An artificial intelligence-based methylene blue online detection method and system

Abstract

The present application relates to the technical field of online chemical component spectrum analysis, and discloses a methylene blue online detection method and system based on artificial intelligence, which comprises the following steps: injecting a chemical disturbance agent actively and collecting a time sequence of spectra, eliminating background interference by differential processing, analyzing the dynamic change mode of the spectral characteristics of a target object based on an artificial intelligence model to determine the concentration of the target object, and comparing the actual dynamic mode with a theoretical model to realize online self-diagnosis and confidence evaluation of the measurement process. The present application converts a static spectrum analysis problem which is susceptible to background matrix drift into a dynamic response rate problem of an analysis target object to standard stimulation, thereby solving the selectivity and reliability problems of long-term measurement in a dynamic complex environment and improving the environmental immunity and self-calibration capability of the system.

Description

Technical Field

[0001] This invention relates to an online detection method and system for methylene blue based on artificial intelligence, belonging to the field of online chemical composition spectral analysis technology. Background Technology

[0002] Currently, in applications such as industrial wastewater treatment, the use of ultraviolet-visible spectroscopy combined with chemometric models for online monitoring of specific pollutants is a commonly adopted technique. Its advantage lies in its ability to respond quickly to the process. However, the effectiveness of this technique depends on a core premise: that the background chemical matrix of the water sample to be tested remains stable. In actual industrial environments, normal fluctuations in upstream production conditions will cause continuous and irregular changes in the types and concentrations of background interfering substances in the wastewater.

[0003] The continuous changes in this background matrix will cause unpredictable deformations in the background absorption patterns of the spectral signals acquired by the online spectrometer. This will cause any static calibration model built based on historical data to deviate from the statistical distribution of its input data during training, and the measurement accuracy will gradually decrease. Since this deviation occurs gradually, the system usually does not issue an immediate warning. The result is long-term material waste or excessive pollutant emissions, constituting an uncertainty in process control. To address this problem, simply increasing the algorithm complexity of the calibration model or increasing the frequency of manual offline calibration cannot fundamentally solve the issue. Specifically, existing technologies mainly suffer from the following limitations: 1. The predictive ability of any static model is limited by the range covered by its training dataset. The model itself lacks an effective adaptation mechanism for the unknown background interference that constantly occurs during the production process; 2. Excessive downtime and manual offline sampling and recalibration increase the workload of operation and maintenance, and also weaken the real-time and continuous nature that online monitoring should possess, causing it to deviate from the original intention of continuous automated monitoring in application.

[0004] To overcome this bottleneck, some researchers in this field have attempted to improve the detection methods themselves to enhance their anti-interference capabilities. However, their underlying logic still relies on static measurements, thus remaining limited in the face of complex and ever-changing interference. For example, Chinese invention patent CN118655105B discloses a rapid detection method and device for the equivalent mud content of concrete aggregates. This method measures the spectral absorbance of the filtrate after the sample is mixed with methylene blue solution at a specific wavelength and calculates the so-called equivalent mud content based on a pre-calibrated simplified linear model. While this method simplifies the operation to some extent, it is essentially still a single-point static spectral analysis mode. The core defect of this mode is... The problem lies in its heavy reliance on a fixed calibration model, making it unable to identify and address spectral or chemical kinetic interferences from unknown coexisting substances in the water sample background online. If the actual water sample contains non-target substances that react with methylene blue or absorb at that wavelength, the static model will output biased and undetectable erroneous results. This is precisely the dilemma faced by existing technologies in complex industrial environments. This situation indicates an inherent technical contradiction in existing methods: attempting to use a fixed data model to handle a dynamically changing, open physical system. The core constraint is the lack of a mechanism in existing analytical procedures to respond to environmental changes in real time and perform self-correction. Therefore, the technical problem this invention aims to solve is how to establish a new online detection method that is independent of the stability of the background environment and can immediately calibrate against dynamic background interference during measurement, thereby separating the measurement results from changes in the background matrix. Summary of the Invention

[0005] This invention provides an artificial intelligence-based online detection method and system for methylene blue. Its main purpose is to solve the problem that the measurement accuracy gradually decreases when using a static calibration model for online monitoring in the prior art because it cannot adapt to the dynamic chemical background of the water sample being tested.

[0006] To achieve the above objectives, this invention provides an online methylene blue detection method based on artificial intelligence, the method comprising:

[0007] Step a: Collect the baseline absorption spectrum of the water sample to be tested flowing through a detection point;

[0008] Step b: After completing step a, inject a dose of chemical perturbation agent into the water sample to be tested to actively stimulate dynamic changes in the spectral absorption characteristics of methylene blue in the water sample, and continuously acquire multiple absorption spectra within the time window after injection to form a spectral time sequence.

[0009] Step c: Perform a differential operation on each spectrum in the spectral time series with the baseline absorption spectrum acquired in step a to generate a differential spectral time series that only represents dynamic changes, and extract the actual dynamic change pattern that represents the intensity decay process of the methylene blue characteristic absorption peak from the differential spectral time series.

[0010] Step d: Input the actual dynamic change pattern into the artificial intelligence model to determine the concentration of methylene blue in the water sample to be tested;

[0011] Step e: Input the concentration of methylene blue determined in step d into a normalized dynamic response model to generate an ideal dynamic change pattern corresponding to the concentration. The normalized dynamic response model is established in advance by processing a to d of a standard methylene blue solution without kinetic interference.

[0012] Step f involves comparing the actual dynamic change pattern obtained in step c with the ideal dynamic change pattern generated in step e, and determining the confidence level of the concentration measurement results based on the difference between the two.

[0013] Preferably, the process of determining the confidence level of the concentration measurement result based on the difference between the two in step f specifically includes: treating the actual dynamic change pattern and the ideal dynamic change pattern as two high-dimensional vectors and calculating the morphological deviation between them to obtain a kinetic distortion index; comparing the kinetic distortion index with a pre-set confidence threshold; marking the concentration measurement result as high confidence when the kinetic distortion index is less than the confidence threshold; and marking the concentration measurement result as low confidence when the kinetic distortion index is not less than the confidence threshold, and outputting a warning message indicating the presence of kinetic interference.

[0014] Preferably, the actual dynamic change mode in step c includes at least one of the following: the initial decay rate of the characteristic absorption peak intensity, the nonlinearity parameter of the characteristic absorption peak intensity decay curve, and the time required for the characteristic absorption peak intensity to reach a steady state. In step b, the chemical perturbator is an oxidant that can chemically react with methylene blue to decompose it.

[0015] Preferably, when the concentration measurement results are continuously marked as low confidence, the method also automatically initiates active interference offsetting, which includes: performing a two-stage perturbation measurement on the water sample to be tested, including injecting a first dose and a second dose of a chemical perturbant, wherein the first dose and the second dose are different; obtaining a first dynamic change pattern caused by the first dose perturbation and a second dynamic change pattern caused by the second dose perturbation; and determining the methylene blue concentration that has been corrected for kinetic interference based on a joint analysis of the nonlinear difference relationship between the first dynamic change pattern and the second dynamic change pattern.

[0016] Preferably, the dynamic distortion index in step f1 The calculation follows these rules: ,in, The decay curve function corresponding to the actual dynamic change mode obtained from step c. The decay curve function corresponding to the ideal dynamic change mode generated in step e. For time variables, This represents the total duration of the time window in step b.

[0017] Preferably, the method further includes a self-diagnosis of the chemical probe integrity, which includes: ensuring that when acquiring the baseline absorption spectrum in step a, the baseline absorption spectrum also contains a spectral fingerprint from the chemical perturbator itself within a preset wavelength range; extracting the spectral fingerprint of the chemical perturbator itself from the acquired baseline absorption spectrum; comparing the extracted spectral fingerprint with a pre-stored reference spectral fingerprint representing the chemical perturbator in a brand-new state; and determining the chemical purity state of the chemical perturbator or the accuracy state of its injection dose based on the comparison result of step a.

[0018] Preferably, in step a, determining the chemical purity of the chemical perturbator specifically involves analyzing whether new absorption peaks appear or the proportion of existing characteristic peaks changes relative to the reference spectral fingerprint; determining the accuracy of its injection dose specifically involves analyzing whether the integrated area or peak height of the spectral fingerprint undergoes a systematic shift relative to the integrated area or peak height of the reference spectral fingerprint.

[0019] Preferably, the method further includes: continuously recording the trend data of the chemical purity state or the accuracy state of its injected dosage determined by step a over time; and based on the trend data, inferring the remaining effective life of the chemical disturbance agent or its injection system through a pre-established prediction model, and generating predictive maintenance instructions before its failure.

[0020] Preferably, the artificial intelligence model in step d is a gradient boosting machine model based on decision trees or a recurrent neural network model.

[0021] An artificial intelligence-based online methylene blue detection system, comprising:

[0022] A flow pool equipped with detection points;

[0023] A spectral acquisition device, the optical path of which passes through the detection point;

[0024] A micro-injection device, the outlet of which is connected to a flow-through cell;

[0025] A control and data processing unit is connected to a spectral acquisition device and a micro-injection device. The control and data processing unit is configured to: control the spectral acquisition device to acquire the baseline absorption spectrum of the water sample flowing through the detection point; after acquiring the baseline absorption spectrum, control the micro-injection device to inject a dose of chemical perturbation agent into the water sample, and simultaneously control the spectral acquisition device to continuously acquire multiple absorption spectra within a time window after injection to form a spectral time series; and perform a difference operation between each spectrum in the spectral time series and the baseline absorption spectrum to generate a single-tablet representation. The method involves generating a differential spectral time series characterized by dynamic changes, and extracting the actual dynamic change pattern representing the attenuation process of the characteristic absorption peak intensity of methylene blue from this differential spectral time series. The actual dynamic change pattern is then input into an artificial intelligence model stored in the unit to determine the concentration of methylene blue in the water sample to be tested. The determined concentration of methylene blue is then input into a normalized dynamic response model stored in the unit to generate an ideal dynamic change pattern corresponding to that concentration. The obtained actual dynamic change pattern and the generated ideal dynamic change pattern are compared in morphology, and the confidence level of the concentration measurement result is determined based on the difference between the two.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] 1. By actively injecting a chemical perturbation agent into the water sample after acquiring the baseline absorption spectrum, and continuously acquiring and analyzing the spectral time series induced by the perturbation; this combination of technical features transforms the difficult problem of resolving the concentration of the target substance from a complex spectrum mixed with dynamic background interference into a kinetic problem that only needs to analyze the response rate of the target substance's own spectral characteristics. This makes the concentration measurement result directly related to a repeatable physicochemical process actively induced by the perturbation, rather than relying on the absolute intensity of the spectral signal that is easily affected by background matrix drift, thereby avoiding the problem of frequent failure of the static calibration model due to environmental changes.

[0028] 2. While determining the methylene blue concentration based on the actual observed dynamic change pattern, this concentration result is also input into a pre-set standardized dynamic response model to generate an ideal dynamic response pattern corresponding to the concentration. The actual observed dynamic pattern is then immediately compared with this ideal pattern. This process uses the measurement output results to verify the measurement process itself in real time, generating a judgment criterion for whether there is kinetic interference in the measurement process for each concentration measurement. This allows the system to not only output the concentration value but also output a self-assessment of the reliability of the value, solving the problem that traditional measurement methods cannot perceive the effectiveness of their own measurement process online.

[0029] 3. Upon identification of kinetic disturbances, a further two-stage perturbation measurement involving two different doses of chemical perturberants is performed. A joint analysis is then conducted based on the differences between the first and second dynamic change modes induced by the two perturbations. Since the interaction between perturberants of different intensities and kinetic disturbances exhibits a nonlinear relationship, while the reaction with the target substance is close to a linear relationship within a certain range, this method of analyzing the differences makes it possible to separate the unknown interference effect from the concentration calculation. This allows the system to transform from passively alerting to potential measurement deviations to actively correcting these deviations, thereby restoring measurement accuracy under complex operating conditions.

[0030] 4. When acquiring the baseline absorption spectrum of the water sample to be tested, the spectral fingerprint of the chemical perturbator itself is also superimposed on the spectral information. By comparing the real-time acquired spectral fingerprint with the pre-stored reference spectral fingerprint, this design utilizes the baseline acquisition step in the measurement process to continuously monitor the chemical properties of the chemical perturbator itself or the stability of its injection dosage, which serves as the measurement benchmark, in a way that reuses information. This allows the entire measurement system to perform self-diagnosis of the health status of its internal core tools while evaluating external samples, providing support for the long-term reliability and traceability of measurement results. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall analysis and self-diagnosis process of the detection method of the present invention;

[0032] Figure 2 This is a schematic diagram illustrating the changes in the KDI index when the system of the present invention identifies dynamic disturbances;

[0033] Figure 3 This is a schematic diagram of the hardware composition and core control architecture of the detection system of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] This invention discloses an artificial intelligence-based online detection method and system for methylene blue. Its overall architecture includes a flow cell with an optical detection point, a spectral acquisition device with the optical path passing through the detection point, a micro-injection device with its outlet connected to the flow cell, and a control and data processing unit that connects to the spectral acquisition device and the micro-injection device. The core analysis process of this method mainly consists of a baseline spectral acquisition and reference establishment stage, an active micro-perturbation injection and time-series spectral capture stage, a differential processing and dynamic feature extraction stage, an artificial intelligence model concentration inversion stage, and a parallel online self-diagnosis and confidence assessment stage. Each stage is sequentially connected, and data flows unidirectionally, forming a closed-loop online analysis procedure with built-in calibration capabilities. In an industrial wastewater online monitoring application, the water sample to be tested flows continuously through a quartz flow cell with an optical path of 10 mm. A detection point is set on this flow cell, initially defined as containing methylene blue with a concentration ranging from 0.1 mg / L to 50 mg / L, and other components. The chemical matrix consists of associated organic matter, both in type and concentration, which is dynamically changing due to fluctuations in upstream production conditions. The environmental specifications required for this process are defined as follows: a spectral acquisition device, specifically a deuterium-tungsten hybrid light source UV-Vis fiber optic spectrometer with a spectral resolution of at least 1 nm and a signal acquisition frequency set to at least 10 Hz; a micro-injection device, specifically a precision micro-pump controllable by an external digital signal with a minimum injection volume accuracy of at least 1 μL; and a control and data processing unit, specifically an industrial computer with an embedded microprocessor, which pre-stores all algorithm models and control logic required for subsequent analysis processes. To establish a stable reference for subsequent dynamic change analysis, the control and data processing unit first executes step a, namely, acquiring the baseline absorption spectrum of the water sample flowing through the detection point. In this step, the control and data processing unit sends a command to the spectral acquisition device to acquire the full-band absorption spectrum of the water sample flowing through the flow cell at the current moment and records this spectral data as the baseline absorption spectrum. ,Should It contains the superposition absorption information of methylene blue and all background interferences at the current moment, which serves as the basis for subsequent differential operations to suppress background interference.

[0036] Considering that the continuous changes in the background chemical matrix of water samples are the main factor leading to the reduced effectiveness of traditional static calibration models in continuous industrial monitoring, this method adopts an active excitation of the target analyte response, transforming the static spectral analysis problem into a dynamic reaction rate analysis problem. Accordingly, after completing step a, the system executes step b, which involves injecting a dose of chemical perturbator into the water sample to actively excite dynamic changes in the spectral absorption characteristics of methylene blue in the sample. Within the time window after injection, multiple absorption spectra are continuously acquired to form a spectral time series. Specifically, the control and data processing unit records... Immediately afterwards, a series of pulse signals are sent to the micro-injection device, driving it to instantaneously inject a preset dose of chemical disruptor into the flow cell. In this embodiment, the chemical disruptor is determined to be an oxidant capable of undergoing a rapid redox reaction with methylene blue, such as a 0.5 mol / L ozone aqueous solution. The injection dose is set according to a calibration procedure. The goal of this procedure is to ensure that this dose is sufficient to produce an observable reaction with methylene blue at a concentration upper limit of 50 mg / L within 5 seconds, but its absolute amount is far insufficient to affect the background organic matter, which can reach several thousand mg / L. For example, the injection dose can be set to 20 μL. At the instant the injection is completed, the control and data processing unit synchronously triggers the spectral acquisition device to enter high-frequency continuous shooting mode, continuously acquiring spectra at 100 ms intervals for 5 seconds, obtaining a total of 50 absorption spectra, denoted as... This set of spectra constitutes a spectral time series. Through this step, a kinetic process involving the fading of methylene blue due to oxidation is recorded in a high-time-resolution data format.

[0037] To separate the dynamic variation information of methylene blue from the complex mixed spectrum, the system is configured to execute step c, which involves performing a differential operation on each spectrum in the spectral time series with the baseline absorption spectrum acquired in step a, to generate a differential spectral time series that only represents the dynamic variations. From this differential spectral time series, the actual dynamic variation pattern representing the intensity decay process of the characteristic absorption peak of methylene blue is extracted. The control and data processing unit then processes the spectral time series... Each spectrum in Perform calculation A differential spectral time series was obtained. Because background interferences that do not participate in the reaction and The absorption contributions in the two signals are basically the same, and this difference operation can suppress the influence of the background signal, making... The primary focus is on retaining the absorption peak information of methylene blue that decreases due to reduced concentration; subsequently, the processing unit processes each differential spectrum... In the analysis, the characteristic absorption peak of methylene blue was located at approximately 664 nm, and the peak height or integral area of ​​this peak was extracted as the intensity index at that moment. Thus, a time series is obtained. This time series forms a decay curve function. This refers to the actual dynamic change pattern; to perform quantitative characterization, the processing unit further extracts a set of numerical features from the decay curve function, specifically including the curve's... The initial decay rate at any given time, the nonlinearity parameter characterizing the curvature of the curve, and the time taken for the intensity to decay to 10% of its initial value—this set of eigenvectors together constitutes the final quantitative description of the actual dynamic change pattern.

[0038] After obtaining the quantitative characteristics representing the reaction kinetics, the system needs to correlate them with the initial concentration of methylene blue. This process is completed in step d, which involves inputting the actual dynamic change pattern into an artificial intelligence model to determine the concentration of methylene blue in the water sample. In this embodiment, the artificial intelligence model is a gradient booster model based on a decision tree, which has been trained through an offline model establishment procedure. This procedure includes preparing a series of methylene blue standard solutions with known concentrations, for example from 0.1 mg / L to 50 mg / L, with 2 mg / L intervals, free of kinetic interferences. For each standard solution, steps a to c are repeated to obtain its corresponding dynamic change pattern feature vector. This feature vector is used as input, and its known true concentration is used as output to supervise the training of the gradient booster model until the model converges. During online measurement, the actual dynamic change pattern extracted in step c is used... The feature vector is input into the trained model, which then outputs a corresponding predicted value for methylene blue concentration. A limitation of traditional online analysis methods is their inability to perceive the effectiveness of the measurement process itself online, especially when unknown kinetic disturbances that can alter the reaction rate are present in the wastewater. To address this issue, this method introduces an online self-diagnostic mechanism, initiated in step e, where the methylene blue concentration determined in step d is input into a standardized dynamic response model to generate an ideal dynamic change pattern corresponding to that concentration. The standardized dynamic response model is a mathematical model established during the model building phase by processing a to d steps of a methylene blue standard solution free of kinetic disturbances. It can be implemented as a function or lookup table, capable of returning a theoretical decay curve function that should appear at any input concentration in a pure chemical environment. For example, when the predicted concentration output in step d is 15.5 mg / L, the model will generate a standard decay curve corresponding to 15.5 mg / L methylene blue under undisturbed conditions.

[0039] The nonlinearity parameter of the characteristic absorption peak intensity decay curve mentioned in step c is specifically calculated as follows: First, the decay curve function corresponding to the actual dynamic change mode is... The least squares method is used to fit a single exponential decay function. Then, the sum of squared residuals between the original curve and the fitted function is calculated. and this The value serves as a nonlinearity parameter quantifying the curvature of the curve; while the normalized dynamic response model described in step e responds to a concentration value not used in the offline calibration. At that time, it generates an ideal dynamic change pattern. The specific method is to locate the points corresponding to the discrete concentration points stored within the model. Two adjacent concentration points and and its corresponding standard attenuation curve and And by performing linear interpolation on each data point on the time axis. To generate a value related to concentration. The system obtains a precise and complete ideal dynamic change pattern curve. After acquiring the observed dynamic pattern and the theoretical ideal pattern, the system compares the two to assess the reliability of the measurement. This process is completed in step f, which involves comparing the morphology of the actual dynamic change pattern obtained in step c with the ideal dynamic change pattern generated in step e, and determining the confidence level of the concentration measurement result based on the difference between the two. Here, the control and data processing unit will... and Treating them as two high-dimensional vectors, and obtaining a dynamic distortion index by calculating the morphological deviation between them. ;Should The calculation can follow these rules: ,in, For time variables, The total duration of the time window in step b is 5 seconds; The value quantifies the degree to which the actual reaction process deviates from the ideal state; a value close to 0... The value indicates that the measurement process was not disturbed by dynamics; subsequently, the processing unit will calculate... The value is compared with a pre-set confidence threshold. The threshold is compared. The setup is based on a calibration experiment, which involves repeated testing in samples containing typical concentrations of interfering substances and recording the results. The distribution of values, and select a statistical quantile that can effectively distinguish between interference and non-interference, for example, it can be set to... ;when When the processing unit marks the concentration measurement result output in step d as high confidence; when If the concentration measurement result is low confidence, a warning message indicating a kinetic disturbance can be simultaneously output to the central control system.

[0040] In some operating conditions, kinetic disturbances may persist, limiting the application of a system that can only provide alarms. To proactively correct for such disturbances, this method also includes an alternative implementation of proactive disturbance offsetting. Specifically, when the concentration measurement result is consecutively marked as low confidence (e.g., three times consecutively), the system automatically activates the proactive disturbance offsetting mode. This mode first executes step g, which involves performing a two-stage disturbance measurement on the water sample to be tested, including the injection of a first dose and a second dose of a chemical disturbance agent. With the second dose Different, for example , Next, step h is executed, which obtains the first dynamic change pattern induced by the first dose perturbation. The second dynamic change pattern induced by the second dose perturbation Finally, step i is executed by an AI model specifically trained for this pattern, based on... and The joint analysis of the nonlinear differences between the two is used to determine the methylene blue concentration that has been corrected for kinetic disturbances. The principle is that the reaction concentration effect between methylene blue and the disturbance agent exhibits an approximately linear superposition within a certain range, while the interaction between the kinetic disturbance and disturbance agents of different intensities usually shows a nonlinear relationship. This AI model, by learning this non-ideal proportional relationship, can reverse-calculate the systematic deviation caused by the disturbance and deduct it from the concentration calculation, thereby restoring the accuracy of the measurement under specific operating conditions. A key point is that the dedicated artificial intelligence model used for joint analysis in step i has the following offline training and construction procedures: First, a methylene blue standard solution test set covering a preset concentration range and containing known kinetic disturbances at different concentration levels is prepared, wherein the kinetic disturbance is a non-target substance capable of undergoing a consumption reaction with the chemical disturbance agent; second, a two-stage perturbation measurement involving the injection of a first and second dose of the chemical disturbance agent is performed on each solution sample in the test set to obtain a paired first dynamic change pattern. With the second dynamic change mode Secondly, from respectively and The initial decay rate and nonlinearity parameters are extracted, and the ratio or difference between the corresponding features of the two modes is calculated to form a combined feature vector that can characterize the nonlinear difference relationship. Finally, the combined feature vector is used as input, and the known true concentration of methylene blue in the corresponding sample is used as the output label to train a gradient boosting decision tree regression model until the prediction error of the model on the independent validation set is lower than the preset accuracy requirement, thereby obtaining a dedicated model that can determine the corrected concentration from the disturbed dual dynamic mode.

[0041] To ensure the long-term stability and traceability of the entire measurement system, this method may also include a procedure for self-diagnosing the integrity of the chemical probe. To achieve this function, during system hardware construction, a transparent tube storing the chemical perturbant to be used can be fixed in the optical path of the spectral acquisition device. This design ensures that when the baseline absorption spectrum is acquired in step a, the baseline absorption spectrum also contains a spectral fingerprint from the chemical perturbant itself, within a preset wavelength range. Upon initial installation, the system records the reference spectral fingerprint of the perturbant in its brand-new state. And store it; in each routine measurement, after executing step a, the system will initiate a diagnostic logic in parallel, that is, execute step k, from the acquired baseline absorption spectrum. In this study, the spectral fingerprint of the current chemical perturbator was extracted using mathematical methods. Then steps l and m are executed to... and The comparison is used to determine the chemical purity status of the chemical disruptor or the accuracy status of its injected dosage; specifically, through analysis Compared to The chemical purity is determined by whether new absorption peaks appear or the proportion of existing characteristic peaks changes; this is achieved through analysis. The integral area or peak height relative to The system determines the accuracy of the injected dose by checking for systematic deviations. In addition, the system can perform steps n and o, which continuously record the trend data of these states over time and infer the remaining effective life of the chemical disturbance agent or its injection system through a pre-established prediction model based on time series analysis. It also generates predictive maintenance instructions before the chemical disturbance agent fails, thereby changing the maintenance method of the instrument from passive maintenance to predictive maintenance.

[0042] Example 1: In a continuously operating textile dyeing and printing wastewater treatment facility, the upstream production conditions frequently change due to dye batches and process adjustments. This results in irregular, slow, but continuous changes in the types and concentrations of associated organic matter in the wastewater discharged into the treatment system, in addition to the target pollutant methylene blue. An online spectroscopic analysis system constructed using a conventional static chemometric model gradually deviates from the offline laboratory test results in the weeks following initial calibration. However, because this deviation is gradually accumulated, the system does not trigger an alarm, leading to downstream treatment units continuously overdosing treatment agents based on this erroneous data. To address this measurement failure scenario caused by background matrix drift, the technical solution disclosed in the aforementioned specific implementation method will be used for online detection. At the beginning of each measurement cycle, the system first executes step a, acquiring a baseline absorption spectrum containing all background interferences and the methylene blue signal at that time. This operation provides an instantaneous reference that perfectly matches the current water sample matrix for subsequent differential calculations, enabling the measurement process to adapt to dynamic changes in the background and avoiding the timeliness issues caused by relying on fixed historical data for calibration. Next, the system executes step b, injecting a dose of chemical disturbance agent into the water sample through a micro-injection device, and simultaneously acquiring high-frequency spectra to form a spectral time sequence. This step transforms a difficult problem requiring static analysis from mixed spectra into a problem that only needs to analyze the dynamic response of the target object itself. The background matrix, which was originally the main source of interference, does not participate in the rapid chemical reaction caused by the disturbance agent, and its influence is pre-isolated at the information processing level in this measurement method.

[0043] Subsequently, the system executes step c, by comparing each spectrum in the spectral time series with... Perform differential operations to generate differential spectral time series, and extract the actual dynamic change patterns from them. This differential operation utilizes the instantaneous baseline acquired in step a, allowing the spectral distortion caused by the slow drift of the background matrix to be considered constant within a short measurement time window of a few seconds and effectively suppressed during the subtraction process. This results in a dynamic signal with a high signal-to-noise ratio that characterizes only the methylene blue concentration decay process. Obtaining this signal is a prerequisite for the reliable analysis by the artificial intelligence model in subsequent step d. The two work together to separate the concentration measurement from the changes in the spectral background. Furthermore, to verify the reliability of this concentration value under the current chemical environment, the system executes steps e and f in parallel, i.e., generating an ideal dynamic change pattern based on this concentration value. and calculate and The dynamic distortion index between In this scenario, since the associated organic matter in the wastewater does not affect the reaction rate between methylene blue and the disruptor, the calculated... If the value is less than the preset confidence threshold, the system marks the concentration measurement result as high confidence and outputs it. This online self-diagnosis and confidence assessment mechanism, combined with the active micro-perturbation injection and differential processing mechanism, ensures that the measurement results are not affected by spectral background, while the latter confirms that the measurement process is not affected by kinetic background. The two work together to enable the system to output not only the concentration value, but also a self-assessment of the confidence of the value. This measurement method shifts the basis of analysis from a variable and uncontrollable external environmental factor, namely the spectral background of the water sample, to a stable and repeatable internal engineering parameter, namely the physicochemical reaction rate of the target substance itself induced by the standard chemical perturbator.

[0044] Example 2: To objectively verify the measurement accuracy and reliability of the technical solution of the present invention in a complex matrix environment containing dynamic spectral background interference and chemical reaction kinetic interference, this example constructs a comparative experiment including a control group and an experimental group. This experiment aims to quantify the performance of the method of the present invention compared to traditional static chemometrics methods under specific conditions simulating industrial wastewater. The experimental platform consists of a spectral acquisition device, a micro-injection device, and a control and data processing unit, and a peristaltic pump with controllable flow rate circulates the water sample to be tested through the detection point. To simulate common interference types in industrial settings, the experiment... Two interfering agent solutions were continuously and slowly injected into the circulation line using an additional micro-pump: one was a lignin sulfonate solution that does not participate in the chemical reaction but has broad-spectrum absorption, used to simulate spectral background drift; the other was a sodium sulfite solution that has no spectral characteristics but can react with the chemical perturbator to simulate kinetic interference. In the experiment, the selection of the chemical perturbator was consistent with the specific implementation method, and its injection dosage was set according to a principle of balancing reaction rate and reagent consumption, that is, the minimum effective dose was selected on the premise of ensuring that the intensity of the characteristic absorption peak of the target analyte decays by more than 80% within a 5-second measurement window.

[0045] The experiment consisted of two sample groups. The control group used a static chemometric model based on partial least squares to determine the concentration. This model was trained using the spectra of a series of methylene blue standard solutions prepared in pure water. The test group, i.e., the sample group of this invention, used a full-process method based on active perturbation and dynamic response analysis for measurement. Its built-in artificial intelligence model and normalized dynamic response model were trained and constructed offline using the same standard solution dataset as the control group model. The experimental procedure was set as follows: first, a series of methylene blue solutions with known true concentrations were prepared in pure water, and samples under two simulated disturbance conditions were measured. The measurement results of the control group and the sample group of this invention were recorded. The sample group of this invention also recorded its output kinetic distortion index. The confidence level determination results and several representative working condition data are recorded in Table 1.

[0046] Table 1: Comparison of measurement results of the two methods under different working conditions.

[0047] Table 1 shows that, under stable baseline conditions without interference, both methods can provide measurement results close to the true concentration. When spectral background drift interference is introduced, the measured values ​​of the control group deviate, with a relative error exceeding 30%, while the measured values ​​of the sample group of this invention remain stable, with a relative error of less than 2%. The value remained unchanged, and the confidence level was determined to be high. This result indicates that the method of the present invention can effectively suppress the influence of spectral background drift through active perturbation and differential processing. When kinetic interference is introduced, the control group is unaffected because the interfering substance has no spectral characteristics, but the measured value of the sample group of the present invention shows a positive deviation. Simultaneously, the calculated value also shows a change. If the value rises above 0.15, exceeding the confidence threshold of 0.05, the system will accordingly mark the measurement result as low confidence. This experimental result confirms that the method disclosed in this invention has advantages in measurement accuracy when dealing with spectral background drift. Furthermore, its online self-diagnosis and confidence assessment mechanism can mitigate the impact of unknown factors on the chemical reaction kinetics during the measurement process. The quantitative analysis of the indicators can identify unreliable states of the measurement results online, thereby preventing the system from outputting incorrect concentration data when it is in a failure state.

[0048] Example 3: This example combines Figures 1 to 3 This document describes an online methylene blue detection method and system based on artificial intelligence, such as... Figure 1As shown, baseline spectral acquisition is performed first to establish an immediate reference for subsequent differential operations. This includes the superposition information of the target analyte and the background matrix. Parallel to this step is a chemical probe integrity self-diagnosis process. This process monitors the chemical purity and injection dosage accuracy by analyzing the spectral fingerprint of the perturbator in the baseline spectrum and can output predictive maintenance instructions based on the analysis results. After baseline acquisition, the process enters the active micro-perturbation injection and time-series spectral capture stage. Chemical perturbators are injected to excite the dynamic response of the target analyte, and its spectral dynamics are recorded at high frequency. Subsequently, in the differential processing and dynamic feature extraction stage… Background spectral interference is suppressed by performing differential operations with the baseline spectrum, and the actual dynamic pattern characterizing the spectral features of the target analyte is separated and quantified. This actual dynamic pattern is input into the concentration inversion module of the artificial intelligence model to calculate the concentration of methylene blue. At the same time, the concentration result is input into a normalized dynamic response model to generate a theoretically ideal dynamic change pattern in a pure chemical environment. Finally, in the online self-diagnosis and confidence assessment stage, the confidence level of the concentration measurement result is determined by comparing the difference between the actual dynamic pattern and the ideal dynamic change pattern, and the concentration measurement result with a confidence level label is finally output.

[0049] like Figure 2 As shown, the dynamic distortion index calculated by this mechanism over a continuous monitoring period of 72 hours is illustrated. The trend curve of the value changing with running time, where the horizontal axis represents the running time in hours and the vertical axis represents the value of the value. Values, where the solid line in the graph represents real-time calculated values. The dashed line represents the pre-set confidence threshold, which is 0.05. As shown in the graph, approximately 46 hours ago, the measured... The value consistently fluctuated below the confidence threshold, indicating that the measurement process was not affected by dynamics. However, after approximately 46 hours, The value shows a sustained and significant increase, and then stabilizes above the confidence threshold. This phenomenon clearly indicates the presence of substances in the water sample that can interfere with reaction kinetics. Based on this, the system can mark the subsequent concentration measurement results as low confidence.

[0050] like Figure 3As shown, the system mainly consists of a spectral acquisition device, a micro-injection device, and a core control and data processing unit. Specifically, the control and data processing unit can be an industrial computer. The spectral acquisition device and the control and data processing unit communicate through a signal connection containing control and data signals. The micro-injection device receives instructions from the control and data processing unit through a signal connection containing control signals. Inside the control and data processing unit, there is a pre-installed software module that implements the method of this invention. Its logical flow is as follows: the top-level control and analysis application schedules and calculates the entire measurement process. During the calculation process, the application calls the built-in artificial intelligence model to perform concentration inversion and calls the normalized dynamic response model to generate an ideal dynamic pattern for confidence assessment.

[0051] Example 4: When an online detection system using the method of this invention, deployed at a chemical production wastewater discharge outlet, has been running continuously for several months, the logs of its control and data processing unit show that the concentration measurement results for methylene blue have been continuously marked as low confidence for the past 72 hours. This phenomenon is triggered by the online self-diagnosis and confidence assessment mechanism, specifically manifested in the calculated kinetic distortion index. If the value remains consistently above the preset confidence threshold, it indicates the presence of one or more kinetic disruptors in the water sample that can alter the degradation rate of methylene blue, and that this disruption is persistent rather than transient. To restore the accuracy of concentration measurement under such persistent kinetic interference, the system is configured to automatically execute an active interference countermeasurement procedure. The system first executes a procedure that includes injecting the first dose. for Measurements of chemical perturbators captured the first dynamic change patterns induced by them. Following this, after the water sample was replaced, a second dose was administered. for Measurements of chemical perturbators captured the second dynamic change pattern. The dedicated analysis module within the control and data processing unit, through... and The nonlinear relationship between the morphological differences between the two is jointly analyzed, and a set of equations for separating the concentration contribution and the interference contribution is established and solved. Finally, a methylene blue concentration value that has been corrected for kinetic interference is output.

[0052] To ensure the long-term effectiveness of the online self-diagnosis and confidence assessment mechanism, especially the key parameter used as the basis for judgment, namely the confidence threshold. In addition to the above settings, this system also includes a periodic parameter calibration procedure. This procedure is set to be executed automatically every 30 days. The steps are as follows: First, the system pauses the measurement of the actual water sample and switches to the internal pipeline, injecting pure water to clean the flow cell, and measuring the kinetic response at this time as the zero-point reference. Second, the system automatically extracts three methylene blue standard solutions of known concentrations without interfering substances from the built-in standard storage tank, such as 5.0 mg / L, 20.0 mg / L, and 40.0 mg / L, and repeats the complete measurement process three times for each standard solution. The system records the kinetic distortion index of these nine measurements. Values ​​and calculate their mean. with standard deviation Ultimately, the system will set a new confidence threshold. Set as This new threshold is then used in subsequent online measurement judgments. Through this procedure, the system can autonomously adapt to slow changes in the measurement baseline noise level caused by factors such as light source aging or optical path contamination, thus maintaining the reliability of the confidence assessment system. Simultaneously, in this periodic parameter calibration procedure, the system also verifies the effectiveness of the artificial intelligence model. While measuring the three standard solutions mentioned above, the system compares the model's output concentration prediction with the known true concentration of the standard solution. If the measurement error of any standard solution exceeds the preset allowable error range (e.g., 5%) for two consecutive calibration cycles, the system will record a maintenance instruction in the log suggesting model retraining, prompting operators to perform a complete offline retraining of the artificial intelligence model using a new series of standard solutions to update the model parameters. This mechanism ensures that the core algorithm model of the measurement system can be periodically verified and maintained, thereby maintaining the measurement accuracy of the entire system during long-term operation.

[0053] Example 5: To establish a standardized dynamic response model for online self-diagnosis and confidence assessment, this example discloses an offline calibration and data filling procedure; this procedure is performed at a constant temperature of 25°C. The process is carried out in a controlled environment. First, a series of methylene blue standard solutions with concentrations ranging from 0.1 mg / L to 50 mg / L are prepared using high-purity water, totaling 50 concentration gradient points. For each concentration point, the standard solution is used to perform ten complete measurements using the calibrated online detection system of this invention, i.e., steps a to c, to obtain ten corresponding actual dynamic change pattern curves. Subsequently, the control and data processing unit averages the ten curves corresponding to each concentration point to generate a high signal-to-noise ratio baseline dynamic change pattern representing that concentration, and stores it and the extracted quantitative features in a lookup table. This lookup table constitutes a standardized dynamic response model, which can return the corresponding ideal dynamic change pattern based on any input concentration value during the online measurement stage.

[0054] To ensure the accuracy of the self-diagnostic function of the chemical probe integrity, this embodiment also discloses a pre-calibration procedure to be performed after the initial installation of the system or after each replacement of the chemical perturber. After the procedure is started, the system first flushes the flow cell and pipeline with pure water until the reading of the spectral acquisition device is stable. Then, with the water sample flow stopped, the micro-injection device is controlled to inject a brand-new, unused chemical perturber into the optical path area of ​​the spectral acquisition device and leave it there. The spectral acquisition device then acquires a high-resolution absorption spectrum, and the control and data processing unit subtracts the pre-stored pure water background spectrum to obtain a pure chemical perturber spectral fingerprint. This spectral fingerprint is stored by the system as a reference spectral fingerprint and used as a benchmark for comparing the chemical purity status or the accuracy of the injected dosage of the chemical perturber during subsequent online monitoring.

[0055] Example 6: To ensure the stability of the hardware status and the adaptation of key process parameters of the online detection system during long-term operation, this example discloses a standardized self-test and parameter optimization procedure executed during system startup or periodic maintenance. After the procedure is started, the system first injects pure water into the flow cell through internal pipelines and controls the spectral acquisition device to acquire spectral data. The control and data processing unit analyzes the data and compares the real-time light intensity of the light source with the reference light intensity calibrated at the factory. If it is lower than the preset threshold, a light source maintenance warning is generated. At the same time, the unit also evaluates the dark current noise level of the detector to confirm that its working status is normal.

[0056] After completing the hardware self-test, the system enters parameter optimization mode to determine the appropriate chemical disturbance agent injection dosage for a specific range of water sample concentrations. In this mode, a methylene blue standard solution with a concentration at the upper limit of the test range is first injected into the flow cell through the pipeline. Starting from a preset minimum dosage, the system performs a complete disturbance-response measurement and calculates the percentage decay of the methylene blue characteristic absorption peak within a preset time window. If the decay percentage is lower than a preset reaction completion target threshold, the control and data processing unit increases the injection dosage in a fixed increment and repeats the measurement and calculation until the decay percentage is greater than or equal to the target threshold. The system then stores the injection dosage used in this measurement as the first dosage for routine measurements. And store a predetermined multiple, such as twice the first dose, as the second dose for active interference hedging measurements. .

[0057] To further verify the necessity of the online self-diagnosis and confidence assessment steps (i.e., steps e and f) included in this invention for ensuring the reliability of measurement results, the following comparative examples are provided.

[0058] Comparative Example 1: This comparative example aims to demonstrate that, even with the use of active perturbation and dynamic response analysis techniques, the measurement system cannot guarantee the reliability of results when faced with specific interferences, especially in the absence of the online self-diagnosis and confidence assessment steps claimed in this invention. The experimental platform, testing procedures, selection of chemical perturbators, and introduction methods of the interfering substances (lignin sulfonate and sodium sulfite) in this comparative example are completely consistent with those described in Example 2 above. The only difference is that the analytical method used in this comparative example, after calculating the methylene blue concentration using an artificial intelligence model (i.e., completing the present invention),... Steps a to d of the method directly output the concentration value as the final measurement result, without performing steps e (generating an ideal dynamic change pattern) and f (comparing the actual and ideal patterns and performing confidence assessment) in the method of this invention. Using this analytical method that omits the online self-diagnosis and confidence assessment steps, a series of water samples were measured under the same test conditions as in Example 2, and the measurement results are recorded in Table 2. For easy direct comparison, Table 2 also lists the measurement data of the control group (based on a static chemometric model) in Example 2 and the test group using the method of this invention.

[0059] Table 2: Comparison of measurement results of the three methods under different working conditions.

[0060] As shown in Table 2, under conditions of no interference or only spectral background drift, both the method of Comparative Example 1 and the method claimed in this invention can provide accurate measurement results, and are superior to the control group using a static model. However, when a kinetic disruptor (sodium sulfite) without spectral characteristics but capable of altering the reaction rate is introduced into the water sample, the measurement value of the method of Comparative Example 1 shows a relative error of over 70%. Because this method lacks a self-diagnostic mechanism for the effectiveness of the measurement process, it cannot identify that the measurement has been contaminated by the kinetic process, and thus directly outputs a biased and undetectable erroneous result. In stark contrast, the test group using the method of this invention, although its internal artificial intelligence model was also affected by kinetic interference and output a higher initial concentration value (e.g., 8.5 mg / L), its subsequent online self-diagnosis and confidence assessment steps, through the calculation of the kinetic distortion index... (When its value rises above 0.15), the difference between the actual dynamic pattern and the ideal dynamic pattern is successfully identified, and the measurement result is correctly marked as low confidence. The experimental results show that without the step of comparing the actual dynamic change pattern with the ideal dynamic change pattern and determining the confidence level, the analysis system will be unable to identify and respond to chemical kinetic interference from the water sample background online, thus outputting erroneous concentration data with serious deviations that cannot be detected by the system itself under certain operating conditions. This precisely proves that the online self-diagnosis and confidence assessment links included in this invention are key technical components to ensure the necessary reliability of the online monitoring system in complex industrial environments.

[0061] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An online methylene blue detection method based on artificial intelligence, characterized in that, The method includes: Step a: Collect the baseline absorption spectrum of the water sample to be tested flowing through a detection point; Step b: After completing step a, inject a dose of chemical perturbation agent into the water sample to be tested to actively stimulate dynamic changes in the spectral absorption characteristics of methylene blue in the water sample, and continuously acquire multiple absorption spectra within the time window after injection to form a spectral time sequence. Step c: Perform a differential operation on each spectrum in the spectral time series with the baseline absorption spectrum acquired in step a to generate a differential spectral time series that only represents dynamic changes, and extract the actual dynamic change pattern that represents the intensity decay process of the methylene blue characteristic absorption peak from the differential spectral time series. Step d: Input the actual dynamic change pattern into the artificial intelligence model to determine the concentration of methylene blue in the water sample to be tested; Step e: Input the concentration of methylene blue determined in step d into a normalized dynamic response model to generate an ideal dynamic change pattern corresponding to the concentration. The normalized dynamic response model is established in advance by processing a to d of a standard methylene blue solution without kinetic interference. Step f: Compare the actual dynamic change pattern obtained in step c with the ideal dynamic change pattern generated in step e, and determine the confidence level of the concentration measurement result based on the difference between the two. Furthermore, the process of determining the confidence level of the concentration measurement result based on the difference between the two in step f specifically includes: treating the actual dynamic change pattern and the ideal dynamic change pattern as two high-dimensional vectors and calculating the morphological deviation between them to obtain a kinetic distortion index; comparing the kinetic distortion index with a pre-set confidence threshold; marking the concentration measurement result as high confidence when the kinetic distortion index is less than the confidence threshold; and marking the concentration measurement result as low confidence when the kinetic distortion index is not less than the confidence threshold, and outputting a warning message indicating the presence of kinetic interference. The dynamic distortion index in step f The calculation follows these rules: ,in, The decay curve function corresponding to the actual dynamic change mode obtained from step c. The decay curve function corresponding to the ideal dynamic change mode generated in step e. For time variables, This represents the total duration of the time window in step b.

2. The online methylene blue detection method based on artificial intelligence according to claim 1, characterized in that, The actual dynamic change mode in step c includes at least one of the following: the initial decay rate of the characteristic absorption peak intensity, the nonlinearity parameter of the characteristic absorption peak intensity decay curve, and the time required for the characteristic absorption peak intensity to reach a steady state. In step b, the chemical perturbator is an oxidant that can chemically react with methylene blue to decompose it.

3. The online methylene blue detection method based on artificial intelligence according to claim 1, characterized in that, When the concentration measurement results are continuously marked as low confidence, the method also automatically initiates active interference offsetting, which includes: performing a two-stage perturbation measurement on the water sample to be tested, which includes injecting a first dose and a second dose of chemical perturbant, wherein the first dose and the second dose are different; obtaining a first dynamic change pattern caused by the first dose perturbation and a second dynamic change pattern caused by the second dose perturbation; and determining the methylene blue concentration that has been corrected for kinetic interference based on the joint analysis of the nonlinear difference relationship between the first dynamic change pattern and the second dynamic change pattern.

4. The online methylene blue detection method based on artificial intelligence according to claim 1, characterized in that, The method also includes a self-diagnosis of chemical probe integrity, which includes: ensuring that when acquiring the baseline absorption spectrum in step a, the baseline absorption spectrum also contains a spectral fingerprint from the chemical perturbator itself within a preset wavelength range; extracting the spectral fingerprint of the chemical perturbator itself from the acquired baseline absorption spectrum; comparing the extracted spectral fingerprint with a pre-stored reference spectral fingerprint representing the chemical perturbator in a brand-new state; and determining the chemical purity state of the chemical perturbator or the accuracy state of its injection dose based on the comparison result of step a.

5. The online methylene blue detection method based on artificial intelligence according to claim 4, characterized in that, In step a, the chemical purity of the chemical perturbator is determined, specifically by analyzing whether new absorption peaks appear or the proportion of existing characteristic peaks changes relative to the reference spectral fingerprint; the accuracy of the injected dose is determined, specifically by analyzing whether the integral area or peak height of the spectral fingerprint has undergone a systematic shift relative to the integral area or peak height of the reference spectral fingerprint.

6. The online methylene blue detection method based on artificial intelligence according to claim 5, characterized in that, The method further includes: continuously recording the trend data of the chemical purity state or the accuracy state of its injected dosage determined by step a over time; and based on the trend data, inferring the remaining effective life of the chemical disturbance agent or its injection system through a pre-established prediction model, and generating predictive maintenance instructions before its failure.

7. The online methylene blue detection method based on artificial intelligence according to claim 1, characterized in that, In step d, the artificial intelligence model is either a gradient boosting machine model based on decision trees or a recurrent neural network model.

8. An artificial intelligence-based online methylene blue detection system, used to execute the method of any one of claims 1 to 7, characterized in that, The system includes: A flow pool equipped with detection points; A spectral acquisition device, the optical path of which passes through the detection point; A micro-injection device, the outlet of which is connected to a flow-through cell; A control and data processing unit is connected to a spectral acquisition device and a micro-injection device. The control and data processing unit is configured to: control the spectral acquisition device to acquire the baseline absorption spectrum of the water sample flowing through the detection point; after acquiring the baseline absorption spectrum, control the micro-injection device to inject a dose of chemical perturbation agent into the water sample, and simultaneously control the spectral acquisition device to continuously acquire multiple absorption spectra within a time window after injection to form a spectral time series; and perform a difference operation between each spectrum in the spectral time series and the baseline absorption spectrum to generate a single-tablet representation. The method involves generating a differential spectral time series characterized by dynamic changes, and extracting the actual dynamic change pattern representing the attenuation process of the characteristic absorption peak intensity of methylene blue from this differential spectral time series. The actual dynamic change pattern is then input into an artificial intelligence model stored in the unit to determine the concentration of methylene blue in the water sample to be tested. The determined concentration of methylene blue is then input into a normalized dynamic response model stored in the unit to generate an ideal dynamic change pattern corresponding to that concentration. The obtained actual dynamic change pattern and the generated ideal dynamic change pattern are compared in morphology, and the confidence level of the concentration measurement result is determined based on the difference between the two.

Images