Microbial electrode online monitoring system and method based on multi-modal fusion and self-repairing mechanism

Abstract

The invention relates to a microbial electrode online monitoring system and method based on multi-modal fusion and a self-repairing mechanism, and belongs to the technical field of environmental monitoring and bioelectrochemistry. Aiming at the problems of high cost, weak anti-interference capability, poor long-term stability, difficulty in precise decoupling of multi-parameter coupling and the like in the existing microbial electrode monitoring technology, the method comprises the following steps: synchronously acquiring multi-modal signals of a dual-wavelength differential near infrared spectrum and an electrochemical impedance spectrum; a front decoupling mechanism based on equivalent circuit fitting and reinforcement learning is combined to clearly distinguish change sources of charge transfer resistance and diffusion resistance; performing deep fusion on the decoupled features by using a convolutional neural network, and outputting a microbial activity score and a metabolite concentration; and a self-repairing electrode material and a closed-loop feedback control mechanism are introduced, and repairing is automatically triggered when electrode performance degradation is monitored. According to the invention, low-cost, high-precision and strong-anti-interference long-term stable in-situ monitoring of the microbial electrode is realized.

Description

Technical Field

[0001] This invention belongs to the field of environmental monitoring and bioelectrochemical technology, and relates to an online monitoring system and method for microbial electrodes based on multimodal fusion and self-healing mechanisms. Background Technology

[0002] The impedance characteristics of microbial electrodes are key indicators reflecting microbial metabolic activity, extracellular electron transport efficiency, and biofilm state. On the other hand, near-infrared spectroscopy can non-destructively and in real-time monitor microbial metabolites, such as extracellular polymers and glucose. Therefore, combining electrochemical impedance spectroscopy with near-infrared spectroscopy to create a multi-dimensional monitoring method has become an important direction for accurately assessing microbial state.

[0003] However, existing technologies have many limitations that hinder their practical application. First, the systems are costly and difficult to integrate. Traditional multimodal analysis often relies on Raman spectroscopy, which requires a separate and precise optical platform, resulting in high overall costs and making it difficult to achieve system miniaturization and integration, thus unsuitable for in-situ deployment. Second, existing technologies generally have weak anti-interference capabilities. In complex real-world environments with high humidity and high salinity, the strong absorption effect of water molecules and power frequency interference significantly increase impedance measurement errors, while near-infrared spectral signals are also limited in application due to strong water interference. Third, the systems lack long-term stability. The impedance signal of microbial electrodes can drift significantly due to continuous biofilm coverage or electrochemical corrosion, and existing technologies lack effective self-repair mechanisms, making long-term stable in-situ monitoring difficult. Fourth, multi-parameter coupling problems are severe. In traditional impedance analysis methods, such as analysis using Nyquist plots, it is impossible to accurately distinguish the root cause of the decrease in charge transfer resistance—whether it stems from increased microbial activity indicated by increased extracellular polymers or from improved mass transfer due to thinning of the biofilm. Neither impedance spectroscopy nor near-infrared spectroscopy alone can meet the requirements for precise multi-parameter analysis and decoupling of microbial states.

[0004] Therefore, there is an urgent need in this field for a low-cost, highly integrated, anti-interference-capable, and long-term stable microbial electrode online monitoring solution that can accurately decouple key parameters. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide an online monitoring system and method for microbial electrodes based on multimodal fusion and self-healing mechanisms, so as to solve the problems of high cost, weak anti-interference ability, poor long-term stability and difficulty in accurately decoupling multi-parameter coupling in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An online monitoring system for microbial electrodes based on multimodal fusion and self-healing mechanisms includes: Near-infrared spectroscopy module: used to acquire near-infrared spectral data of microbial electrodes; The three-electrode system includes a working electrode, a counter electrode, and a reference electrode, which are used in conjunction with an external electrochemical impedance analysis module to obtain electrochemical impedance spectroscopy data. The working electrode is made of a self-healing electrode material. Multi-parameter decoupling module: used to receive the electrochemical impedance spectroscopy data and the near-infrared spectroscopy data, and clearly distinguish and output the decoupled charge transfer resistance and diffusion resistance through a pre-decoupling mechanism; Data fusion module: used to fuse the decoupled impedance characteristics and near-infrared spectral characteristics, and output microbial activity scores and metabolite concentrations; Self-healing feedback module: Based on the impedance data drift monitored by the electrochemical impedance analysis module, when the impedance data drift exceeds a preset threshold, a closed-loop repair action is triggered on the working electrode.

[0007] Furthermore, the near-infrared spectroscopy module includes a miniature fiber optic spectrometer with a working wavelength range of 900~1700nm, and uses a dual-wavelength differential absorption method with 970nm and 1450nm as characteristic wavelengths to suppress water interference.

[0008] Furthermore, the multi-parameter decoupling module includes an equivalent circuit fitting unit and a reinforcement learning optimization unit, used for: The electrochemical impedance spectroscopy data are fitted with an equivalent circuit to obtain the initial charge transfer resistance and diffusion resistance; Verification was performed by combining the intensity of specific peaks in the near-infrared spectral data; Reinforcement learning algorithms are used to dynamically optimize fitting parameters in order to achieve parameter decoupling.

[0009] Furthermore, the state space of the reinforcement learning algorithm includes the impedance phase angle change rate, the specific peak intensity, and the signal-to-noise ratio, and its reward function aims to maximize the decoupling accuracy of the charge transfer resistor and the diffusion resistor.

[0010] Furthermore, the working electrode in the three-electrode system is composed of a carbon paper surface-modified self-healing hydrogel, which contains a polyvinyl alcohol-chitosan matrix and carbon nanotubes dispersed therein.

[0011] Furthermore, the self-healing feedback module includes an impedance monitoring unit, a microfluidic pump, and a repair fluid reservoir; the impedance monitoring unit monitors the impedance data of the external electrochemical impedance analysis module at 1kHz, 10kHz, and 100kHz in real time, and when the impedance drift at any frequency point exceeds 5% for three consecutive monitoring, the microfluidic pump is triggered to inject the repair fluid.

[0012] Furthermore, the data fusion module is implemented based on a convolutional neural network (CNN) and deployed on a heterogeneous system-on-chip (SoC). The input of the CNN is the normalized decoupled impedance characteristics and the near-infrared spectral characteristics, and the output is the microbial activity score and the metabolite concentration.

[0013] A method for online monitoring of microbial electrodes based on multimodal fusion and self-healing mechanisms includes the following steps: S1: Pretreatment stage, sterilize the working electrode and inoculate it with target microorganisms, and initialize the parameters for dual-modal acquisition and analysis; S2: Dual-modal signal acquisition stage, simultaneously acquiring electrochemical impedance spectroscopy data and near-infrared spectroscopy data of the microbial electrode; S3: Multi-parameter decoupling stage, through the pre-decoupling steps of equivalent circuit fitting, near-infrared spectroscopy verification and reinforcement learning optimization, outputs the decoupled charge transfer resistance and diffusion resistance; S4: Self-repair feedback stage, real-time monitoring of impedance data drift at key frequency points, and triggering repair action on the working electrode when the impedance drift exceeds a preset threshold. S5: Data fusion and output stage, deeply fuses the decoupled impedance characteristics and near-infrared spectral characteristics, and calculates and outputs microbial activity scores and metabolite concentrations through a deep learning model.

[0014] Furthermore, in step S2, the near-infrared spectral data acquisition is synchronized with the electrochemical impedance spectroscopy data acquisition, specifically, one near-infrared spectrum is acquired for every 10 impedance frequency points.

[0015] Furthermore, in step S5, the formula used to normalize the impedance characteristics is as follows:

[0016] in, For frequency Complex impedance at that point, These are the minimum and maximum values ​​of the complex impedance at all frequencies, respectively. The formula for performing Standard Normal Transform (SNV) processing on the near-infrared spectral data is as follows:

[0017] in, For the original spectrum at wavelength i The absorbance value at that point This represents the average absorbance across all wavelengths of the spectrum. The standard deviation is denoted as .

[0018] The beneficial effects of this invention are as follows: (1) The near-infrared spectroscopy module is used to replace the traditional Raman spectroscopy module, and the heterogeneous system chip is used to achieve a high degree of integration of decoupling, fusion and feedback control functions, which greatly reduces the hardware cost of the whole system and significantly improves the integration, making it very suitable for on-site deployment.

[0019] (2) By using dual-wavelength differential and advanced spectral processing algorithms, the interference of water molecules on the near-infrared spectrum is effectively suppressed; at the same time, the specially designed three-electrode system ensures the stability of the electrochemical impedance signal in complex environments such as high salt content, and controls the measurement error caused by external interference to a low level.

[0020] (3) Through the pre-decoupling mechanism that combines equivalent circuit fitting, near-infrared spectroscopy verification and reinforcement learning optimization, the root causes of changes in charge transfer resistance and diffusion resistance can be clearly distinguished. Combined with the deep fusion of multimodal data from convolutional neural networks, the accuracy of assessing microbial activity and metabolic state is greatly improved.

[0021] (4) The innovative self-healing electrode material and closed-loop feedback control mechanism can quickly trigger and complete the repair action when the electrode performance decay is detected, effectively suppressing the signal drift caused by biofilm coverage or electrode corrosion, extending the service life of the electrode, and reducing the complexity and cost of long-term operation and maintenance.

[0022] (5) Based on the hardware acceleration design of heterogeneous chips, the data preprocessing and inference speed of the convolutional neural network model are optimized, so that the system data update rate can match the conventional electrochemical impedance scanning rate, and the real-time tracking and feedback of the dynamic changes of microbial metabolism are realized.

[0023] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a block diagram of the device structure; Figure 2 This is a correlation diagram between impedance spectroscopy and near-infrared spectroscopy. Figure 3 This is a comparison of impedance drift between self-healing electrodes and traditional electrodes. Detailed Implementation

[0025] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0027] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0028] I. Technical Solution (I) Components of the apparatus The overall structure of the device includes a near-infrared spectroscopy module, a three-electrode system, a multi-parameter decoupling module, a self-healing feedback module, and a data fusion module. It interfaces with an external electrochemical impedance spectroscopy module, and all modules work collaboratively. The specific structure is as follows: Figure 1 As shown.

[0029] Near-infrared spectroscopy module: Employs a miniature fiber optic spectrometer (900~1700nm), which is less expensive than traditional Raman spectrometers; eliminates 80% of water interference through dual-wavelength differential absorption (970nm is a strong absorption peak for water, and 1450nm is a sensitive peak for microbial metabolites), and improves the characteristic peak recognition rate to 90% by combining it with the OSC algorithm (based on partial least squares regression PLSR to separate the spectra of water and target substances); the fiber optic probe uses a sapphire window, which can withstand high-salt environments of 0~30g / L NaCl, and has a transmittance of >90%, making it suitable for complex detection scenarios (such as industrial wastewater, MFC anode chamber).

[0030] Electrochemical impedance analysis module: Self-developed high-performance broadband impedance analysis module, continuously adjustable from 100Hz to 100kHz, with a minimum frequency step of 0.1Hz, an amplitude range of 5~10mV peak-to-peak value, and an amplitude accuracy of ≤0.1mV, ensuring a microbial survival rate of >98%.

[0031] Three-electrode system: Working electrode: Carbon paper surface modified with self-healing hydrogel (polyvinyl alcohol-chitosan + 0.5wt% carbon nanotubes), the imine bond cross-linking structure gives it self-healing ability (repair rate >90% after 30s after breakage), carbon nanotubes improve conductivity (resistance <50Ω), surface roughness Ra=8~12μm, which is conducive to biofilm adhesion and does not affect electron transfer. Counter electrode: A platinum ring (2mm in diameter) surrounds the working electrode to ensure uniform current distribution and avoid impedance measurement deviation caused by local current concentration, adapting to the current acquisition requirements of external electrochemical workstations or electrochemical impedance analysis modules. All-solid-state reference electrode: Ag / silver tetraphenylborate-polyvinyl chloride film structure, which is stable in high-salt environment (drift <8mV / month) and has a response time <2s, avoiding the leakage problem of liquid reference electrode and ensuring that the electrochemical impedance analysis module obtains a stable potential reference.

[0032] Multi-parameter decoupling module: The core function is to distinguish between Rct (reflecting electron transfer efficiency) and Rd (reflecting mass transfer efficiency) to solve the parameter coupling problem. Initial Rct and Rd are obtained by fitting the impedance spectrum output from the external electrochemical impedance analysis module using an equivalent circuit. The physical meaning of the parameters is verified by combining the near-infrared 1450nm peak intensity (directly related to EPS content) – if Rct decreases and the 1450nm peak intensity increases, it indicates enhanced microbial activity (increased EPS); if Rct decreases and the 1450nm peak intensity decreases, it indicates thinner biofilm (improved mass transfer). Then, the equivalent circuit fitting parameters are dynamically optimized through reinforcement learning to ensure decoupling accuracy (error <3%). The state space of the reinforcement learning includes the impedance phase angle change rate (reflecting impedance signal stability, data sourced from the external electrochemical impedance analysis module), the 1450nm peak intensity (reflecting metabolic activity), and the signal-to-noise ratio (reflecting signal quality). The reward function aims to maximize the decoupling accuracy of Rct and Rd, achieving adaptive optimization.

[0033] ⑤ Self-healing feedback module: It consists of a microfluidic pump, a repair solution reservoir (0.1wt% chitosan-polyvinyl alcohol prepolymer solution), and an impedance monitoring unit. The monitoring unit collects the impedance at three key frequency points (1kHz, 10kHz, and 100kHz) output by an external electrochemical impedance analysis module in real time (1kHz reflects the charge transfer at the electrode interface, and 10kHz and 100kHz reflect the mass transfer process of the solution). When the drift at any frequency point is >5% for three consecutive times, the microfluidic pump is triggered to inject 10μL of repair solution. The prepolymer solution repairs biofilm defects or electrode corrosion on the electrode surface through dynamic cross-linking of imine bonds, with a repair response time of <30s. After repair, the impedance recovery rate is verified to be >95% by the external electrochemical impedance analysis module. During 30 days of long-term monitoring, the impedance drift is stable at <5%, which is far superior to traditional electrodes (>25%), and there is no need to frequently replace the electrodes.

[0034] ⑥ Data fusion module: Multimodal feature deep fusion is achieved based on CNN, deployed on the ZYNQ7010 heterogeneous SoC to balance real-time performance and integration. A CNN convolution kernel accelerator is designed at the PL end, reducing inference time from 800ms in pure software to 75ms. Input features are decoupled impedance features (Rct / Rd, real and imaginary parts of impedance, approximately 400 features, data sourced from the electrochemical impedance analysis module) and filtered near-infrared spectral features (approximately 800 wavelength points), which are concatenated by channel and input into the CNN. The CNN passes through 3... Convolutional layers extract local features (kernel sizes 5×1, 3×1, 3×1, adapted to one-dimensional spectral and impedance features), three max-pooling layers reduce feature dimensionality (pooling kernel 2×1, retaining key information), and two fully connected layers integrate cross-modal features. The final output is a microbial activity score (0~100) and metabolite concentration. The activity score is calculated based on Rct (electron transfer efficiency) and the 1450nm peak intensity (metabolic activity). A score of 85~90 corresponds to a stable MFC power generation state (power density >12mW / m³). 2 (This can be verified with the assistance of an external electronic load tester).

[0035] (II) Analytical Methods and Steps Pretreatment stage: The working electrode is sterilized by ultraviolet light for 30 min (to eliminate contamination and avoid interference with the detection of target microorganisms). After inoculating with target microorganisms (such as Shewanella oneidensis MR-1 on the MFC anode and degrading bacteria in environmental water), it is placed in a 30℃ incubator. After 48 h, the biofilm coverage is verified to be >80% by fluorescence staining (such as SYTO9 staining). The device parameters (near-infrared spectral scanning step size 10 nm / step, spectral acquisition interval 10 impedance frequency points) and the external electrochemical impedance analysis module parameters (voltage amplitude 10 mV peak-to-peak value, frequency scanning range 100 Hz~100 kHz, sampling points 200 points / frame) are initialized to ensure that the data acquisition of the two is synchronized.

[0036] Dual-modal signal acquisition stage: The external electrochemical impedance analysis module outputs a sinusoidal voltage according to the set parameters, and acquires the impedance spectrum (200 points / frame) through a three-electrode system. The real part, imaginary part and phase angle data of the impedance are transmitted to the multi-parameter decoupling module in real time. The near-infrared spectroscopy module starts synchronously with the impedance acquisition. Every 10 impedance frequency points are acquired, corresponding to one near-infrared spectrum acquisition. The spectral data are first subjected to 970nm / 1450nm dual-wavelength differential to eliminate water interference, and then the characteristic wavelengths are screened by standard normal variable transformation (SNV) (to eliminate baseline drift and scattering effects) and partial least squares regression (PLSR) (to retain peaks related to metabolites such as 1450nm) to obtain preprocessed near-infrared characteristic data.

[0037] Multi-parameter decoupling stage: The multi-parameter decoupling module receives impedance spectrum data from the external electrochemical impedance analysis module, fits it using an equivalent circuit, and outputs candidate values ​​for Rct and Rd. Combining the pre-processed near-infrared 1450nm peak intensity, it determines the physical meaning of the Rct change (enhanced activity or improved mass transfer). The reinforcement learning algorithm is then activated, using the impedance phase angle change rate, 1450nm peak intensity, and signal-to-noise ratio as state variables to dynamically adjust the equivalent circuit fitting parameters (such as the initial fitted value of charge transfer resistance and the time constant of diffusion resistance) to maximize the decoupling accuracy of Rct and Rd, and finally outputs accurate Rct and Rd values ​​(error <3%).

[0038] Self-repair feedback phase: The self-repair feedback module receives impedance data at 1kHz, 10kHz, and 100kHz monitored by the electrochemical impedance analysis module in real time. If the impedance drift at any frequency point is >5% for three consecutive monitoring (e.g., the impedance at 1kHz increases from 300Ω to 315Ω), the microfluidic pump is immediately triggered to inject 10μL of repair solution. After the repair is completed, the impedance data is re-acquired by the electrochemical impedance analysis module to verify that the impedance recovery rate is >95%. If the standard is not met, the injection is repeated (up to 3 times to avoid excessive repair solution affecting microbial activity).

[0039] Data fusion and output stage: The decoupled impedance characteristics (Rct / Rd, real part + imaginary part of impedance) are normalized (eliminating dimensional differences, such as converting impedance from Ω to the 0~1 range), and then spliced ​​with the preprocessed near-infrared feature data; the spliced ​​feature data is input into the CNN model, and the model outputs the microbial activity score (accuracy >92%, percentage of samples within ±5 error) and metabolite concentration (e.g., glucose 0~20mmol / L, mean square error (MSE) <1.2mmol / L); the data fusion module transmits the results to the host computer through standard interfaces (e.g., RS485, Ethernet), and simultaneously displays the activity score, impedance spectrum, near-infrared spectrum and metabolic concentration curve in real time on a 10.1-inch touch screen to meet the needs of on-site monitoring.

[0040] (1) Data preprocessing Impedance spectrum preprocessing: The broadband impedance spectrum is normalized to eliminate the influence of dimensions. The normalization formula is:

[0041] in, For frequency Complex impedance at that point, These represent the minimum and maximum values ​​of the complex impedance at all frequencies. Simultaneously, key features of the impedance spectrum, such as charge transfer resistance, are extracted. (Obtained through equivalent circuit fitting), phase angle wait.

[0042] Near-infrared spectral preprocessing: Standard normal variable transformation (SNV) is used to eliminate baseline drift and scattering effects in the spectrum. The formula is as follows:

[0043] in, For the original spectrum at wavelength i The absorbance value at that point This represents the average absorbance across all wavelengths of the spectrum. The standard deviation is then used. Partial least squares regression (PLSR) is then used to screen for characteristic wavelengths related to microbial metabolism (such as the characteristic peak at 1450 nm associated with extracellular polymeric substances (EPS)).

[0044] (2) CNN model structure and training Model structure: Input layer: The input consists of preprocessed impedance spectrum (Rct / Rd, real part of impedance + imaginary part, about 400 features in total) and near-infrared spectrum (about 800 wavelength points). The two are concatenated by channel and used as input. The input dimension is (1, 1200) (assuming the number of impedance spectrum features is 400 and the number of near-infrared spectrum wavelength points is 800).

[0045] Convolutional layers: There are 3 convolutional layers with kernel sizes of 5×1, 3×1 and 3×1 respectively. Each convolutional layer is followed by batch normalization and ReLU activation function to extract local features of bimodal data.

[0046] Pooling layer: A max pooling layer is added after each convolutional layer. The pooling kernel size is 2×1, which is used to reduce the feature dimension and retain key information.

[0047] Fully connected layer: Contains two fully connected layers with 256 and 128 nodes respectively, and finally connects to the output layer.

[0048] Output layer: The output consists of microbial activity scores (0~100) and metabolite concentrations (such as glucose concentration, ranging from 0~20 mmol / L), so the output dimension is 2.

[0049] Model training: Dataset: Collect dual-modal data (impedance spectroscopy and near-infrared spectroscopy) of different microbial growth stages and metabolic states, and label them with microbial activity scores (determined by traditional biochemical methods, such as the MTT assay for cell activity) and metabolite concentrations (determined by high-performance liquid chromatography (HPLC) and other methods). Construct training, validation and test sets, with the training set accounting for 70%, the validation set accounting for 15% and the test set accounting for 15%.

[0050] Loss function: For microbial activity scores (regression problem), mean squared error (MSE) loss is used; for metabolite concentrations (regression problem), MSE loss is also used. The total loss function is a weighted sum of the two:

[0051] in, and This is a weighting coefficient, adjusted according to the importance of the task; here it is all set to 0.5.

[0052] Optimizer: The Adam optimizer is used, with the learning rate initialized to 0.001. During training, a learning rate decay strategy is adopted, with the learning rate multiplied by 0.9 every 10 epochs.

[0053] Training process: The number of training epochs is 100, and the batch size is 32. During training, the model performance is monitored using a validation set. Training is stopped when the validation set loss no longer decreases to prevent overfitting.

[0054] Low cost and high integration: Near-infrared spectroscopy replaces Raman spectroscopy, reducing hardware costs by more than 70%; ZYNQ7010 heterogeneous SoC achieves integrated "decoupling-fusion-feedback" and is suitable for on-site monitoring (such as MFC stack and environmental water monitoring stations).

[0055] Strong anti-interference capability: The dual-wavelength differential + orthogonal signal correction (OSC) algorithm eliminates 90% of water interference, and the sapphire window of the near-infrared probe is resistant to high-salt environments; the platinum ring counter electrode and the all-solid-state reference electrode of the three-electrode system ensure that the external electrochemical impedance analysis module obtains a stable impedance signal, and the measurement error caused by power frequency interference is controlled to <8% (better than the >15% of traditional non-anti-interference designs), meeting the detection needs of complex scenarios.

[0056] Precise parameter decoupling: Through the pre-decoupling mechanism of "equivalent circuit fitting + near-infrared verification + reinforcement learning", the problem that traditional Nyquist plots cannot distinguish the cause of Rct changes is completely solved; combined with CNN multimodal fusion, the accuracy of microbial activity assessment is >92%, which is far better than single impedance method (<75%) or single near-infrared method (<80%), and can accurately determine the metabolic state of microorganisms (such as enhanced activity, mass transfer limitation, electrode contamination).

[0057] Excellent long-term stability: self-healing electrode + closed-loop feedback mechanism, impedance drift <5% after 30 days, repair response time <30s, avoiding the frequent replacement of traditional electrodes due to excessive biofilm coverage or corrosion (traditional electrode life is usually <15 days), reducing the operation and maintenance costs of long-term monitoring.

[0058] Real-time adaptation: CNN inference time 75ms, data update rate 10Hz, matching the conventional scan rate (1~10Hz) of the external electrochemical impedance analysis module, can track the dynamic changes of microbial metabolism in real time (such as the activity enhancement during the MFC initiation period and the activity fluctuation during the degradation of environmental pollutants).

[0059] Example 1: Implementation of System Architecture and Decoupling Method Based on Heterogeneous System-on-Chips This embodiment describes in detail the hardware architecture, signal acquisition, and core multi-parameter decoupling method of the online monitoring system of the present invention.

[0060] I. System Hardware Architecture and Module Description like Figure 1 The device structure diagram shown indicates that this system includes a near-infrared spectroscopy module, a three-electrode system, a multi-parameter decoupling module, a data fusion module, a self-healing feedback module, and an interface for interfacing with an external electrochemical impedance analysis module.

[0061] Near-infrared spectroscopy module: Employs a miniature fiber optic spectrometer with an operating wavelength range of 900nm to 1700nm, offering lower cost than traditional Raman spectrometers. This module utilizes a dual-wavelength differential absorption method, selecting 970nm and 1450nm as characteristic wavelengths for water interference suppression. The fiber optic probe features a sapphire window, providing resistance to high-salt environments.

[0062] The three-electrode system includes a working electrode, a platinum ring counter electrode, and an all-solid-state reference electrode.

[0063] Working electrode: Constructed of a self-healing hydrogel modified with carbon paper surface, the hydrogel comprising a polyvinyl alcohol-chitosan matrix and carbon nanotubes dispersed therein. The imine bond cross-linking structure endows it with self-healing ability, while the carbon nanotubes enhance conductivity.

[0064] All-solid-state reference electrode: It adopts an Ag / silver tetraphenylborate-polyvinyl chloride film structure, which can provide a stable potential reference in high-salt environments.

[0065] External electrochemical impedance spectroscopy module: used to output sinusoidal voltage signals and obtain electrochemical impedance spectroscopy data through a three-electrode system.

[0066] The main control core, including the control and operation cores of the multi-parameter decoupling module, data fusion module, and self-healing feedback module, is deployed on the Xilinx ZYNQ7010 heterogeneous system chip.

[0067] PS side: Runs a Linux system and is responsible for the deployment of system communication, human-computer interaction, and data fusion models.

[0068] PL side: Design a one-dimensional convolutional parallel architecture as a convolution kernel accelerator for deep learning models, reducing inference time from 800 milliseconds in pure software to 75 milliseconds.

[0069] II. Workflow: Dual-modal signal acquisition and multi-parameter decoupling The workflow of this embodiment mainly focuses on steps S1 to S3: S1: In the pretreatment stage, the working electrode is sterilized with ultraviolet light for 30 minutes, inoculated with target microorganisms such as electrogenic bacteria, and cultured at 30 degrees Celsius. The scanning step size of the near-infrared spectroscopy module is initialized to 10 nm per step, and the spectral acquisition interval is 10 impedance frequency points. The parameters of the external electrochemical impedance analysis module are initialized, such as a voltage amplitude of 10 mV peak-to-peak value, a frequency scan range of 100 Hz to 100 kHz, and 200 sampling points per frame, ensuring synchronous acquisition of dual modes.

[0070] S2: During the dual-mode signal acquisition phase, the near-infrared spectral data acquisition and the electrochemical impedance spectroscopy data acquisition are synchronized. The external electrochemical impedance analysis module outputs a sinusoidal voltage according to set parameters, acquires the impedance spectrum through a three-electrode system, and transmits the real part, imaginary part, and phase angle data of the impedance to the multi-parameter decoupling module in real time. The near-infrared spectroscopy module starts synchronously with the impedance acquisition, acquiring one near-infrared spectrum for every 10 impedance frequency points. The spectral data is first subjected to dual-wavelength differential analysis at 970nm and 1450nm to eliminate water interference, and then subjected to Standard Normal Variation (SNV) to eliminate baseline drift and scattering effects, obtaining preprocessed near-infrared characteristic data.

[0071] S3: Multi-parameter decoupling stage. The multi-parameter decoupling module performs the following pre-decoupling steps: Equivalent circuit fitting: The impedance spectrum data of the external electrochemical impedance analysis module is fitted using an equivalent circuit model to obtain initial candidate values ​​for charge transfer resistance Rct and diffusion resistance Rd.

[0072] Near-infrared spectroscopy verification: Verification is performed by combining the peak intensity at 1450 nm in the preprocessed near-infrared spectral data. For example, if Rct decreases and the peak intensity at 1450 nm increases, it is determined that the increased electron transfer efficiency is due to enhanced microbial activity; if Rct decreases but the peak intensity at 1450 nm decreases, it is determined that the improved mass transfer is due to the thinning of the biofilm, thus distinguishing the physical meaning of the parameter changes.

[0073] Reinforcement learning optimization: The reinforcement learning algorithm is started, with impedance phase angle change rate, 1450nm peak intensity and signal-to-noise ratio as state variables, and the decoupling accuracy of Rct and Rd as the reward function objective. The equivalent circuit fitting parameters are dynamically adjusted to ensure that the decoupling accuracy error is less than 3%, and finally the accurate decoupled Rct and Rd values ​​are output.

[0074] Through the above steps, this embodiment verifies the integration of the system hardware architecture and the effectiveness of the pre-decoupling mechanism that combines equivalent circuit fitting, near-infrared spectroscopy verification, and reinforcement learning optimization.

[0075] Example 2: Implementation and Application Verification of Self-Healing Feedback and Deep Fusion Methods This embodiment describes in detail the self-healing feedback mechanism and data fusion process of the system of the present invention, and verifies them in combination with the application of anode biofilm monitoring in microbial fuel cells (MFC).

[0076] I. Implementation of the self-healing feedback mechanism The self-healing feedback module consists of an impedance monitoring unit, a microfluidic pump, and a repair fluid reservoir.

[0077] Impedance monitoring unit: Real-time acquisition of impedance data at three key frequency points (1kHz, 10kHz, and 100kHz) output from the external electrochemical impedance analysis module. The 1kHz impedance reflects charge transfer at the electrode interface, while the 10kHz and 100kHz impedances reflect mass transfer processes in the solution.

[0078] Repair fluid storage: Stores 0.1% by weight chitosan-polyvinyl alcohol prepolymer solution.

[0079] Triggering conditions and repair actions: When the impedance drift exceeds 5% for three consecutive monitoring points at any frequency, the impedance monitoring unit immediately triggers a microfluidic pump to inject 10 μL of repair solution. The prepolymer solution dynamically crosslinks with imine bonds to repair biofilm defects or electrode corrosion on the electrode surface. The repair response time is less than 30 seconds.

[0080] II. Workflow: Self-healing feedback and data fusion output The workflow of this embodiment mainly focuses on steps S4 to S5, and is validated by long-term monitoring of the MFC anode biofilm: S4: The self-healing feedback phase system performs long-term monitoring of the MFC anode biofilm. During monitoring, the self-healing feedback module receives impedance data at key frequency points in real time. For example, when the impedance at 1 kHz changes from 300... Rising to 315 If the drift remains above 5% in two subsequent monitoring sessions, the system determines that the electrode performance has degraded and immediately triggers a closed-loop repair process. After the repair solution is injected by the microfluidic pump, impedance data is re-acquired using an external electrochemical impedance analysis module to verify the impedance recovery rate. Experimental results show that during 30 days of long-term monitoring, the impedance drift of the self-healing electrode is consistently below 5%, which is far superior to the drift of more than 25% for traditional electrodes.

[0081] Figure 3 This is a comparison chart of impedance drift between self-healing electrodes and conventional electrodes. From... Figure 3 As can be clearly seen, the impedance drift of the traditional electrode increases rapidly over time, reaching 28% on day 30. However, the self-healing electrode of this invention, through closed-loop feedback, successfully stabilizes the impedance drift below 5%, fully demonstrating the long-term stability of the self-healing feedback mechanism.

[0082] S5: Data Fusion and Output Stage Data preprocessing: The decoupled impedance characteristics Rct / Rd and the sum of the real and imaginary parts of the impedance are normalized to eliminate dimensional differences. Near-infrared spectral data are subjected to SNV processing to eliminate baseline drift and scattering effects.

[0083] Deep fusion: The normalized impedance characteristics are spliced ​​with the preprocessed near-infrared characteristic data and used as input.

[0084] Model computation: The concatenated feature data is input into a deep learning model implemented using a convolutional neural network. The model extracts local features through multiple convolutional layers, reduces feature dimensionality through pooling layers, and finally integrates cross-modal features through a fully connected layer.

[0085] Output: The model ultimately outputs a microbial activity score and metabolite concentration. The activity score is calculated based on a combination of Rct electron transport efficiency and metabolic activity at the 1450 nm peak intensity.

[0086] III. Application Verification and Figure Description Taking MFC anode biofilm monitoring as an example: On day 7, the Rct of the self-healing electrode decreased to 300. The peak intensity at 1450nm increased by 3 times, and the activity score was 82. On days 14 and 30, the system maintained impedance stability through self-healing feedback, with the activity score stabilizing between 85 and 90 and the power density stabilizing at around 12.5 milliwatts per square meter.

[0087] Figure 2The diagram shows the correlation between impedance spectroscopy and near-infrared spectroscopy. Correlation analysis of long-term monitoring data clarified a positive correlation between "decreasing Rct and increasing 1450nm peak intensity," with a Pearson correlation coefficient of 0.92. This correlation physically verifies the accuracy of the S3 decoupling mechanism, indicating that the decrease in Rct indeed stems from increased extracellular polymeric substances (EPS) due to enhanced microbial activity, rather than simply improved mass transfer due to biofilm thinning. When an increase in Rct is detected but the 1450nm peak intensity remains unchanged, the system can accurately determine that it is due to electrode surface contamination rather than decreased microbial activity, guiding the user to perform physical cleaning. This verifies the precise decoupling capability of the decoupling mechanism in practical applications.

[0088] The verification in this embodiment proves that the self-healing feedback mechanism of the present invention can effectively ensure the long-term stable operation of the system, while the data fusion and deep learning model can achieve accurate assessment of the metabolic state of microorganisms.

[0089] 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 present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism, characterized in that: include: External electrochemical impedance spectroscopy module: used to output excitation signal and obtain electrochemical impedance spectroscopy data of the microbial electrode through a three-electrode system; Near-infrared spectroscopy module: used to acquire near-infrared spectral data of the microbial electrode; The three-electrode system includes a working electrode, a counter electrode, and a reference electrode, which are used in conjunction with the external electrochemical impedance analysis module to obtain electrochemical impedance spectroscopy data. The working electrode is made of a self-healing electrode material. Multi-parameter decoupling module: used to receive the electrochemical impedance spectroscopy data and the near-infrared spectroscopy data, and clearly distinguish and output the decoupled charge transfer resistance and diffusion resistance through a pre-decoupling mechanism; Data fusion module: used to fuse the decoupled impedance characteristics and near-infrared spectral characteristics, and output microbial activity scores and metabolite concentrations; Self-healing feedback module: Based on the impedance data drift monitored by the external electrochemical impedance analysis module, when the impedance data drift exceeds a preset threshold, a closed-loop repair action is triggered on the working electrode.

2. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 1, characterized in that: The near-infrared spectroscopy module includes a miniature fiber optic spectrometer with a working wavelength range of 900~1700nm, and uses a dual-wavelength differential absorption method with 970nm and 1450nm as characteristic wavelengths to suppress water interference.

3. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 1, characterized in that: The multi-parameter decoupling module includes an equivalent circuit fitting unit and a reinforcement learning optimization unit, used for: The electrochemical impedance spectroscopy data are fitted with an equivalent circuit to obtain the initial charge transfer resistance and diffusion resistance; Verification was performed by combining the intensity of specific peaks in the near-infrared spectral data; Reinforcement learning algorithms are used to dynamically optimize fitting parameters in order to achieve parameter decoupling.

4. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 3, characterized in that: The state space of the reinforcement learning algorithm includes the impedance phase angle change rate, the specific peak intensity, and the signal-to-noise ratio. Its reward function aims to maximize the decoupling accuracy between the charge transfer resistor and the diffusion resistor.

5. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 1, characterized in that: The working electrode in the three-electrode system is composed of a carbon paper surface-modified self-healing hydrogel, which contains a polyvinyl alcohol-chitosan matrix and carbon nanotubes dispersed therein.

6. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 1, characterized in that: The self-healing feedback module includes an impedance monitoring unit, a microfluidic pump, and a repair fluid reservoir. The impedance monitoring unit monitors the impedance data of the external electrochemical impedance analysis module at 1 kHz, 10 kHz, and 100 kHz in real time. When the impedance drift at any frequency point exceeds 5% for three consecutive monitoring points, the microfluidic pump is triggered to inject the repair fluid.

7. The microbial electrode online monitoring system based on multimodal fusion and self-healing mechanism according to claim 1, characterized in that: The data fusion module is implemented based on a convolutional neural network (CNN) and deployed on a heterogeneous system-on-chip (SoC). The input of the CNN is the normalized decoupled impedance characteristics and the near-infrared spectral characteristics, and the output is the microbial activity score and the concentration of metabolites.

8. A method for online monitoring of microbial electrodes based on multimodal fusion and self-healing mechanisms, characterized in that: Includes the following steps: S1: Pretreatment stage, sterilize the working electrode and inoculate it with target microorganisms, and initialize the parameters for dual-modal acquisition and analysis; S2: During the dual-modal signal acquisition stage, the electrochemical impedance spectroscopy data and near-infrared spectral data processed to suppress water interference of the microbial electrode are acquired simultaneously; S3: Multi-parameter decoupling stage, through the pre-decoupling steps of equivalent circuit fitting, near-infrared spectroscopy verification and reinforcement learning optimization, outputs the decoupled charge transfer resistance and diffusion resistance; S4: Self-repair feedback stage, real-time monitoring of impedance data drift at key frequency points, and triggering repair action on the working electrode when the impedance drift exceeds a preset threshold. S5: Data fusion and output stage, deeply fuses the decoupled impedance characteristics and near-infrared spectral characteristics, and calculates and outputs microbial activity scores and metabolite concentrations through a deep learning model.

9. The online monitoring method for microbial electrodes based on multimodal fusion and self-healing mechanism according to claim 8, characterized in that: In step S2, the near-infrared spectral data acquisition is synchronized with the electrochemical impedance spectroscopy data acquisition, specifically, one near-infrared spectrum is acquired for every 10 impedance frequency points.

10. The online monitoring method for microbial electrodes based on multimodal fusion and self-healing mechanism according to claim 8, characterized in that: In step S5, the formula used to normalize the impedance characteristics is: in, For frequency Complex impedance at that point, These are the minimum and maximum values ​​of the complex impedance at all frequencies, respectively. The formula for performing Standard Normal Transform (SNV) processing on the near-infrared spectral data is as follows: in, For the original spectrum at wavelength i The absorbance value at that point This represents the average absorbance across all wavelengths of the spectrum. The standard deviation is denoted as .

Images