Seawater sample residual chlorine online monitoring method combined with intelligent sensing

Abstract

The invention relates to the technical field of environment monitoring and intelligent sensing, and discloses a seawater sample residual chlorine online monitoring method combined with intelligent sensing. The method comprises the following steps: synchronously acquiring multi-source heterogeneous data through a multi-mode electrochemical sensing array and a micro ultraviolet-visible absorption spectrum module, carrying out feature decoupling by utilizing a deep auto-encoder and an independent component analysis joint model, extracting a free residual chlorine exclusive component, and correcting salinity, turbidity and organic matter interference in combination with a dynamic environment compensation model, so as to obtain the free residual chlorine content. And finally, high-selectivity and high-sensitivity real-time concentration calculation is realized. The device has adaptive signal decoupling and on-line model fine tuning capabilities, inhibits cross interference and long-term drift, realizes reagent-free and continuous on-line monitoring, and improves the accuracy and reliability of residual chlorine detection in the marine environment.

Description

Technical Field

[0001] This invention belongs to the field of environmental monitoring and intelligent sensing technology, specifically relating to an online monitoring method for residual chlorine in seawater samples that combines intelligent sensing. Background Technology

[0002] With the rapid development of marine environmental monitoring, seawater desalination, and coastal industrial cooling water treatment, the demand for real-time and accurate online monitoring of residual chlorine concentration in seawater is becoming increasingly urgent. As a key indicator for measuring disinfection effectiveness and ecological risk, the concentration level of residual chlorine directly affects the safety of marine ecosystems and human health.

[0003] Traditional residual chlorine detection methods mainly rely on electrochemical sensors or colorimetric methods. While these methods have some application in freshwater systems, they face significant challenges in complex seawater matrices. Seawater is rich in high concentrations of salt, bromides, iodides, and organic amines, which readily react with residual chlorine to form structurally similar compounds such as chloramines and hypobromic acid. This results in severe cross-interference of sensor response signals, making it difficult to achieve specific identification of target substances.

[0004] Detection technologies based on single-sensor principles generally suffer from poor selectivity and low stability due to their lack of ability to distinguish complex components. In recent years, microfluidic chip technology has been explored for use in water quality analysis pretreatment due to its advantages such as low sample consumption, high separation efficiency, and strong integration. However, existing designs mostly focus on physical filtration or simple laminar flow separation, failing to effectively simulate the dynamic discrimination mechanism of biological systems against mixed odor molecules, and making it difficult to achieve efficient decoupling of interfering substances from target residual chlorine at the molecular level.

[0005] At the same time, although artificial intelligence algorithms have made progress in the field of signal processing, conventional deep learning models rely on a large amount of labeled data and have high computational costs, which cannot meet the requirements of online monitoring for low power consumption and real-time response, and lack biomimetic reference to the coding mechanism of biological olfactory nerves.

[0006] In the existing technology, no solution has yet deeply integrated the dynamic recognition logic of biomimetic olfaction with the separation function of the microfluidic front end. The microfluidic channel design does not introduce biomimetic gradient diffusion or selective adsorption structures, and cannot reproduce the spatiotemporal response differences of biological olfactory epithelium to different volatile molecules; the signal processing link still uses a static classifier, ignoring the impulse response characteristics of the multi-channel sensor array in the time dimension, resulting in limited overall system specificity.

[0007] Especially in high-salinity and highly turbulent seawater environments, these defects are further amplified, making residual chlorine detection results susceptible to the influence of coexisting substances such as chloramines and bromides, resulting in significant false positives or false negatives. This severely restricts the reliability and applicability of online monitoring systems. Therefore, there is an urgent need for a novel online monitoring method for residual chlorine in seawater that integrates biomimetic recognition mechanisms and microfluidic separation technology to overcome the bottlenecks in specificity and robustness of existing technologies. Summary of the Invention

[0008] This invention provides an online monitoring method for residual chlorine in seawater samples using intelligent sensing, aiming to solve the technical problems of difficulty in distinguishing residual chlorine from similar substances such as chloramines and bromides, and insufficient specificity. The method achieves high selectivity, high sensitivity, and real-time online monitoring of free residual chlorine in seawater by constructing a multimodal electrochemical sensing array and a spectral feature fusion identification mechanism, combined with a dynamic environmental compensation model and an adaptive signal decoupling algorithm.

[0009] This invention provides a method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing, comprising: Seawater samples are collected and flow through a multi-channel sensing reaction chamber; Multiple electrochemical response signals are acquired simultaneously using a multimodal electrochemical sensing array; The multimodal electrochemical sensing array includes a first sensing unit, a second sensing unit, and a third sensing unit; The first sensing unit uses a platinum-plated working electrode and is modified with a nano-titanium dioxide catalytic layer to produce a broad-spectrum response to chlorine-containing oxidizing substances. The second sensing unit uses a gold-plated working electrode and is fixed with a molecularly imprinted polymer film with free residual chlorine as a template molecule to specifically identify free residual chlorine. The third sensing unit employs a glassy carbon working electrode and is coated with a polyaniline conductive polymer layer to respond to bromide and chloramine interferences. The absorbance values ​​of seawater samples at four characteristic wavelengths of 210 nm, 254 nm, 310 nm and 365 nm were obtained using a miniature ultraviolet-visible absorption spectroscopy module. The electrochemical response signal and the absorbance value are spatiotemporally aligned to form a multi-source heterogeneous sensing dataset containing 7 dimensions; The multi-source heterogeneous sensing dataset is subjected to component separation processing based on a preset feature decoupling model, and feature components related only to free residual chlorine are extracted. The feature decoupling model is a neural network model based on a joint architecture of deep autoencoder and independent component analysis, which outputs three statistically independent latent variables. The concentration of free residual chlorine in the seawater sample is calculated based on the characteristic components, and the monitoring results are output.

[0010] Preferably, the multi-channel sensing reaction chamber is equipped with a microfluidic structure, including a main sample inlet channel, a diversion channel, a mixing chamber, and a waste liquid outlet; The main sampling channel is connected to a seawater sampling pump to introduce the seawater sample to be tested. The diversion channel divides the seawater sample into three paths, which are respectively introduced into the detection chambers corresponding to the first sensing unit, the second sensing unit, and the third sensing unit. Each detection chamber is equipped with a constant temperature heating plate at the bottom to maintain the reaction temperature at 25 degrees Celsius; each detection chamber is equipped with a miniature stirring magnet at the top, driven by an external rotating magnetic field to ensure that the sample makes uniform contact with the sensing interface.

[0011] Preferably, the miniature ultraviolet-visible absorption spectroscopy module includes a deuterium lamp light source, a quartz colorimeter, a grating spectrometer, and a linear photodiode array detector; The quartz colorimeter is integrated into the bypass channel of the multi-channel sensing reaction chamber; The grating beam splitter disperses the transmitted light according to wavelength and projects it onto the linear photodiode array detector.

[0012] Preferably, the spatiotemporal alignment of the multi-source heterogeneous sensor dataset is achieved through a hardware-level synchronization triggering mechanism; The multimodal electrochemical sensing array and the miniature ultraviolet-visible absorption spectroscopy module share the same clock source; At each sampling time, the current response values ​​of three electrochemical sensing units and the absorbance values ​​of four characteristic wavelengths are recorded simultaneously to form a 7-dimensional vector; Data collected continuously for 60 seconds forms a time window, which serves as the input to the feature decoupling model.

[0013] Preferably, the feature decoupling model includes an encoder, a decoder, and an independent component constraint layer; The encoder consists of a three-layer fully connected neural network; The decoder structure is symmetrical to the encoder; the independent component constraint layer is embedded between the last layer of the encoder and the first layer of the decoder, forcing the three latent variables of the output to be statistically independent of each other; The model was trained using a standard seawater sample dataset prepared artificially, with the label data being the actual concentrations of free residual chlorine, monochloramine, dichloramine, nitrogen trichloride, bromide ions, and hypobromic acid.

[0014] Preferably, during the model inference stage, real-time collected multi-source heterogeneous sensor data is input into the trained feature decoupling model to obtain three independent components; The independent components corresponding to free residual chlorine were determined using the component-substance mapping table; The component-substance mapping table was established through offline calibration experiments. The specific method is as follows: A single standard substance was added to chlorine-free artificial seawater in incremental amounts, and the trend of the response intensity of each independent component was recorded. Components whose response intensity monotonically increased with the concentration of free residual chlorine and had a weak response to other substances were marked as "chlorine-specific components".

[0015] Preferably, the monitoring method further includes a dynamic environmental compensation step; The dynamic environmental compensation model is used to correct the interference of seawater salinity, turbidity and organic matter background on the monitoring results; The input variables of the dynamic environmental compensation model include seawater temperature, conductivity, dissolved oxygen concentration and total organic carbon content, and the output is the correction coefficient for residual chlorine-specific components. The seawater temperature was measured by a platinum resistance temperature sensor. The conductivity was measured using a four-electrode conductivity cell; the dissolved oxygen concentration was measured using a fluorescence quenching dissolved oxygen probe. The total organic carbon content was calculated using an empirical formula based on the absorbance value at 254 nm.

[0016] Preferably, the monitoring method further includes an adaptive signal decoupling algorithm; When the system detects that the change in residual chlorine concentration is less than 5% within 30 consecutive minutes and the environmental parameters are stable, the model fine-tuning mode is activated. In model fine-tuning mode, the system collects current seawater samples and injects a small amount of free residual chlorine standard solution of known concentration to form a perturbed sample; Input the multi-source heterogeneous sensing data of the perturbed sample into the current model to calculate the deviation between the predicted concentration and the actual added concentration; If the absolute value of the deviation is greater than the set value of 0.02 mg / L, the backpropagation algorithm is triggered to update the weights of the feature decoupling model in small steps.

[0017] Preferably, the monitoring results are uploaded to the central monitoring platform in real time via an industrial Ethernet interface or an RS485 bus; When the residual chlorine concentration is detected to be greater than or equal to the preset safety threshold, an audible and visual alarm is triggered and the event log is recorded. The event log includes timestamps, raw sensor data, decoupled feature values, compensated concentration values, and environmental parameters.

[0018] Preferably, the nano-titanium dioxide catalyst layer is synthesized by the sol-gel method.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By constructing a multimodal electrochemical sensing array composed of specific molecularly imprinted electrodes, broad-spectrum response electrodes, and interference-sensitive electrodes, and combining it with multi-wavelength feature extraction of micro UV-Vis absorption spectroscopy, a differentiated response capability to various chlorine- and bromine-containing oxidizing substances in seawater was formed. 2. A feature decoupling model based on a joint architecture of deep autoencoder and independent component analysis is introduced to separate the exclusive feature component of free residual chlorine from multi-source heterogeneous sensing data, and suppress the cross-interference of coexisting substances such as chloramine and bromide. 3. An integrated dynamic environmental compensation model is used to correct the monitoring signal in real time using seawater physicochemical parameters, which improves the measurement stability and accuracy in complex marine environments; 4. An adaptive signal decoupling algorithm is adopted, which enables the system to have online learning and model fine-tuning capabilities, and can cope with long-term drift factors such as sensor aging and seasonal changes in seawater conductivity. 5. The overall solution achieves high specificity, high precision, reagent-free, and continuous online monitoring of free residual chlorine in seawater, solving the problems of false alarms and missed alarms caused by poor selectivity in existing technologies, and providing reliable technical support for marine ecological environment protection, disinfection control in seawater desalination plants, and safety monitoring of cooling water in nuclear power plants. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the feature decoupling model based on the joint architecture of deep autoencoder and independent component analysis in this invention; Figure 3 This is a flowchart illustrating the data acquisition and spatiotemporal alignment logic of the multimodal electrochemical sensing array and the micro UV-Vis absorption spectroscopy module in this invention. Figure 4 This is a flowchart illustrating the logical process of constructing and applying the dynamic environment compensation model in this invention. Figure 5 This is a flowchart illustrating the logical framework of the online learning and model fine-tuning mechanism of the adaptive signal decoupling algorithm in this invention. Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow between the multi-channel sensing reaction cavity and the multi-source sensing unit in this invention. Detailed Implementation

[0021] refer to Figures 1 to 6This invention provides an online monitoring method for residual chlorine in seawater samples using intelligent sensing. Its core lies in the collaborative operation of a multimodal electrochemical sensing array and a miniature ultraviolet-visible absorption spectroscopy module to construct a highly specific and sensitive free residual chlorine identification mechanism. Furthermore, it integrates a dynamic environmental compensation model and an adaptive signal decoupling algorithm to achieve real-time, continuous, and reagent-free online monitoring of the free residual chlorine concentration in seawater. The specific implementation of the method will be described in detail below according to the steps explicitly listed in the invention description.

[0022] The method first performs step S1: collecting seawater samples and allowing them to flow through a multi-channel sensing reaction chamber. This multi-channel sensing reaction chamber is a microfluidic structure, consisting of a main inlet channel, a diversion channel, a mixing chamber, and a waste outlet. The main inlet channel is connected to a seawater sampling pump, which employs a peristaltic diaphragm structure to extract the seawater sample at a constant flow rate of 50 ml per minute, ensuring no air bubbles are introduced and that the flow rate is stable during the sampling process.

[0023] The shunt channel has a Y-shaped symmetrical layout, which divides the seawater sample in the main injection channel into three paths, which are then introduced into three independent detection chambers. Each detection chamber corresponds to a multimodal electrochemical sensing unit.

[0024] Each detection chamber has an integrated constant temperature heating element at the bottom, with a temperature control accuracy of ±0.2 degrees Celsius, maintaining the reaction temperature at a constant 25 degrees Celsius to eliminate electrochemical response drift caused by seawater temperature fluctuations.

[0025] Each detection chamber has a built-in miniature stirring magnet with a diameter of 1.5 mm, which is driven by an external rotating magnetic field and rotates at 600 revolutions per minute. This creates a uniform vortex in the seawater sample within the detection chamber, ensuring full contact between the sensing interface and the sample, and improving response consistency and repeatability.

[0026] Then, step S2 is performed: multiple electrochemical response signals are simultaneously acquired using a multimodal electrochemical sensing array. The multimodal electrochemical sensing array consists of three functionally independent electrochemical sensing units.

[0027] The first sensing unit uses a platinum-plated working electrode, with a titanium alloy disk on the substrate. The disk has a diameter of 5 mm and a 200-nanometer-thick platinum film deposited on its surface by magnetron sputtering. Then, a nano-titanium dioxide catalytic layer is modified on top of it. This nano-titanium dioxide catalytic layer is synthesized by the sol-gel method, with an average particle size of 20 nanometers and a porosity of 45%. It is used to generate a broad-spectrum current response to chlorine-containing oxidizing substances, including free residual chlorine, hypochlorite, chloramine, etc.

[0028] The second sensing unit uses a gold-plated working electrode. The substrate is a stainless steel cylinder with a 150-nanometer-thick gold film electroplated on the surface. A molecularly imprinted polymer film is fixed on it. The molecularly imprinted polymer film uses free residual chlorine as a template molecule, methacrylic acid as a functional monomer, and ethylene glycol dimethacrylate as a crosslinking agent. It is polymerized in situ under ultraviolet light initiation. The film thickness is 500 nanometers. It has a high specific recognition ability for free residual chlorine and a cross-response of less than 5% to interfering substances such as chloramine and bromide.

[0029] The third sensing unit uses a glassy carbon working electrode with a diameter of 4 mm. A layer of polyaniline conductive polymer with a thickness of 300 nm is coated on the surface by spin coating. This polyaniline conductive polymer layer has a selective oxidation response to bromide ions, hypobromic acid and chloramines. Its current signal can be used as a reference indicator for interfering components.

[0030] The three sensing units share the same reference electrode (silver / silver chloride electrode) and the same pair of auxiliary electrodes (platinum wire), forming a three-electrode system. The operating potentials are set to 0.8 V, 0.6 V, and 0.9 V (relative to the reference electrode), respectively, and are uniformly controlled by a potentiostat. All electrochemical signals are amplified by a low-noise analog front-end circuit and then sampled by a 24-bit analog-to-digital converter at a frequency of 10 times per second, with the output being a current value in microamps.

[0031] Next, step S3 is performed: the absorption spectrum data of the seawater sample in a specific wavelength band is acquired using a miniature ultraviolet-visible absorption spectroscopy module. This miniature ultraviolet-visible absorption spectroscopy module is integrated into the bypass channel of the multi-channel sensing reaction chamber. This bypass channel is connected in parallel with the main sample inlet channel, ensuring that the sample flows through the optical detection zone simultaneously with the electrochemical detection chamber. The module includes a deuterium lamp source, a quartz cuvette, a grating spectrometer, and a linear photodiode array detector.

[0032] The deuterium lamp emits a continuous spectrum covering a wavelength range of 190 nm to 400 nm, with a light intensity stability better than 0.5%. The quartz cuvette has an optical path length of 10 mm, and its inner wall is treated with a superhydrophilic coating to prevent air bubble adhesion. The light-transmitting window is made of fused silica, with a transmittance greater than 80% at 200 nm. The grating beam splitter is a planar reflective type with a line density of 1200 lines per millimeter and a dispersion resolution of 0.3 nm per pixel.

[0033] The linear photodiode array detector contains 1024 pixel units with an effective photosensitive area of ​​30 mm in length, and each pixel corresponds to one wavelength channel. The system extracts absorbance values ​​at only four characteristic wavelengths: 210 nm (corresponding to hypochlorite ions). (Leap), 254 nm (corresponding to the absorption of aromatic organic compounds and total organic carbon), 310 nm (corresponding to chloramines) (Transition) and 365 nm (corresponding to charge transfer absorption of brominated compounds). Absorbance values ​​were calculated using Beer-Lambert's law, with deionized water as a blank reference. The sampling frequency was synchronized with the electrochemical module at 10 times per second.

[0034] Then, step S4 is executed: the electrochemical response signal and absorption spectral data are spatiotemporally aligned to form a multi-source heterogeneous sensing dataset. Spatiotemporal alignment is achieved through a hardware-level synchronization triggering mechanism. The system's main control chip has a built-in single high-stability clock source with a frequency of 100 MHz, which generates a unified sampling trigger signal via a frequency divider and simultaneously sends it to both the electrochemical signal acquisition module and the spectral data acquisition module.

[0035] At each trigger moment, the system synchronously records the current value of the first sensing unit. Current value of the second sensing unit The current value of the third sensing unit and absorbance values ​​at four wavelengths. , forming a 7-dimensional vector 600 vectors were continuously collected over 60 seconds, forming a time window data block, which served as the input for the subsequent feature decoupling model. The data block was stored in memory as a circular buffer; when new data was written, the oldest data was overwritten, ensuring that the analysis always used data from the most recent 60 seconds.

[0036] Next, step S5 is performed: based on a preset feature decoupling model, the multi-source heterogeneous sensing dataset is subjected to component separation processing to extract feature components that are only related to free residual chlorine. The feature decoupling model is a neural network model based on a joint architecture of deep autoencoder and independent component analysis. This neural network model includes an encoder, a decoder, and an independent component constraint layer.

[0037] The encoder consists of a 3-layer fully connected neural network: the input layer has 7 nodes, the first hidden layer has 16 nodes with a rectified linear unit activation function, the second hidden layer has 8 nodes with a hyperbolic tangent activation function, and the output layer has 3 nodes with no activation function, outputting 3 latent variables. .

[0038] The decoder structure is symmetrical to the encoder. The first hidden layer has 8 nodes, the second hidden layer has 16 nodes, and the output layer has 7 nodes, used to reconstruct the original 7-dimensional input vector. An independent component constraint layer is embedded between the encoder output and the decoder input, and its objective function is... To maximize the non-Gaussianity of the latent variables, the negative entropy approximation is used as a measure, with the specific expression as follows: ; , The first output of the encoder One latent variable, It is a constant, taking the value 1. For standard Gaussian random variables, For mathematical expectation, The negative entropy approximation function is used. Model training employs a manually prepared standard seawater sample dataset containing one hundred different combinations, covering concentration gradients of free residual chlorine (0–0.2 mg / L), monochloramine (0–1.5 mg / L), dichloramine (0–0.8 mg / L), nitrogen trichloride (0–0.5 mg / L), bromide ions (0–50 mg / L), and hypobromic acid (0–0.6 mg / L). Label data represents the true concentrations of each component, calibrated using a combination of standard iodometric titration and ion chromatography. The training loss function is... The weighted sum of reconstruction error and independent component constraint terms: ; For the original input, To reconstruct the output, The value is 0.9. The value is 0.1, representing the first value output by the encoder. There are 600 latent variables. After training, the model is deployed in an embedded processor. During inference, 600 7-dimensional vectors collected in real time are input sequentially, and after being filtered by moving average, the corresponding 3 independent component sequences are output.

[0039] Then, step S6 is executed: the concentration of free residual chlorine in the seawater sample is calculated based on the characteristic components, and the monitoring results are output. During the model inference phase, the system acquires three independent components. , , The time series was analyzed. Independent components corresponding to free residual chlorine were identified using a pre-established component-substance mapping table. This mapping table was established through offline calibration experiments: single standard substances (e.g., only free residual chlorine) were added incrementally to chlorine-free artificial seawater, and the changes in the response intensity of each independent component were recorded.

[0040] Experiments show that, The composition of the substance increases monotonically with increasing free residual chlorine concentration, with a correlation coefficient greater than 0.99, while the absolute value of the response slope to other substances is less than 0.05. Therefore, This is labeled as a "residual chlorine-specific component." The value of this residual chlorine-specific component was converted to a concentration value after linear regression calibration; the calibration equation is as follows. , This is the sensitivity coefficient. The intercept was obtained by fitting the calibration data using the least squares method. Final concentration value. This refers to the monitoring results of free residual chlorine.

[0041] Furthermore, the method includes a dynamic environmental compensation step. Seawater salinity, turbidity, and organic background can interfere with the sensing signal.

[0042] To address this, a dynamic environmental compensation model is integrated into the system. The input variables for this dynamic environmental compensation model include seawater temperature. Electrical conductivity Dissolved oxygen and total organic carbon . The temperature is measured by a platinum resistance temperature sensor integrated into the reaction chamber, with an accuracy of ±0.1 degrees. The conductivity was measured using a four-electrode conductivity cell with an electrode spacing of 5 mm and an excitation frequency of 1 kHz. Measured by a fluorescence quenching probe and calculated based on the phase difference method; absorbance at 254 nanometers Empirical formula Estimate, This is the regional correction factor, with an initial value of 2.5.

[0043] The compensation model is a multilayer sensor, and the output is a component specific to residual chlorine. Correction coefficient The final compensated residual chlorine characteristic value is Substitute the values ​​into the calibration equation to calculate the final concentration.

[0044] The method further includes an adaptive signal decoupling algorithm. The system continuously monitors the rate of change of residual chlorine concentration and the stability of environmental parameters. When the concentration change is less than 5% within 30 consecutive minutes, and , , When the fluctuations are all less than their respective thresholds (0.5 degrees Celsius, 50 microsiemens per centimeter, and 0.2 milligrams per liter, respectively), the model fine-tuning mode is activated.

[0045] In model fine-tuning mode, the system controls a micro-injection pump to inject 0.1 ml of a 1 mg / L free residual chlorine standard solution into the current seawater sample, creating a perturbed sample. Multi-source heterogeneous data from the perturbed sample are collected and input into the current feature decoupling model to obtain the predicted concentration. Actual added concentration Precise calculations based on injection volume and flow rate. If When the set value is 0.02 mg / L, the backpropagation algorithm is triggered, and the model weights are updated step by step with a learning rate of 0.001. After the update, the prediction error of the perturbation sample is verified. If it still exceeds the limit, the fine-tuning is repeated, up to 3 times; otherwise, the system switches back to normal monitoring mode.

[0046] Monitoring results are uploaded to the central monitoring platform in real time via an industrial Ethernet interface or RS485 bus. The system has a built-in audible and visual alarm device. When the residual chlorine concentration is greater than or equal to the preset safety threshold of 0.1 mg / L, the alarm is triggered and an event log is recorded. The log includes a timestamp, the original 7-dimensional vector mean, and... value, Value, final concentration , , , , Alarm status and other information are stored in non-volatile memory and can be retrieved remotely.

[0047] The implementation of the above method relies on a complete monitoring system. This system includes a multi-channel sensing reaction chamber, a multimodal electrochemical sensing array, a miniature ultraviolet-visible absorption spectroscopy module, an environmental parameter sensing module, a main control processing unit, and a communication and alarm module.

[0048] The multi-channel sensing reaction chamber is made of polytetrafluoroethylene, which is highly corrosion resistant. The internal flow channels are precision injection molded to ensure that the flow uniformity error is less than 3%.

[0049] The three units of the multimodal electrochemical sensing array are packaged on the same ceramic substrate, and the electrode leads are connected to the printed circuit board by gold wire bonding. The shielding layer is wrapped to suppress electromagnetic interference.

[0050] The miniature ultraviolet-visible absorption spectroscopy module adopts a modular design, with the light source and detector fixed in an aluminum alloy heat dissipation frame. The thermal expansion coefficients are matched to avoid optical path deviation.

[0051] The environmental parameter sensing module is integrated into the side wall of the reaction chamber, and each sensor probe is in direct contact with the flowing seawater, with a response time of less than 10 seconds.

[0052] The main control processing unit uses an ARM Cortex-A53 quad-core processor, runs an embedded Linux system, and is equipped with a neural network inference engine, supporting floating-point operation acceleration. The communication module supports Modbus TCP and Modbus RTU protocols, and the alarm module includes a red LED light and a buzzer with a sound pressure level greater than 80 decibels.

[0053] The entire system needs to be calibrated throughout the entire process before deployment.

[0054] First, a series of free residual chlorine solutions with varying concentrations were prepared using standard seawater. These solutions were then flowed through the system sequentially, and the outputs of each module were recorded to establish an initial calibration curve.

[0055] Then, interference solutions containing chloramine and bromide were introduced respectively to verify the specificity of the second sensing unit and the interference response capability of the third sensing unit.

[0056] Further adjustments were made to seawater temperature, salinity, and organic matter content to test the effectiveness of the dynamic environmental compensation model.

[0057] Finally, an adaptive fine-tuning process was executed to confirm that the model update mechanism was functioning correctly. During daily operation, the system automatically performs a zero-point calibration every 24 hours, flushing the flow path with chlorine-free artificial seawater and collecting baseline signals for drift compensation.

[0058] In summary, this embodiment constructs a highly specific and robust online monitoring scheme for residual chlorine in seawater through four major technical pillars: multimodal sensor fusion, deep feature decoupling, dynamic environmental compensation, and online adaptive learning. It effectively solves the technical problem of distinguishing residual chlorine in seawater from similar substances such as chloramines and bromides, and achieves accurate, continuous, and maintenance-free monitoring of free residual chlorine.

[0059] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing, characterized in that, include: Seawater samples are collected and flow through a multi-channel sensing reaction chamber; Multiple electrochemical response signals are acquired simultaneously using a multimodal electrochemical sensing array; The multimodal electrochemical sensing array includes a first sensing unit, a second sensing unit, and a third sensing unit; The first sensing unit uses a platinum-plated working electrode and is modified with a nano-titanium dioxide catalytic layer to produce a broad-spectrum response to chlorine-containing oxidizing substances. The second sensing unit uses a gold-plated working electrode and is fixed with a molecularly imprinted polymer film with free residual chlorine as a template molecule to specifically identify free residual chlorine. The third sensing unit employs a glassy carbon working electrode and is coated with a polyaniline conductive polymer layer to respond to bromide and chloramine interferences. The absorbance values ​​of seawater samples at different characteristic wavelengths were obtained using a miniature ultraviolet-visible absorption spectroscopy module. The electrochemical response signal and the absorbance value are spatiotemporally aligned to form a multi-source heterogeneous sensing dataset. The multi-source heterogeneous sensing dataset is subjected to component separation processing based on a preset feature decoupling model, and feature components related only to free residual chlorine are extracted. The feature decoupling model is a neural network model based on a joint architecture of deep autoencoder and independent component analysis, which outputs multiple statistically independent latent variables. The concentration of free residual chlorine in the seawater sample is calculated based on the characteristic components, and the monitoring results are output.

2. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 1, characterized in that, The multi-channel sensing reaction chamber is equipped with a microfluidic structure, including a main sample inlet channel, a diversion channel, a mixing chamber, and a waste liquid outlet; The main sampling channel is connected to a seawater sampling pump to introduce the seawater sample to be tested. The diversion channel divides the seawater sample into three paths, which are respectively introduced into the detection chambers corresponding to the first sensing unit, the second sensing unit, and the third sensing unit. Each detection chamber is equipped with a constant temperature heating plate at the bottom to maintain the reaction temperature at 25 degrees Celsius; each detection chamber is equipped with a miniature stirring magnet at the top, driven by an external rotating magnetic field to ensure that the sample makes uniform contact with the sensing interface.

3. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 2, characterized in that, The miniature ultraviolet-visible absorption spectroscopy module includes a deuterium lamp light source, a quartz colorimeter, a grating spectrometer, and a linear photodiode array detector; The quartz colorimeter is integrated into the bypass channel of the multi-channel sensing reaction chamber; The grating beam splitter disperses the transmitted light according to wavelength and projects it onto the linear photodiode array detector.

4. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 3, characterized in that, The spatiotemporal alignment of the multi-source heterogeneous sensor dataset is achieved through a hardware-level synchronization triggering mechanism; The multimodal electrochemical sensing array and the miniature ultraviolet-visible absorption spectroscopy module share the same clock source; At each sampling moment, the current response values ​​of multiple electrochemical sensing units and the absorbance values ​​of multiple characteristic wavelengths are recorded simultaneously to form a multidimensional vector; Continuously collected data forms a time window, which serves as the input to the feature decoupling model.

5. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 4, characterized in that, The feature decoupling model includes an encoder, a decoder, and an independent component constraint layer; The encoder consists of a three-layer fully connected neural network; The decoder structure is symmetrical to the encoder; the independent component constraint layer is embedded between the last layer of the encoder and the first layer of the decoder, forcing the multiple latent variables of the output to be statistically independent of each other; The model was trained using a standard seawater sample dataset prepared artificially, with the label data being the actual concentrations of free residual chlorine, monochloramine, dichloramine, nitrogen trichloride, bromide ions, and hypobromic acid.

6. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 5, characterized in that, During the model inference stage, real-time collected multi-source heterogeneous sensor data is input into the trained feature decoupling model to obtain multiple independent components; The independent components corresponding to free residual chlorine were determined using the component-substance mapping table; The component-substance mapping table was established through offline calibration experiments. The specific method is as follows: A single standard substance was added to chlorine-free artificial seawater in incremental amounts, and the trend of the response intensity of each independent component was recorded. Components whose response intensity monotonically increased with the concentration of free residual chlorine and had a weak response to other substances were marked as residual chlorine-specific components.

7. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 6, characterized in that, The monitoring method also includes a dynamic environmental compensation step; The dynamic environmental compensation model is used to correct the interference of seawater salinity, turbidity and organic matter background on the monitoring results; The input variables of the dynamic environmental compensation model include seawater temperature, conductivity, dissolved oxygen concentration and total organic carbon content, and the output is the correction coefficient for residual chlorine-specific components. The seawater temperature was measured by a platinum resistance temperature sensor. The conductivity was measured using a four-electrode conductivity cell; the dissolved oxygen concentration was measured using a fluorescence quenching dissolved oxygen probe. The total organic carbon content was calculated using an empirical formula based on absorbance values.

8. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 7, characterized in that, The monitoring method also includes an adaptive signal decoupling algorithm; When the system detects that the change in residual chlorine concentration is less than the specified value over a continuous period of time and the environmental parameters are stable, the model fine-tuning mode is activated. In model fine-tuning mode, the system collects current seawater samples and injects a small amount of free residual chlorine standard solution of known concentration to form a perturbed sample; Input the multi-source heterogeneous sensing data of the perturbed sample into the current model to calculate the deviation between the predicted concentration and the actual added concentration; If the absolute value of the deviation is greater than the set value, the backpropagation algorithm is triggered to update the weights of the feature decoupling model in small steps.

9. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 8, characterized in that, The monitoring results are uploaded to the central monitoring platform in real time via an industrial Ethernet interface or RS485 bus. When the residual chlorine concentration is detected to be greater than or equal to the preset safety threshold, an audible and visual alarm is triggered and the event log is recorded. The event log includes timestamps, raw sensor data, decoupled feature values, compensated concentration values, and environmental parameters.

10. The method for online monitoring of residual chlorine in seawater samples combined with intelligent sensing according to claim 9, characterized in that, The nano-titanium dioxide catalyst layer was synthesized by the sol-gel method.

Images