval locks the "molecular fingerprint" range of a specific pollutant, the actual spectral response in this interval does not show maximum, most stable or most representative signal everywhere. In addition, only at a certain specific wavelength point, the response signal is the strongest, the noise is the smallest, and the sensitivity is the best, which can be used as the optimal position for "quantitative detection". In addition, due to the influence of water background, instrument system, other impurities, etc., occasional noise, background interference or overlapping false peaks are easy to appear, and other wavelengths in the characteristic response interval often have more background noise or non-specific interference. The overall operation is easy to dilute the target signal, lower the detection limit, and even bring false positives / false negatives, so by selecting the optimal response characteristic wavelength, the accuracy and robustness of concentration quantification can be greatly improved.

[0052] In step S140, a concentration correction model is solved according to the optimal response characteristic wavelength, the perturbation operation parameter and the environmental parameter, so as to obtain the concentration of the target trace organic pollutant.

[0053] Here, the core task of the concentration correction model is to establish a mathematical model that can quantitatively map the perturbation response signal and the concentration of the target pollutant by solving the identified optimal response characteristic wavelength, the perturbation operation parameter and the environmental parameter. The input related signal data is mapped to the concentration value of the corresponding pollutant.

[0054] Exemplarily, the concentration correction model is expressed by the following formula:

[0055] C target = f (△S (λ * ,t), A perturb , E env ), formula (1)

[0056] In the formula, C target is the concentration of the target trace organic pollutant, T is the perturbation time interval, △S(λ * ,t) is the response intensity of the perturbation response residual spectrum at the optimal response characteristic wavelength λ * and the perturbation time t, A perturb is the perturbation operation parameter, E env is the environmental parameter, and f(·) is the pre-trained concentration correction model function.

[0057] Specifically, the concentration correction model can be trained based on a large amount of experimental data or historical sample data. It can automatically correct signal fluctuations according to disturbance response and environmental changes, adapt to different water environments, and accurately estimate the concentration of pollutants.

[0058] It should be understood that the type of concentration correction model function can be diverse and will not be limited here. For example, the concentration correction model can employ multiple regression models (such as linear regression, support vector regression, Lasso regression, etc.) or machine learning methods (such as neural networks, random forests, etc.) to map known perturbation responses, characteristic wavelengths, and environmental parameters to the concentration of the target pollutant. Furthermore, there is generally a nonlinear relationship between pollutant concentration and response signal (especially when pollutant concentration is low), and various non-restricted nonlinear models (such as support vector machine regression, random forests, neural networks, etc.) can be used to set the concentration correction model function.

[0059] Through the embodiments of this application, by comprehensively utilizing perturbation operation, time-series difference analysis, optimal response wavelength identification, and multi-factor concentration correction model, and by flexibly adjusting perturbation operation parameters and dynamically selecting response characteristic wavelengths, the sensitivity, specificity, and stability issues of trace organic pollutant detection in environmental water bodies are effectively solved. It can also provide accurate and reliable concentration data for detection scenarios with different pollutants and environmental conditions.

[0060] Figure 2 A flowchart illustrating an example of determining perturbation operation parameters that match a target water sample, according to an embodiment of this application, is shown.

[0061] like Figure 2 As shown, in step S210, the target pollutant type of the target trace organic pollutant is obtained.

[0062] It should be noted that different types of pollutants have different molecular structures, physicochemical properties, and response mechanisms to disturbances. Only by first identifying the type of pollutant can we ensure that the excitation of disturbances is specific and efficient.

[0063] Here, the methods for obtaining the target pollutant type can be diverse, such as specifying it based on user input or automatic identification. For example, the system automatically parses the type of pollutant to be detected by analyzing the detection task sheet, such as "phenols," "polycyclic aromatic hydrocarbons," or "organic amines."

[0064] In step S220, a set of calibration parameters matching the target pollutant type is determined from a preset pollutant-calibration parameter set. The pollutant-calibration parameter set pre-stores multiple pollutant types and corresponding calibration parameter sets.

[0065] Here, the preset pollutant-calibration parameter set establishes a mapping relationship between the pollutant type and the optimal disturbance parameter, and by directly calling the pre-stored parameter, the system can quickly obtain the matching disturbance condition, thereby improving the detection efficiency and consistency. An example of the mapping relationship is as follows:

[0066] Phenols: physical micro-vibration, 20-35 Hz, 60-80 seconds;

[0067] Polycyclic aromatic hydrocarbons: slight change in ionic strength, NaCl 0.5-0.8 mM, stirring for 25-35 seconds;

[0068] Organic amines: slight adjustment of pH, pH = 8.0-9.0, slow titration for 2-3 minutes.

[0069] In this way, the system can automatically match the corresponding calibration parameter group according to the detection task sheet, and the calibration parameter group can define multiple parameter combinations of disturbance operation amplitude and disturbance operation duration. Thus, the debugging and pre-calibration time can be greatly shortened, the detection throughput and engineering efficiency can be improved, and the disturbance operation can be highly standardized and reproducible under different detection batches and different sites.

[0070] In step S230, based on the multiple calibration disturbance operation parameters in the calibration parameter group, the target water body sample is subjected to a pre-disturbance operation.

[0071] It should be noted that for different types of pollutants, the matched exclusive disturbance type is different; for the same type of pollutant, the disturbance type is consistent, and the parameters can be adaptively optimized. Here, each calibration disturbance operation parameter respectively defines a corresponding disturbance operation type, a disturbance operation amplitude, and a disturbance operation duration. Specifically, to adapt to the differences in water body background and pollutant components, multiple groups of pre-defined disturbance operation parameters are used for pre-experiment to form a standardized and engineered parameter screening process.

[0072] Exemplarily, the calibration parameter group is composed of multiple disturbance parameter combinations (T i ,A i ,D i ), where T i represents the disturbance operation type (such as physical micro-vibration, pH adjustment, etc.), A i represents the disturbance operation amplitude (such as vibration amplitude, pH amplitude, etc.), and D i represents the disturbance operation duration. For each group of parameters, the system automatically controls the corresponding disturbance execution unit to sequentially apply each group of pre-disturbance operations in multiple water body samples. For example, for the detection of phenols, differential vibration frequency and vibration time can be implemented on each water body sample to achieve differential disturbance operation.

[0073] In step S240, for each calibration perturbation operation parameter, the perturbation response residual spectrum corresponding to the calibration perturbation operation parameter is obtained, and the maximum target response of the perturbation response residual spectrum is determined in the target trace organic pollutant characteristic wavelength interval and the perturbation time interval, and the maximum background response is determined in the background wavelength interval irrelevant to the target trace organic pollutant and the perturbation time interval.

[0074] Here, after each set of perturbation parameters is applied, the perturbation response spectrum is collected in real time, and the maximum target response and the maximum background response are extracted by differentiating the perturbation response spectrum from the baseline spectrum, so as to reflect the sensitivity of the target pollutant and the system noise characteristics.

[0075] Specifically, under the i th set of perturbation parameters, the perturbation response spectrum S pert,i (λ, t) is collected, and the baseline spectrum S base (λ, t0) is obtained, so as to calculate the perturbation response residual spectrum:

[0076] △S i (λ, t) = S pert,i (λ, t) - S base (λ, t0), formula (2)

[0077] In the target pollutant characteristic wavelength interval Λ target and the perturbation time interval T, the maximum target response is determined:

[0078]

[0079] In the background irrelevant wavelength interval Λ bg and the perturbation time interval T, the maximum background response is determined:

[0080]

[0081] Here, the maximum value extraction method is used to fully avoid the influence of accidental perturbation and noise, highlight the real physical signal under the perturbation-response mechanism, and realize the accurate quantification of the response sensitivity of each perturbation parameter and the system noise intensity.

[0082] In step S250, the ratio of the maximum target response to the maximum background response of each calibration perturbation operation parameter is calculated to obtain the target correlation score.

[0083] Here, the target response and the background response are normalized in the form of ratio, highlighting the parameter set with the optimal "signal-to-noise ratio", and ensuring that the screened perturbation parameters can greatly stimulate the target pollutant response and maximize the background noise suppression.

[0084] Specifically, for each set of perturbation parameters, the target correlation score is calculated:

[0085]

[0086] wherein ε is a very small positive number to prevent the denominator from being zero; all G i The target correlation score vector is recorded.

[0087] Thus, the target correlation score is calculated by optimizing the signal-to-noise ratio, so that the system can automatically avoid high-noise interference areas, and improve the target pollutant detection capability and method disturbance resistance.

[0088] In step S260, the disturbance operation parameter corresponding to the maximum target correlation score is selected as the disturbance operation parameter matched with the target water sample.

[0089] Here, the disturbance parameter group with the highest target correlation score is selected by global screening as the optimal disturbance strategy for the current sample.

[0090] Specifically, the optimal disturbance parameter group is selected by the maximum value criterion:

[0091] (T * , A * , D * )=argmax i G i , formula (6)

[0092] Further, the optimal disturbance parameter group is automatically issued to the system control unit to implement the disturbance operation on the target water sample and the corresponding trace organic pollutant detection operation.

[0093] By the embodiments of the present application, the front-end disturbance screening mechanism based on multiple sets of calibration disturbance operation parameters can systematically apply a series of disturbance parameter combinations according to the differences in the physical and chemical backgrounds and potential pollutant components of different target water samples, and calculate the ratio of the maximum target response to the background response of the disturbance response residual spectrum of each parameter group. Thus, the optimization of all parameters is based on the objective spectral response (i.e., the quantitative ratio of the maximum signal response to the background noise in the characteristic wavelength interval), which ensures the accurate screening of the optimal disturbance operation strategy, realizes the maximum response enhancement of the target trace organic pollutant and minimization of the background interference, and significantly improves the signal-to-noise ratio and accuracy of the detection.

[0094] The technical implementation principle of the embodiments of the present application will be described below:

[0095] The presence of trace organic pollutants in water bodies is often influenced by solubility, complexation, adsorption, micro-environment, etc. The intrinsic spectral signals of the pollutants are often submerged by the main components, impurities and background noise of the water body, making it difficult to directly distinguish. The application of perturbation operations such as physical micro-vibration, pH adjustment, ion strength change or periodic light, etc. will essentially change the interaction between the pollutant molecules and their environment, causing systematic changes in the energy level structure, electron cloud distribution or dissociation state in the solution of the pollutant molecules. This change will lead to a repeatable and controllable dynamic response of the pollutant's absorption / fluorescence / Raman spectrum at a specific wavelength, while the background components have weak or almost no response to the same perturbation due to the stability of their physical and chemical properties.

[0096] Therefore, by perturbing the residual spectrum, the specific dynamic signal of the pollutant can be effectively revealed. For the same sample, the spectra collected before and after the perturbation / during the perturbation period are differentiated (i.e. residual), which can effectively eliminate the water body baseline, background and external interference, realize the difference amplification of the pollutant signal and background blanking. In addition, different perturbation operation parameters (type, amplitude, duration) and different pollutant molecules form a unique "perturbation-response" dynamic matching relationship. Such response contains the chemical fingerprint information of the pollutant itself.

[0097] Experimental Example 1: UV absorption enhancement of phenolic pollutants by physical micro-vibration.

[0098] Specifically, for a low-concentration phenol water sample, UV absorption spectra are collected without perturbation and under low-amplitude mechanical vibration (such as 20 Hz, micro-vibration for 1 minute).

[0099] Figure 3 An example of the UV absorption spectrum response simulation effect diagram of phenolic pollutants under physical micro-vibration is shown.

[0100] Specifically, micro-vibration can promote the more uniform distribution of phenol molecules and short-term exposure to the surface of the water body, resulting in a short-term increase in absorption intensity at the characteristic absorption wavelength (such as 270 nm), while the baseline and irrelevant components change little. As shown in Figure 3 The yellow solid line is the UV absorption spectrum of the unperturbed phenol sample, and the characteristic absorption peak of phenol (about 270 nm) is weak. The orange solid line is the UV absorption spectrum collected after applying micro-vibration (such as 20 Hz, 1 minute), and the characteristic absorption peak is significantly enhanced. The dashed line represents the "post-micro-vibration-unperturbed" residual spectrum, which clearly shows the dynamic response enhancement at the phenol characteristic wavelength.

[0101] Therefore, by Figure 3 It can be seen that physical micro-vibration can effectively amplify the spectral signal of phenol molecules, and residual analysis significantly improves the identifiability and quantitative sensitivity of trace pollutants in complex water body backgrounds.

[0102] Experimental Example 2: pH fine-tuning induces Raman signal change of organic amine.

[0103] Specifically, low-concentration organic amine water samples were collected at pH = 7.0 and pH = 8.5, respectively.

[0104] Figure 4 An example of the simulation effect diagram showing the Raman spectral response of organic amine trace pollutants under different pH conditions.

[0105] As shown in Figure 4 , the yellow curve represents the Raman spectrum of the water sample at pH = 7.0, showing weak characteristic Raman peaks (such as 1350 cm -1 and 1550 cm -1 ); the orange curve is the Raman spectrum of the same sample after the pH is adjusted to 8.5, and the intensity of the main characteristic peaks is significantly enhanced. The dashed line is the difference between the two (i.e. the "residual spectrum"), which produces a clear positive peak at the characteristic peak position, highlighting the amplification effect of pH fine-tuning on the Raman response of the target pollutant. The gray vertical line marks the characteristic peak position.

[0106] The results show that pH disturbance can induce changes in the Raman activity of target organic amine molecules, and residual analysis can significantly enhance the spectral recognition and quantitative sensitivity of pollutants in complex backgrounds.

[0107] Figure 5 An example of the simulation effect diagram showing the fluorescence spectral response of polycyclic aromatic hydrocarbon trace organic pollutants under different ionic strength conditions.

[0108] Specifically, the fluorescence intensity of polycyclic aromatic hydrocarbons (such as naphthalene) in water is easily affected by the ionic strength of the solution.

[0109] By slowly adding low-concentration inorganic salts to the water sample, the time-series change of the fluorescence spectrum was collected, and it was found that the characteristic peaks of the pollutants showed a repeatable kinetic response, while the fluorescence of the main background water body (such as the solvent itself) changed very weakly. As shown in Figure 5 , the yellow curve is the fluorescence spectrum of the water sample at the initial low ionic strength, and the 375 nm and 410 nm are the typical polycyclic aromatic hydrocarbon characteristic peaks; the orange curve is the fluorescence spectrum of the same sample after the ionic strength is increased by adding inorganic salts, and the characteristic peak intensity is significantly increased, indicating that the kinetic behavior of the pollutant molecules is excited by the disturbance. The dashed line is the difference between the two (i.e. the residual spectrum), which forms a clear positive peak at the two characteristic wavelengths, clearly highlighting the fluorescence enhancement signal of the target pollutant under the disturbance response, greatly weakening the influence of the background and other interfering components. The gray vertical line marks the characteristic peak position.

[0110] Therefore, by Figure 5The results verify that the spectral response of polycyclic aromatic hydrocarbons can be effectively amplified by ion intensity perturbation, and the residual analysis greatly improves the detection sensitivity and identification ability in complex water environment.

[0111] It should be noted that in real environment water, the signal of trace organic pollutants is often overwhelmed by the main background, impurities and instrument noise. The characteristic peaks of the original spectrum are often weak or even covered by the background. Direct analysis of the original spectrum not only has low sensitivity, but also is easy to produce false positive or false negative.

[0112] In the above simulation figures of each perturbation experiment, in order to visualize the "amplification" of the residual response, the ideal clean background is selected. However, in actual water environment and complex multi-component samples, the original spectrum peaks are often disordered, peak type drifts, and characteristic peaks are submerged. Direct detection is often ineffective. In contrast, through perturbation operation, the molecular dynamics behavior of trace pollutants can be actively excited without destroying the sample structure, so that the response is "amplified" in a specific wavelength range, and the main background component has little effect. Further, through residual analysis (i.e. after perturbation-before perturbation / control group), most of the background, system error and environmental noise unrelated to perturbation can be effectively "subtracted", and only the perturbation response signal itself is retained. Thus, the dynamic response of the pollutant molecule can also be presented in a high signal-to-noise ratio manner in a high noise and complex background, greatly reducing the misjudgment rate.

[0113] Figure 6 An operation flowchart of an example of identifying the optimal response characteristic wavelength based on the perturbation response residual spectrum according to an embodiment of the present application is shown.

[0114] As shown in Figure 6 , in step S610, in the characteristic response wavelength interval Λ tgt of the target trace organic pollutant, the main peak response is calculated based on the perturbation response residual spectrum for each candidate response wavelength.

[0115] As described above, each type of pollutant has a corresponding characteristic response wavelength interval Λ tgt , which can be obtained through literature data or historical measurement.

[0116] Exemplarily, Λ tgt = {λ1, λ2, …, λ n}, wherein λ1, λ2, …, λ n are the typical response wavelengths of the target pollutant, and the spectral characteristic region of the target pollutant is accurately locked.

[0117] Further, in the characteristic response wavelength interval, the response signal of the target pollutant is a main peak signal. Through each candidate wavelength point λ jThe maximum response is extracted to ensure that the most representative signal is selected as the basis for concentration quantification.

[0118] Specifically, the disturbance response residual spectrum △S(λ j ,t) collected after the disturbance operation is processed to extract the main peak response of each wavelength λ j The response intensity of each wavelength λ j is determined by the maximum value, that is, the maximum response value in the disturbance time sequence.

[0119] R(λ j )=max t∈T △S(λ j ,t), formula (7)

[0120] In the formula, △S(λ j ,t) is the response intensity at the candidate response wavelength λ j and the disturbance time t in the disturbance response residual spectrum, and R(λ j ) represents the main peak response at λ j .

[0121] Thus, by accurately extracting the main peak response, the response signal of the target pollutant is significantly enhanced, and the interference of irrelevant signals and background noise is excluded.

[0122] In step S620, the signal-to-noise ratio of the main peak response corresponding to each candidate response wavelength is calculated respectively.

[0123] The signal-to-noise ratio (SNR) is a key indicator for evaluating the quality of spectral response. By calculating the signal-to-noise ratio of each wavelength λ j , the stability of the target signal and the noise level can be measured.

[0124] Specifically, the signal-to-noise ratio (SNR) is obtained by ratio calculation of the main peak response R(λ j ) of each wavelength λ j and the standard deviation σ bg (λ j ) of the background noise.

[0125] The background noise σ bg (λ j ) is estimated by calculating the spectral standard deviation of the non-response interval of λ j .

[0126]

[0127] In the formula, σ bg (λ j ) is the background noise standard deviation at λ j , and SNR(λ j) represents the signal-to-noise ratio of the main peak response R(λ j ) of the main peak response.

[0128] The higher the signal-to-noise ratio, the stronger the stability of the response signal, and by maximizing the signal-to-noise ratio, the response quality of the pollutant is optimized, thereby reducing the possibility of misjudgment.

[0129] In step S630, for each candidate response wavelength, the fingerprint correlation between the main peak response corresponding to the candidate response wavelength and the standard fingerprint template response is calculated.

[0130] It should be noted that the spectral response of each target pollutant has a specific "fingerprint" feature, and by correlation analysis with a known standard fingerprint template, the presence and concentration of the target pollutant can be further confirmed.

[0131] Specifically, the response intensity of each wavelength λ j is matched with the standard fingerprint template R ref (λ j ) by Pearson correlation coefficient, thereby confirming the consistency of each wavelength response with the standard template, thereby improving the accuracy of the response signal.

[0132]

[0133] In the formula, R ref (λ j ) represents the response intensity of the standard fingerprint template at λ j , Cov represents the covariance, C(λ j ) represents the fingerprint correlation, σ R (λ j ) and are the standard deviation of the main peak response at λ j and the standard deviation of the standard fingerprint template, respectively.

[0134] It should be noted that the standard fingerprint template refers to the reference data of the spectral main peak response distribution with wavelength for a specific target pollutant under a set (or series) of specific experimental / standard conditions after collection and normalization. Through the standard fingerprint template, it expresses the "characteristic spectral response curve" or "fingerprint curve" of the pollutant under "ideal pure environment" for a specific response wavelength.

[0135] Thus, by correlation matching, the response signal is ensured to match the standard template in the database, and the selected wavelength is ensured to best match the standard response mode of the target pollutant, further reducing the risk of misidentification or missed detection.

[0136] In step S640, the comprehensive discrimination score is calculated.

[0137] Specifically, by comprehensively considering the signal-to-noise ratio and the correlation, the comprehensive discrimination score of each wavelength is calculated, so as to comprehensively evaluate the representativeness of each wavelength to the response of the target pollutant.

[0138] Specifically, the corresponding weight coefficients a and b are set for different indicators, and the weights should satisfy a+b=1. Further, the signal-to-noise ratio and the correlation are combined by weighted calculation to obtain the comprehensive discrimination score.

[0139] Q(λ j )=a·SNR(λ j )+b·C(λ j ), formula (10)

[0140] In the formula, Q(λ j ) represents the comprehensive discrimination score at λ j , and a and b represent the discrimination weight.

[0141] In this way, by using the weighted score method, the signal-to-noise ratio and the correlation can be considered comprehensively, and the selected optimal response wavelength can maximize the signal quality and detection accuracy.

[0142] In step S650, the wavelength point corresponding to the maximum comprehensive discrimination score is selected as the optimal response characteristic wavelength λ * .

[0143] Here, the most representative wavelength is selected from the candidate response wavelengths by the comprehensive discrimination score.

[0144] Specifically, the comprehensive discrimination scores Q(λ j ) of all candidate response wavelengths are compared, and the wavelength λ * corresponding to the maximum value is selected as the optimal response characteristic wavelength.

[0145]

[0146] By the embodiment of the present application, the optimal response characteristic wavelength determination method based on the correlation between the main peak response and the standard fingerprint template is used, which can perform multi-point matching between the disturbance response residual spectrum of the current sample and the calibrated standard fingerprint template of the target pollutant, and the weight of the signal-to-noise ratio and the correlation is comprehensively considered, which not only ensures the significance of the response signal, but also ensures that the wavelength selection has the physical specificity of the target pollutant. Therefore, the anti-interference ability and the specificity of the detection of trace organic pollutants in a complex water body are significantly improved, and the probability of false positives and component overlap interference is reduced.

[0147] As to the selection of the concentration correction model function f(·) in the embodiments of the present application, it should be noted that the traditional linear or multiple linear regression model can only fit a simple ideal scenario where the response varies linearly with the concentration, and in real water body environment, the model is easily affected by system errors, interaction effects and nonlinear background interference, resulting in large errors and poor adaptability. When the relationship between the disturbance and the environment variable and the concentration presents an exponential, saturation, threshold or other complex nonlinear variation, the linear model is unable to fit, and even gives completely wrong concentration prediction.

[0148] In contrast, the kernel mapping function (such as Gaussian RBF kernel, Sigmoid kernel, neural network, etc.) can effectively map the high-dimensional complex nonlinear relationship of the input variables (normalized main peak response, disturbance parameters, environmental parameters, etc.) to a space that is easier to linearly divide, and automatically capture the implicit coupling relationship and dynamic characteristics between variables. In addition, the model can not only depict complex responses at one time, but also continuously adapt to changes in the field through incremental learning (fine-tuning parameters), automatically correct the effects of system long-term drift, background changes and new disturbance patterns, and improve the reliability and applicability of long-term operation.

[0149] In some examples of the embodiments of the present application, the concentration correction model function adopts a nonlinear kernel mapping model based on Gaussian radial basis function, and the advantages of which compared to other kernel functions will be described in detail in other parts of the text combined with the experimental part.

[0150] Specifically, the spectral intensity values of the disturbance response residual spectrum at the optimal response characteristic wavelength and in the disturbance time interval are normalized.

[0151] Here, the disturbance response residual spectrum directly reflects the dynamic spectral change of the water sample under micro-disturbance, but the absolute spectral intensity is easily affected by multiple sources such as sample background, instrument sensitivity and system noise. By normalizing the residual spectrum at each disturbance time t, the sample background, instrument drift and batch effect are eliminated.

[0152] For each sample, the normalized spectral intensity value of the disturbance response residual spectrum at the optimal response characteristic wavelength λ * and in the disturbance time interval is denoted as f(λ * ,t).

[0153]

[0154] In the formula, △S norm (λ * ,t) represents the normalized spectral intensity value of the disturbance response residual spectrum at the optimal response characteristic wavelength λ * and at the disturbance time t; μ bg (λ * ) and σ bg (λ *respectively represent the mean and standard deviation of all spectral intensity values before applying perturbation at λ *

[0155] The perturbation operation parameters and environmental parameters are normalized respectively.

[0156] Similarly, the perturbation operation parameters (such as amplitude, pH adjustment amplitude, etc.) and environmental parameters (such as temperature, conductivity, etc.) have different physical dimensions and statistical distributions, and normalization can avoid feature imbalance and numerical instability in subsequent modeling.

[0157] Specifically, the perturbation parameters A perturb and the environmental parameters E env are mean-standard deviation normalized respectively:

[0158]

[0159] In the formula, μ A and σ A respectively represent the mean and standard deviation of the perturbation parameters, A norm represents the normalized perturbation operation parameter vector; E norm represents the normalized environmental parameter vector, μ E and σ E respectively represent the mean and standard deviation of the environmental parameters.

[0160] Through the above normalization operation, it is ensured that all input features are in the same dimension and numerical scale, providing balanced feature space input for Gaussian kernel mapping.

[0161] The normalized features are spliced to construct the input vector x:

[0162] x = [△S norm (λ * , t), A norm , E norm ], formula (14)

[0163] The input vector is processed by a nonlinear kernel mapping model based on Gaussian radial basis kernel function to predict the concentration of the target trace organic pollutants.

[0164]

[0165] In the formula, K represents the number of Gaussian radial basis kernel function hidden nodes, w k represents the weight parameter of the kth hidden node obtained by pre-training, b represents the bias term obtained by pre-training, c k represents the center vector of the kth hidden node, σ k represents the width parameter of the kth Gaussian radial basis kernel, and ‖x-c k ​‖ represents the Euclidean distance between the input feature and the center of the k-th hidden node.

[0166] Here, by automatically learning the high-order nonlinear coupling mapping between input features through the Gaussian Radial Basis Function (RBF), the complex influence mechanism of perturbation signals, perturbation parameters, and environmental parameters on the response of trace organic pollutants can be adaptively characterized, thereby improving the quantitative accuracy and system robustness under low concentration and dynamic environments.

[0167] The Gaussian kernel forms a "local peak" for each sample point in the input variable space, meaning the model's influence on each sample decreases exponentially with increasing distance from the input variables. It is suitable for modeling the influence of perturbation responses and changes in environmental parameters on the main peak response (i.e., there are local differences in the perturbation responses among water samples rather than a globally linear pattern). Furthermore, the Gaussian kernel can map the original finite-dimensional space to an infinite-dimensional high-dimensional feature space, effectively capturing the nonlinear relationship between target concentration and multiple factors such as perturbation parameters and environmental parameters, involving higher-order coupling.

[0168] Through the embodiments of this application, the system integrates the normalization processing of the residual spectrum of the disturbance response, the normalization of multi-source parameters and feature stitching, and uses a nonlinear mapping model with Gaussian radial basis kernel to predict the concentration of the normalized feature vector. This realizes the automatic modeling and engineering application of the complex nonlinear coupling relationship between the disturbance response signal, the disturbance operation parameters and the environmental background parameters, which significantly improves the model's adaptability to dynamic disturbances and the quantitative accuracy of complex environmental changes.

[0169] To verify the effectiveness of the nonlinear concentration correction model based on Gaussian RBF kernel proposed in this application for the detection of trace organic pollutants in water, several water samples of different types were selected. For each sample, spectral intensity time-series data (ΔS(λ)) at its optimal response characteristic wavelength were collected. * ,t)), disturbance operating parameters (A perturb The data includes environmental parameters (such as water temperature, pH, conductivity, etc., which are uniformly normalized) and the corresponding known real pollutant concentrations. The data sample covers the range of low concentration (0.01 mg / L) to high concentration (1.0 mg / L), and has high representativeness and engineering authenticity.

[0170] This experiment used Gaussian RBF kernel support vector regression (SVR), multinomial kernel SVR, Sigmoid kernel SVR, and linear kernel SVR to model the data. All models used ΔS(λ) as the basis for the model. * ,t)A perturb The system takes environmental parameters as joint input and pollutant concentration as output. Each model automatically fine-tunes its parameters through grid search and cross-validation to ensure optimal model performance.

[0171] To comprehensively evaluate the performance of the model, the following three indicators are mainly used:

[0172] Goodness of fit R 2 : measure the explanatory power of the predicted results to the actual concentration;

[0173] Root mean square error (RMSE): measure the average error between the predicted value and the true value;

[0174] Abnormal point detection rate: the proportion of samples with a prediction error absolute value greater than 0.1 mg / L.

[0175] Table 1. Comparison of experimental results

[0176]

[0177]

[0178] As shown in the experimental results in Table 1, the Gaussian RBF kernel model achieves the highest goodness of fit and the lowest root mean square error on the test set, and the abnormal point detection rate is also much lower than other kernel functions. Although the polynomial kernel has certain high-order fitting ability, it shows slight overfitting when the variable dimension increases, and the generalization ability is not as good as the Gaussian RBF kernel. The Sigmoid kernel and the linear kernel cannot effectively capture the nonlinear relationship under complex disturbance and environmental variables, and the prediction error and abnormal detection rate are high.

[0179] Figure 7 The prediction error distribution comparison simulation diagram of an example of the concentration prediction experiment of trace organic pollutants under different kernel function models is shown.

[0180] As shown in the results in Figure 7 , the residual distribution of the Gaussian RBF kernel model is the most concentrated in the entire concentration interval, the fluctuation amplitude is the smallest, the error is always close to zero, and it is significantly better than the polynomial kernel, the Sigmoid kernel and the linear kernel model. Other kernel models have larger error fluctuations and abnormal points in different concentration intervals, showing lower fitting accuracy and engineering stability. Therefore, it is verified that the Gaussian RBF kernel nonlinear correction model has high accuracy and strong robustness in predicting the concentration of trace organic pollutants in complex water environment.

[0181] The details of the exemplary embodiment of the training of the concentration correction model will be further described below.

[0182] Specifically, a plurality of standard samples of target trace organic pollutants with known concentrations are prepared, and the disturbance response residual spectrum data of each standard sample at the optimal response characteristic wavelength under disturbance operation, as well as the corresponding disturbance operation parameters and environmental parameters, are collected to form a training sample set.

[0183] Exemplarily, standard water samples with known concentrations are prepared in laboratory conditions, such as 10-20 groups, covering the full range of the detection system (e.g. 0.01-1.0 mg / L).

[0184] For each group of water samples, data acquisition is performed under the conditions of unified disturbance operation parameters (such as oscillation amplitude, pH adjustment range, light period, etc.) and environmental parameters (such as temperature, conductivity).

[0185] The collected data includes the disturbance response residual spectrum of each standard sample at the optimal response characteristic wavelength λ * under the disturbance operation. Disturbance parameters Environmental parameters and the true concentration of the group.

[0186] Based on the optimal response characteristic wavelength and the disturbance timing interval in the training samples, the disturbance response residual spectrum data is normalized, and the principal component analysis method is used to reduce the dimension of the normalized timing spectrum intensity vector to extract the principal component features.

[0187] Here, the timing residual spectrum intensity between different samples is standardized, and the principal components are extracted, which not only eliminates non-target factors such as instruments and sample backgrounds, but also effectively compresses high-dimensional features, improves model generalization and calculation efficiency.

[0188] Specifically, the disturbance response residual spectrum of each group of training samples is normalized:

[0189]

[0190] wherein, represents the normalized disturbance response residual spectrum sequence of the i-th training sample at its optimal response characteristic wavelength λ * and the disturbance timing interval t, and are the mean and standard deviation of all training samples at λ * under the background period (or undisturbed time).

[0191] The normalized spectrum intensity sequence of all training samples is constructed into a matrix, and principal component analysis (PCA) is used:

[0192]

[0193] wherein, Z (train) represents the principal component feature matrix of all training samples, and N represents the total number of training samples.

[0194] The first d principal components (such as 2 or 3) are retained, and the principal component features corresponding to each group of samples are

[0195] Thus, the non-target variables such as noise, background, instrument drift, etc. are effectively removed, and the most core dynamic characteristics representing the change of the perturbation response are extracted.

[0196] The perturbation operation parameters and the environmental parameters in the training samples are normalized.

[0197] Specifically, the mean-standard deviation normalization is performed on the perturbation operation parameters and each dimension of the environmental parameters, respectively.

[0198]

[0199] wherein, represents the normalized value of the perturbation operation parameters of the i-th training sample, and respectively represent the mean and the standard deviation of the perturbation operation parameters of all training samples, represents the normalized value of the environmental parameters (such as temperature, conductivity, pH, etc.) of the i-th training sample, and respectively represent the mean or the standard deviation of all training samples in the environmental parameter component. Thus, the weight balance in the subsequent kernel function mapping is ensured, and the model convergence speed and the generalization ability are improved.

[0200] For each training sample, the principal component features corresponding to the training sample, the normalized perturbation operation parameters, and the normalized environmental parameters are spliced into a training input feature vector.

[0201] Based on the K-means clustering algorithm, all training input feature vectors are clustered to determine the hidden node centers and the kernel width parameters of the Gaussian radial basis kernel function.

[0202] Here, the features from different sources are fused into a unified vector, the K-means clustering is used to initialize the Gaussian kernel center and the kernel width, and the adaptability of the model to the data distribution is enhanced. Specifically, for each group of training samples, the principal component features, the normalized perturbation parameters, and the normalized environmental parameters are spliced to obtain a feature vector:

[0203]

[0204] All are clustered by the K-means clustering algorithm, and K cluster centers c k are taken as the RBF hidden node centers, and the kernel width σ k is set as the average Euclidean distance between all cluster centers. Thus, the kernel function center and the bandwidth are adaptively selected according to the spatial distribution of the training samples, and the expression ability and the stability of the model under complex data structure are improved.

[0205] Based on the training input feature vector and the corresponding known concentration, the weight and bias parameters of the Gaussian radial basis kernel function are jointly updated by the least square method combined with the Adam optimization algorithm to minimize the mean square error loss of the training samples.

[0206] More specifically, the normalized features, hidden node centers and bandwidths are adopted to realize the nonlinear mapping from the input to the concentration output through the joint optimization of the weight and bias.

[0207] Based on all the training samples, the nonlinear kernel mapping model of the Gaussian radial basis kernel function constructed as shown in equation (15) is trained and optimized by constructing the following loss function:

[0208]

[0209] The weight {w k} and bias b are jointly trained by the least square method combined with the Adam optimization algorithm until the loss function converges, so that the model can automatically adapt to the nonlinear coupling relationship between the high-dimensional principal component features and the disturbance, environmental variables, and realize high-precision concentration inversion.

[0210] Through the embodiments of the present application, the centers and bandwidths of the Gaussian kernel function are adaptively determined by K-means clustering, so that the subsequent optimization of the model parameters is based on the true depiction of the essential structure of the input feature space, realizing adaptive initialization of the Gaussian radial basis kernel structure, and enabling the concentration correction model to automatically capture the high-dimensional nonlinear coupling relationship between the complex water disturbance signal, environmental variables and pollutant concentration, thereby significantly improving the modeling accuracy and generalization ability of the model.

[0211] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of actions, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application. In the above embodiments, the description of each embodiment is focused on, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0212] Figure 8 A structural block diagram of an example of a water body trace organic pollutant detection system according to an embodiment of the present application is shown.

[0213] As Figure 8 shown, the water body trace organic pollutant detection system 800 includes a water body sample collection unit 810, a sample micro-disturbance unit 820, an optimal wavelength identification unit 830 and a trace concentration identification unit 840.

[0214] The water body sample collection unit 810 is configured to collect an environmental parameter of a target water body sample to be detected and obtain baseline spectral data of the target water body sample.

[0215] The sample micro-disturbance unit 820 is configured to determine a disturbance operation parameter matched with the target water body sample and apply a micro-disturbance to the target water body sample according to the disturbance operation parameter to obtain disturbance response spectral data; the disturbance operation parameter includes a disturbance operation type, a disturbance operation amplitude, and a disturbance operation duration.

[0216] The optimal wavelength identification unit 830 is configured to perform time series difference analysis on the disturbance response spectral data and the baseline spectral data to obtain a disturbance response residual spectrum, and identify an optimal response characteristic wavelength from a characteristic response wavelength interval of a target trace organic pollutant based on the disturbance response residual spectrum.

[0217] The trace concentration identification unit 840 is configured to solve a concentration correction model according to the optimal response characteristic wavelength, the disturbance operation parameter, and the environmental parameter to obtain a concentration of the target trace organic pollutant.

[0218] The concentration correction model is expressed by the following formula:

[0219] C target = f (△S (λ * ,t), A perturb ,E env ),

[0220] In the formula, C target is the concentration of the target trace organic pollutant, T is a disturbance time interval, △S (λ * ,t) is a response intensity of the disturbance response residual spectrum at the optimal response characteristic wavelength λ * and the disturbance time t, A perturb is the disturbance operation parameter, E env is the environmental parameter, and f (·) is a pre-trained concentration correction model function.

[0221] In some embodiments, the present application provides a non-volatile computer readable storage medium, wherein the storage medium stores one or more programs including execution instructions, the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to perform the steps of any of the water body trace organic pollutant detection methods described above.

[0222] In some embodiments, the embodiments of the present application further provide a computer program product, comprising a computer program stored on a non-volatile computer readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the steps of any of the water body trace organic pollutant detection methods described above.

[0223] In some embodiments, the embodiments of the present application further provide an electronic device, comprising: at least one processor, and a memory connected to the at least one processor in communication, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the water body trace organic pollutant detection method.

[0224] The above product can execute the method provided by the embodiments of the present application, and has the corresponding function modules and beneficial effects of executing the method. Technical details not described in detail in the embodiments can be referred to the method provided by the embodiments of the present application.

[0225] The electronic device of the embodiments of the present application exists in various forms, including but not limited to: mobile communication devices, ultra-mobile personal computer devices, portable entertainment devices, or other onboard electronic devices with data interaction functions.

[0226] The device embodiments described above are only schematic, and the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e. they may be located in one place, or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0227] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus a general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions or the part that contributes to the related art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in the various embodiments or some parts of the embodiments.

[0228] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing examples, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting trace amounts of organic pollutants in a body of water, characterized by, The method comprises: Collecting environmental parameters of a target water sample to be detected, and obtaining baseline spectral data of the target water sample; Determine the perturbation operation parameters matched with the target water sample, and apply a micro-perturbation to the target water sample according to the perturbation operation parameters to obtain perturbation response spectral data; the perturbation operation parameters include perturbation operation type, perturbation operation amplitude and perturbation operation duration; Perform time series difference analysis on the perturbation response spectral data and the baseline spectral data to obtain perturbation response residual spectrum, and identify the optimal response characteristic wavelength from the characteristic response wavelength interval of the target trace organic pollutant based on the perturbation response residual spectrum; Solve the concentration correction model according to the optimal response characteristic wavelength, the perturbation operation parameters and the environmental parameters to obtain the concentration of the target trace organic pollutant; The concentration correction model is expressed by the following formula: C target = f(AS(λ * ,t), A perturb , E env ), wherein C target is the concentration of the target trace organic contaminant, T is the perturbation time interval, S(λ * , t) is the response intensity of the perturbation response residual spectrum at the optimal response characteristic wavelength λ * and the perturbation time t, A perturb is the perturbation operating parameter, E env is the environmental parameter, and f(·) is the pre-trained concentration correction model function.

2. The method of claim 1, wherein, The perturbation operation type includes at least one of the following: physical micro-vibration, pH value fine adjustment, ion strength micro-change or light intensity periodic fluctuation.

3. The method according to claim 1 or 2, characterized in that, The determination of the perturbation operation parameters matched with the target water sample comprises: Based on a plurality of calibration perturbation operation parameters in a calibration parameter group, a leading perturbation operation is applied to the target water sample respectively; each calibration perturbation operation parameter respectively defines a corresponding perturbation operation type, a perturbation operation amplitude and a perturbation operation duration; For each calibration perturbation operation parameter, obtain the perturbation response residual spectrum corresponding to the calibration perturbation operation parameter, and determine the maximum target response of the perturbation response residual spectrum in the characteristic wavelength interval of the target trace organic pollutant and the perturbation time interval, and determine the maximum background response in the background wavelength interval irrelevant to the target trace organic pollutant and the perturbation time interval; Calculate the ratio of the maximum target response to the maximum background response of each calibration perturbation operation parameter to obtain a target correlation score; Screen out the perturbation operation parameter corresponding to the maximum target correlation score as the perturbation operation parameter matched with the target water sample.

4. The method of claim 3, wherein, For the determination of the calibration parameter group, comprising: Obtain the target pollutant type of the target trace organic pollutant; Determine the calibration parameter group matched with the target pollutant type from a preset pollutant-calibration parameter set; the pollutant-calibration parameter set pre-stores a plurality of pollutant types and corresponding calibration parameter groups.

5. The method of claim 1, wherein, The identification of the optimal response characteristic wavelength from the characteristic response wavelength interval of the target trace organic pollutant based on the perturbation response residual spectrum comprises: in a characteristic response wavelength interval Λ of the target trace organic contaminant tgt For each candidate response wavelength, a main peak response is calculated based on the perturbation response residual spectrum: R(λ j ) = max t∈T △S(λ j ,t), where ΔS(λ j ,t) is the response intensity at the candidate response wavelength λ j and the perturbation time t in the perturbation response residual spectrum, R(λ j ) represents the main peak response at λ j . For each candidate response wavelength, calculate the signal-to-noise ratio of the main peak response corresponding to the candidate response wavelength respectively: where σ bg (λ j ) is the standard deviation of the background noise at λ j , and SNR(λ j ) represents the signal-to-noise ratio of the main peak response R(λ j ). For each candidate response wavelength, calculate the fingerprint correlation degree between the main peak response corresponding to the candidate response wavelength and the standard fingerprint template response: where R ref (λ j ) represents the response intensity of the standard fingerprint template at λ j , Cov represents the covariance, C(λ j ) represents the fingerprint correlation, σ R (λ j ) and are the standard deviation of the main peak response at λ j and the standard deviation of the standard fingerprint template, respectively; Calculate the comprehensive discrimination score: Q(λ j ) = a · SNR(λ j ) + β · C(λ j ), where Q(λ j ) represents the overall discriminant score at λ j , and α and β represent discriminant weighting weights, and α + β = 1. Select the wavelength point corresponding to the maximum comprehensive discrimination score as the optimal response characteristic wavelength λ * :

6. The method of claim 5, wherein, The concentration correction model function adopts a nonlinear kernel mapping model based on Gaussian radial basis function; The solving of the concentration correction model according to the optimal response characteristic wavelength, the perturbation operation parameters and the environmental parameters to obtain the concentration of the target trace organic pollutant comprises: normalizing all spectral intensity values of the perturbation response residual spectrum within the optimal response characteristic wavelength and the perturbation time interval: where ΔS norm (λ * ,t) represents the normalized spectral intensity of the perturbation response residual spectrum at the optimal response characteristic wavelength λ * and the perturbation time t; μ bg (λ * ) and σ bg (λ * ) represent the mean and standard deviation of all spectral intensities before the perturbation is applied at λ * . normalizing the perturbation operation parameters and the environmental parameters respectively: where μ A and σ A denote the mean and standard deviation of the disturbance parameters, respectively, A norm denotes the normalized disturbance operating parameter vector; E norm denotes the normalized environment parameter vector, μ E and σ E denote the mean and standard deviation of the environment parameters, respectively. splicing each normalized feature to construct an input vector x: x = [AS norm (λ * ,t), A norm , E norm ], processing the input vector by using a nonlinear kernel mapping model based on a Gaussian radial basis kernel function to predict the concentration of the target trace organic pollutant; where K represents the number of Gaussian radial basis kernel function hidden nodes, w k represents the weight parameter of the kth hidden node obtained by pre-training, b represents the bias term obtained by pre-training, c k represents the center vector of the kth hidden node, σ k represents the width parameter of the kth Gaussian radial basis kernel, ‖x-c k ‖ represents the Euclidean distance between the input feature and the center of the kth hidden node.

7. The method of claim 6, wherein, for training of the concentration correction model, including: preparing a plurality of standard samples of the target trace organic pollutant with known concentrations, collecting the perturbation response residual spectrum data of each standard sample at the optimal response characteristic wavelength under perturbation operation, and the corresponding perturbation operation parameters and environmental parameters, to form a training sample set; based on the optimal response characteristic wavelength and the perturbation time interval in the training sample, normalizing the perturbation response residual spectrum data, and using principal component analysis method to reduce the dimension of the normalized time series spectral intensity vector to extract principal component features; normalizing the perturbation operation parameters and the environmental parameters in the training sample; for each training sample, splicing the principal component features, normalized perturbation operation parameters and normalized environmental parameters corresponding to the training sample into a training input feature vector; based on the K-means clustering algorithm, clustering all training input feature vectors to determine the hidden node center and kernel width parameters of the Gaussian radial basis kernel function; based on the training input feature vector and the corresponding known concentration, updating the weight and bias parameters of the Gaussian radial basis kernel function by least squares method combined with Adam optimization algorithm to minimize the mean square error loss of the training sample.

8. A system for detecting trace organic contaminants in a body of water, comprising: The system comprises: a water sample collection unit for collecting the environmental parameters of the target water sample to be detected and obtaining baseline spectrum data of the target water sample; a sample perturbation unit for determining perturbation operation parameters matched with the target water sample and applying micro-perturbation to the target water sample according to the perturbation operation parameters to obtain perturbation response spectrum data; the perturbation operation parameters include perturbation operation type, perturbation operation amplitude and perturbation operation duration; an optimal wavelength identification unit for performing time series difference analysis on the perturbation response spectrum data and the baseline spectrum data to obtain perturbation response residual spectrum, and identifying the optimal response characteristic wavelength from the characteristic response wavelength interval of the target trace organic pollutant based on the perturbation response residual spectrum; a trace concentration identification unit for solving a concentration correction model according to the optimal response characteristic wavelength, the perturbation operation parameters and the environmental parameters to obtain the concentration of the target trace organic pollutant; the concentration correction model is expressed by the following formula: C target = f(△S(λ * ,t), A perturb , E env ), In the formula, C target is the concentration of the target trace organic pollutant, T is the perturbation time interval, and S(λ * , t) is the response intensity of the perturbation response residual spectrum at the optimal response characteristic wavelength λ * and the perturbation time t, A perturb is the perturbation operation parameter, E env is the environmental parameter, and f(·) is the pre-trained concentration correction model function.

Images