A bionic colon micro-ecological monitoring and adjusting system and method based on an electronic nose

Abstract

The application provides a bionic colon micro-ecological monitoring and adjusting system and method based on an electronic nose, relates to the field of in-vitro intestinal fermentation simulation and micro-ecological process control, and comprises the following steps: a bionic colon culture module provides an anaerobic continuous flow fermentation environment for intestinal flora and generates headspace gas reflecting the metabolic state of the flora; a gas sensing module outputs a multi-dimensional response signal after pre-treating the headspace gas; a processing module performs compensation processing and feature extraction on the multi-dimensional response signal, compares the feature representation with a steady-state odor fingerprint reference model, obtains the deviation degree and drift direction of the current micro-ecological state, and generates a differential regulation instruction; and a regulation execution module adjusts the substrate supply amount, liquid dilution rate and pH of the bionic colon culture module respectively according to the differential regulation instruction. The application realizes long-term online monitoring and self-adaptive adjustment of the micro-ecological state in the continuous flow bionic colon culture process.

Description

Technical Field

[0001] This invention relates to the field of in vitro intestinal fermentation simulation and microecological process control technology, and in particular to a biomimetic colonic microecological monitoring and regulation system and method based on an electronic nose. Background Technology

[0002] In vitro gut fermentation models are important tools for studying gut microbiota metabolism, succession, prebiotic utilization, and drug biotransformation in the gut. Compared to static batch fermentation systems, continuous-flow biomimetic colon models more closely resemble the continuous input of nutrients, continuous removal of metabolites, and relatively stable environmental parameters in the human colon environment, thus possessing high application value in microecological research and functional evaluation. Existing in vitro biomimetic colon devices typically employ multi-stage or single-stage continuous-flow structures, simulating the physiological environment of different segments of the colon by adjusting pH, substrate input, and hydraulic retention time, providing relatively controllable in vitro culture conditions for the gut microbiota.

[0003] In terms of process monitoring, the current assessment of the microecological status of continuous flow biomimetic colon systems mainly relies on offline liquid-phase detection methods, including 16S rRNA sequencing, short-chain fatty acid quantification, and metabolomics analysis. While these methods offer rich information, they have long detection cycles, making it difficult to support real-time feedback control. Furthermore, frequent liquid-phase sampling can disturb the anaerobic environment and effective working volume of the reactor, hindering the maintenance of stability during long-term continuous culture. Although some studies have attempted to use electronic noses or gas sensor arrays for online characterization of fermentation headspace gases to quickly reflect metabolic activity and abnormal putrefactive gas generation trends, these explorations generally lack low-disturbance sampling designs, sensor anti-condensation and anti-toxication protection mechanisms required for long-term continuous operation, and a quantitative stability index system that can be effectively coupled with control strategies.

[0004] Chinese invention patent application number 202211649753.6 discloses an electronic nose odor analysis system and method applicable to multiple scenarios. The system consists of an industrial control computer, an analyzer, and a sampler. The analyzer incorporates a constant temperature chamber and a ring-shaped gas-sensitive sensor array, supporting both pin-type offline headspace sampling and multi-channel cyclic sampling modes. At the data processing level, it integrates dimensionality reduction algorithms such as principal component analysis and linear discriminant analysis, as well as a neural network-based qualitative and quantitative odor analysis method, enabling odor identification and concentration estimation for various gas samples. However, its sensor signal processing link primarily focuses on feature extraction and classification of individual sampling batches, failing to perform low-disturbance, long-term online monitoring of the headspace gas in the colon, resulting in reduced accuracy of evaluation results under long-term online operation conditions in continuous flow fermentation processes. Summary of the Invention

[0005] In view of this, the present invention provides a bionic colon microecological monitoring and regulation system and method based on an electronic nose, which solves the technical problems that the existing electronic nose odor analysis system lacks a steady-state reference model and deviation quantification index in the continuous flow fermentation process and cannot form a linkage between odor fingerprint information and a closed-loop control actuator, and realizes long-term online monitoring and adaptive regulation of the microecological state in the continuous flow bionic colon culture process.

[0006] The technical solution of the present invention is realized as follows:

[0007] On the one hand, the present invention provides a bionic colon microecological monitoring and regulation system based on an electronic nose, including:

[0008] A bionic colon culture module for providing an anaerobic continuous flow fermentation environment for intestinal flora and generating headspace gas reflecting the metabolic state of the flora;

[0009] A gas sensing module connected to the gas phase space of the bionic colon culture module, including a sampling pretreatment unit and a multi-channel sensor array. The sampling pretreatment unit preprocesses the headspace gas and then introduces it into the multi-channel sensor array, and the multi-channel sensor array outputs a multi-dimensional response signal reflecting the current odor fingerprint;

[0010] A processing module electrically connected to the gas sensing module, which is used for compensating and processing the multi-dimensional response signal and extracting features, obtaining a feature representation of the current odor fingerprint, comparing the feature representation with a pre-established steady-state odor fingerprint reference model, obtaining the deviation degree and drift direction of the current microecological state relative to the steady state, and generating a differential regulation instruction;

[0011] A regulation execution module electrically connected to the processing module and connected to the bionic colon culture module, which adjusts the substrate supply amount, liquid dilution rate and pH of the bionic colon culture module according to the differential regulation instruction.

[0012] Based on the above technical solution, preferably, the bionic colon culture module includes a culture cavity and a constant temperature control unit, an anaerobic maintenance unit and a continuous flow inlet and outlet component that are configured with the culture cavity. Among them,

[0013] The constant temperature control unit is used to maintain the internal temperature of the culture cavity within a preset temperature range;

[0014] The anaerobic maintenance unit is used to maintain an anaerobic state in the culture cavity;

[0015] The continuous flow inlet and outlet component is used to input substrates into the culture cavity and discharge waste liquid to maintain a continuous flow fermentation state. Among them, a gas phase space is provided at the top of the culture cavity, and the gas phase space is connected to the gas sensing module through a sampling pipeline.

[0016] Based on the above technical solutions, preferably, the multi-channel sensor array includes an electronic nose sensor array, a gas sensor group, and a temperature and humidity sensor; wherein,

[0017] An electronic nose sensor array, including one or more of MOS sensors, CP sensors, and electrochemical cross-sensitive sensors, for detecting metabolic gases;

[0018] The gas sensor group includes electrochemical Sensors, electrochemistry Sensors, NDIR sensor;

[0019] Temperature and humidity sensors are installed in the sampling pipeline to collect air temperature and relative humidity.

[0020] Based on the above technical solutions, preferably, the processing module includes an edge computing unit and a central control unit, wherein the edge computing unit is electrically connected to a multi-channel sensor array and is used to perform signal preprocessing and feature extraction, and the central control unit is communicatively connected to the edge computing unit and is used to perform odor fingerprint analysis, anomaly detection, and instruction generation. The specific steps of the processing module include:

[0021] Each sensor channel is sampled at a preset sampling period. The current output signal of each sensor channel is normalized relative to the baseline reference value established during steady-state operation to obtain the relative response value of each channel. Environmental drift compensation is performed on the relative response value of each channel based on the temperature and relative humidity collected by the temperature and humidity sensor to obtain the compensated signal of each channel.

[0022] Within a preset sliding time window, window features are extracted from the compensated signals of each channel, and the window features of all channels are concatenated to form a high-dimensional feature vector; the window features include one or more of the following: mean, peak value, integral area, maximum slope, recovery slope, volatility, and inter-channel ratio;

[0023] The high-dimensional feature vector is projected to the low-dimensional space through a low-dimensional mapping transformation to obtain a low-dimensional representation vector. The deviation of the low-dimensional representation vector from the pre-established steady-state odor fingerprint reference model is calculated. The deviation reflects the magnitude of the current odor fingerprint's deviation from the steady-state reference domain.

[0024] The drift trajectory of the time series of low-dimensional representation vectors is evaluated to obtain a drift trend index that reflects the dynamic evolution trend of the micro-ecological state. The drift trend index includes a positional direction similarity component and a drift trend direction similarity component.

[0025] Based on the offset direction and drift trend direction of the current sample in the low-dimensional space, the reference directions of each anomaly type are matched and calculated to obtain the comprehensive direction matching score of each anomaly type. Based on this, adaptive anomaly pattern judgment is performed, and the anomaly type identification result of the current drift is output.

[0026] Based on the deviation, drift trend indicators and anomaly type judgment results, state transition judgment is performed in the state machine to determine the current state of the system;

[0027] Based on the current status and the result of the anomaly type determination, differentiated control instructions are generated for the feed pump, dilution and liquid discharge control module, and pH titration module.

[0028] Based on the above technical solution, preferably, the calculation steps for the deviation include:

[0029] Principal component analysis was performed on the steady-state sample set to obtain the steady-state mean vector. and principal component projection matrix ,in , The feature dimension, The principal component dimension after dimensionality reduction and ;

[0030] Establishing a steady-state reference center in low-dimensional space and regularized covariance matrix , will be the moment The high-dimensional feature vector is mapped to a low-dimensional representation vector, and the deviation is calculated:

[0031]

[0032]

[0033]

[0034] in, Indicates time The low-dimensional representation vector, Indicates time The high-dimensional feature vector, Represents the principal component projection matrix transpose, Indicates time The current deviation of the odor fingerprint from the steady-state reference domain. It is the inverse of the regularized covariance matrix. The covariance matrix of the steady-state sample in low-dimensional space. The regularization coefficient is and , It is an identity matrix.

[0035] Based on the above technical solutions, preferably, the states of the state machine include a normal state, a drift state, a protection state, and a recovery state, and the finite state machine is provided with conditions satisfying... The three-level deviation threshold, among which To restore the judgment threshold, This indicates the drift detection threshold. To protect the judgment threshold, the system as a whole adopts a hysteresis mechanism for state machine transitions, and the transition rules between each state are as follows:

[0036] When the system is in a normal state and the deviation is continuous More than one time window At that time, it enters a drift state;

[0037] When deviation Exceed or electrochemical The sensor detected that the hydrogen sulfide concentration exceeded the preset high-level alarm threshold, or an electrochemical... When the sensor detects that the ammonia concentration exceeds the preset high ammonia concentration alarm threshold, or the pH of the bionic colon culture module exceeds the preset safe range, it immediately switches to protection mode.

[0038] When the system is in a drift state or protection state, and the deviation is... continuous A time window below When all the concentrations of the risk gases are below their respective recovery thresholds, the system enters a recovery state.

[0039] When the system is in recovery state, and continuously Deviation of each time window Persistently below Then, it returns to normal.

[0040] Based on the above technical solutions, preferably, the drift trajectory evaluation specifically includes:

[0041] Based on the current position offset vector and the current smooth drift trend vector The similarity components of location and drift trend are calculated by matching them with the reference directions of each anomaly type. The drift anomaly types include acidification drift, putrefaction drift, and other anomalous drifts.

[0042]

[0043]

[0044] in, This indicates the direction of the current sample's position offset relative to the steady-state center and the first... Cosine similarity of the reference direction for anomalies. Indicates the direction from the steady-state reference center to the first The reference direction vector of the mean center of the anomaly type. Indicates the current smooth drift trend direction relative to the first... Cosine similarity of the reference direction for anomalies. This represents the position offset vector of the current low-dimensional representation vector relative to the steady-state reference center. This represents the difference vector between the current low-dimensional representation vector and the low-dimensional representation vectors from multiple time windows prior.

[0045] The comprehensive orientation matching score is calculated based on the position orientation similarity component and the drift trend orientation similarity component:

[0046]

[0047] in, Indicates the first The drift direction matching score corresponding to the reference drift direction of the anomaly type. Weighting coefficients representing positional and directional similarity. The weighting coefficients represent the similarity of the drift trend directions, and satisfy the following conditions: .

[0048] Based on the above technical solutions, preferably, the method for generating the differentiated control instructions is as follows:

[0049] Under normal conditions, the central control unit maintains the current feed flow rate, liquid dilution rate and pH target value unchanged, and does not output control intervention commands.

[0050] During drift, the central control unit determines the control strategy based on the type of drift anomaly:

[0051] When the abnormal drift type is acidification drift, reduce the carbon source feed flow rate and increase the dilution rate, and adjust it in conjunction with the pH control parameters;

[0052] When the abnormal drift type is putrefactive drift, reduce the nitrogen source feed flow rate and maintain or adjust the supply of easily fermentable carbon sources according to the overall metabolic state, while increasing the dilution rate.

[0053] If the abnormal drift type is other abnormal drift, the carbon source feed flow rate and nitrogen source feed flow rate will be adjusted.

[0054] In the protection state, the central control unit adjusts the feed flow rate and liquid dilution rate, and suspends routine microbial intervention operations;

[0055] In the recovery state, the central control unit gradually adjusts the feed flow rate, liquid dilution rate and pH target value according to the preset step size until the state machine returns to the normal state.

[0056] More preferably, the control and execution module includes a feed pump, a dilution and liquid discharge control module, a pH titration module, and a waste liquid collection container, wherein,

[0057] The feed pump is used to adjust the feed flow rates of carbon source and nitrogen source according to differentiated control instructions, so as to control the substrate supply of the biomimetic colon culture module.

[0058] The dilution and dispensing control module is used to adjust the dispensing pump speed to control the dispensing volume per unit time and adjust the liquid dilution rate to the value specified by the command.

[0059] The pH titration module is used to add acid or alkali to the biomimetic colon culture module according to differentiated control instructions in order to adjust the pH to the target value.

[0060] The waste liquid collection container is connected to the liquid outlet of the bionic colon culture module and is used to collect the waste liquid generated during the culture process.

[0061] In addition, the present invention also provides a method for monitoring and regulating the biomimetic colonic microecology based on an electronic nose, which is applied to the biomimetic colonic microecology monitoring and regulation system based on an electronic nose as described above, and includes the following steps:

[0062] In the biomimetic colon culture module, an anaerobic continuous flow fermentation environment is provided for the intestinal flora, and headspace gas generated in the gas phase space of the biomimetic colon culture module is continuously collected.

[0063] The headspace gas is detected by a multi-channel sensor array, and a multi-dimensional response signal reflecting the current odor fingerprint is output.

[0064] The multidimensional response signal is compensated and its features are extracted to obtain the characteristic representation of the current odor fingerprint;

[0065] The feature representation is compared with a pre-established steady-state odor fingerprint reference model to calculate the deviation of the current micro-ecological state from the steady state, and the drift direction is comprehensively evaluated to determine the abnormal mode of the current micro-ecological state.

[0066] Based on the identified abnormal patterns, differentiated control instructions are generated to adjust the substrate supply, liquid dilution rate, and pH of the biomimetic colon culture module.

[0067] The present invention has the following advantages over the prior art:

[0068] (1) Based on the bionic colon culture module, gas sensing module, processing module and regulation execution module, the whole-link automation from headspace gas signal acquisition to microecological state judgment to control command output is realized, avoiding the mode of relying on offline liquid phase detection for manual judgment and manual intervention, and realizing long-term online monitoring and adaptive regulation of microecological state during continuous flow bionic colon culture.

[0069] (2) By extracting window features from the signal after channel compensation, a low-dimensional representation vector is obtained through low-dimensional mapping transformation, and drift trajectory evaluation is performed. Complex high-dimensional sensing information is transformed into a quantitative stability index suitable for threshold determination and closed-loop control, which solves the problem that high-dimensional odor fingerprint signals are difficult to be directly used for threshold determination and feedback control.

[0070] (3) By setting up cross-sensitive sensor groups, dedicated hazardous gas sensor groups and temperature and humidity sensors, the long-term availability of the electronic nose array in sulfur-containing fermentation gas environments can be improved.

[0071] (4) The state of the state machine is determined by the deviation, drift trend index and abnormality type judgment results, and combined with the biological observation period, the over-regulation and control oscillation caused by microecological lag is reduced, and the stability and automation of the long-term operation of the continuous flow biomimetic colon system are improved. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a schematic diagram of the system composition of a biomimetic colonic microecological monitoring and regulation system based on an electronic nose according to the present invention;

[0074] Figure 2 This is a schematic diagram illustrating the construction of a steady-state odor fingerprint reference model for a biomimetic colonic microecological monitoring and regulation system based on an electronic nose, according to the present invention.

[0075] Figure 3 This is a schematic diagram of the state machine state transition of a biomimetic colonic microecological monitoring and regulation system based on an electronic nose according to the present invention.

[0076] Explanation of reference numerals in the attached figures:

[0077] 1. Continuous flow biomimetic colon reactor; 2. Headspace gas online sampling and pretreatment module; 3. Gas sensing module; 4. Edge computing and central control unit; 5. Feed pump; 6. Dilution and liquid discharge control module; 7. pH titration module; 8. Waste liquid collection container. Detailed Implementation

[0078] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0079] This invention provides a biomimetic colonic microecological monitoring and regulation system based on an electronic nose, comprising:

[0080] A biomimetic colon culture module is used to provide an anaerobic continuous flow fermentation environment for the gut microbiota and to generate headspace gas that reflects the metabolic state of the microbiota.

[0081] The gas sensing module is connected to the gas phase space of the bionic colon culture module. It includes a sampling preprocessing unit and a multi-channel sensor array. The sampling preprocessing unit preprocesses the headspace gas and then imports it into the multi-channel sensor array. The multi-channel sensor array outputs a multi-dimensional response signal that reflects the current odor fingerprint.

[0082] The processing module, electrically connected to the gas sensing module, is used to compensate and extract features from the multidimensional response signal to obtain the feature representation of the current odor fingerprint. The feature representation is then compared with a pre-established steady-state odor fingerprint reference model to obtain the degree of deviation and drift direction of the current micro-ecological state relative to the steady state, and to generate differentiated control instructions.

[0083] The control and execution module is electrically connected to the processing module and connected to the biomimetic colon culture module. It adjusts the substrate supply, liquid dilution rate and pH of the biomimetic colon culture module according to differentiated control commands.

[0084] This invention, based on a biomimetic colon culture module, a gas sensing module, a processing module, and a regulation and execution module, achieves full-link automation from headspace gas signal acquisition to microecological state judgment and control command output. It avoids the reliance on offline liquid phase detection methods for manual judgment and intervention, and realizes long-term online monitoring and adaptive regulation of the microecological state during continuous flow biomimetic colon culture.

[0085] Specifically, such as Figure 1As shown, the system includes a continuous flow biomimetic colon reactor 1, a headspace gas online sampling and pretreatment module 2, a gas sensing module 3, an edge computing and central control unit 4, a feed pump 5, a dilution and liquid discharge control module 6, a pH titration module 7, and a waste liquid collection container 8.

[0086] The continuous-flow biomimetic colon reactor 1 provides a constant-temperature, anaerobic, continuous-flow fermentation environment, with a gas phase space at the top. The headspace gas online sampling and pretreatment module 2 is connected to the gas phase space to continuously acquire headspace metabolic gases. The reactor's working volume is preferably 0.5–5 L, with the gas phase space accounting for 5%–30% of the total volume. The operating temperature is preferably 35–39°C. The hydraulic retention time (HRT) under continuous-flow conditions is preferably 8–48 h (more preferably 12–24 h), corresponding to a dilution rate D = 1 / HRT. Microbial inoculation can use fecal microbial suspension or a mixture of target microorganisms, with an inoculation amount preferably 5%–20% of the working volume. After a colonization / stabilization period of 12–48 h, the steady-state modeling stage begins. Unspecified culture media or substrate formulations can be selected according to the experimental purpose, such as standard biomimetic colon culture media, prebiotic substrates, or protein / amino acid substrates.

[0087] The headspace gas online sampling and pretreatment module 2 preferably includes a miniature diaphragm pump, an anti-backflow microporous membrane, a flow stabilization unit, an anti-condensation unit, a gas path switching valve, and an inert gas backflushing interface. After being processed by the module, the sampled gas enters the sensing chamber of the gas sensing module 3. In normal monitoring mode, it can be returned to the gas phase zone of the reactor 1 to reduce disturbance to the gas phase composition and anaerobic state of the reactor; in protection mode, it can be vented through a bypass and the sensing chamber can be backflushed and cleaned by inert gas.

[0088] Electrochemical sensors are preferred for the hazardous gas channels: the hydrogen sulfide (H2S) channel range is preferably 0–200 ppm or 0–500 ppm, with a minimum detection limit preferably ≤1 ppm; the ammonia (NH3) channel range is preferably 0–100 ppm or 0–500 ppm, with a minimum detection limit preferably ≤5 ppm. The NDIR carbon dioxide channel range is preferably 0–20% (v / v) to cover fluctuations in fermentation metabolism. To support temperature and humidity compensation and dew point management, the temperature and humidity sensors preferably have an accuracy within ±2%RH and ±0.5℃ within the range of 0–100%RH and 10–45℃.

[0089] Gas sensing module 3 includes an electronic nose sensor array and an electrochemical sensor array. Sensors, electrochemistry Sensors, NDIR Sensors and temperature / humidity sensors. The electronic nose sensor array preferably includes one or more combinations of MOS sensors, CP sensors, and electrochemical cross-sensitive sensors to form a cross-sensitive response to metabolic gases such as volatile fatty acids, nitrogen-containing volatiles, and sulfur-containing volatiles; electrochemical... Sensors and electrochemistry Sensors are used for online monitoring of hazardous gases; NDIR Sensors are used to characterize overall metabolic activity as a reference; temperature and humidity sensors are used for drift compensation and dew point management.

[0090] The edge computing and central control unit 4 is electrically connected to the gas sensing module 3 and is used to perform signal acquisition, baseline update, temperature and humidity compensation, window feature extraction, low-dimensional mapping, deviation calculation, state determination and control command generation. The control commands are sent to the feed pump 5, the dilution and liquid discharge control module 6 and the pH titration module 7 respectively to adjust the carbon source feed, nitrogen source feed, dilution rate and pH control parameters in a coordinated manner.

[0091] In one embodiment of the present invention, the biomimetic colon culture module includes a culture chamber and a constant temperature control unit, an anaerobic maintenance unit, and a continuous flow inlet / outlet liquid assembly configured in conjunction with the culture chamber.

[0092] The temperature control unit is used to maintain the internal temperature of the culture chamber within a preset temperature range;

[0093] The anaerobic maintenance unit is used to maintain the anaerobic state within the culture chamber;

[0094] The continuous flow inlet / outlet assembly is used to input substrate into the culture chamber and discharge waste liquid to maintain continuous flow fermentation. The culture chamber has a gas phase space at the top, which is connected to the gas sensing module through a sampling pipeline.

[0095] Understandably, the anaerobic maintenance unit maintains the anaerobic state within the chamber by continuously introducing nitrogen or a carbon dioxide / nitrogen mixture. The continuous flow inlet / outlet assembly is equipped with an inlet line driven by a peristaltic pump and an outlet line controlled by a dilution and outlet control module to maintain the hydraulic retention time (HRT) required for continuous flow fermentation. The inoculum is derived from fresh human fecal samples, diluted and homogenized with buffer before being injected into the chamber. Before system operation, it must be continuously acclimatized under the above conditions until the composition of the microbial metabolites tends to stabilize before entering the steady-state modeling stage.

[0096] In one embodiment of the present invention, the multi-channel sensor array includes an electronic nose sensor array, a gas sensor group, and a temperature and humidity sensor; wherein...

[0097] An electronic nose sensor array, including one or more of MOS sensors, CP sensors, and electrochemical cross-sensitive sensors, for detecting metabolic gases;

[0098] The gas sensor group includes electrochemical Sensors, electrochemistry Sensors, NDIR sensor;

[0099] Temperature and humidity sensors are installed in the sampling pipeline to collect air temperature and relative humidity.

[0100] This invention improves the long-term availability of electronic nose arrays in sulfur-containing fermentation gas environments by setting up cross-sensitive sensor groups, dedicated hazardous gas sensor groups, and temperature and humidity sensors.

[0101] In one embodiment of the present invention, the processing module includes an edge computing unit and a central control unit. The edge computing unit is electrically connected to a multi-channel sensor array and is used to perform signal preprocessing and feature extraction. The central control unit is communicatively connected to the edge computing unit and is used to perform odor fingerprint analysis, anomaly detection, and instruction generation. The specific steps of the processing module include:

[0102] Each sensor channel is sampled at a preset sampling period. The current output signal of each sensor channel is normalized relative to the baseline reference value established during steady-state operation to obtain the relative response value of each channel.

[0103]

[0104] in, Indicates the first Each sensing channel at time The relative response value, This represents the raw output of the i-th channel at time t. Indicates the first Each sensing channel at time The baseline response value, represents a very small positive number, i represents the sensor channel number, and t represents the sampling time;

[0105] Environmental drift compensation is performed on the relative response values ​​of each channel based on the temperature and relative humidity collected by the temperature and humidity sensors to obtain the compensated signal of each channel.

[0106] Within a preset sliding time window, window features are extracted from the compensated signals of each channel, and the window features of all channels are concatenated to form a high-dimensional feature vector. , Indicates time High-dimensional feature vectors, R m express 3D real space, m The feature dimension is determined by the number of sensing channels and the number of feature types; among which window features include one or more of the following: mean, peak value, integral area, maximum slope, recovery slope, volatility, and inter-channel ratio.

[0107] The high-dimensional feature vector is projected to the low-dimensional space through a low-dimensional mapping transformation to obtain a low-dimensional representation vector. The deviation of the low-dimensional representation vector from the pre-established steady-state odor fingerprint reference model is calculated. The deviation reflects the magnitude of the current odor fingerprint's deviation from the steady-state reference domain.

[0108] The drift trajectory of the time series of low-dimensional representation vectors is evaluated to obtain a drift trend index that reflects the dynamic evolution trend of the micro-ecological state. The drift trend index includes a positional direction similarity component and a drift trend direction similarity component.

[0109] Based on the offset direction and drift trend direction of the current sample in the low-dimensional space, the reference directions of each anomaly type are matched and calculated to obtain the comprehensive direction matching score of each anomaly type. Based on this, adaptive anomaly pattern judgment is performed, and the anomaly type identification result of the current drift is output.

[0110] Based on the deviation, drift trend indicators and anomaly type judgment results, state transition judgment is performed in the state machine to determine the current state of the system;

[0111] Based on the current status and the result of the anomaly type determination, differentiated control instructions are generated for the feed pump, dilution and liquid discharge control module, and pH titration module.

[0112] Understandably, the relative response values ​​of each channel The calculation method is the same as in the steady-state modeling stage, with the baseline response value... Under normal conditions, slow updates can be performed using a long-period moving average strategy to correct long-term baseline drift of the sensor. When the system is in a drift state or protection state, baseline updates are paused to prevent abnormal responses from being mistakenly included in the baseline, causing reference value contamination. The temperature and humidity compensation strategy is consistent with that in the steady-state modeling stage, and the compensated signal enters the subsequent window feature extraction process.

[0113] This invention extracts window features from the channel-compensated signal, obtains a low-dimensional representation vector through low-dimensional mapping transformation, and evaluates the drift trajectory. This transforms complex high-dimensional sensing information into a quantitative stability index suitable for threshold determination and closed-loop control, solving the problem that high-dimensional odor fingerprint signals are difficult to use directly for threshold determination and feedback control.

[0114] Preferably, the sampling period is 10 seconds to 5 minutes.

[0115] In one embodiment of the present invention, the sampling period is 1 minute, and the system updates the baseline of each channel using an exponential moving average or a moving average method to obtain the baseline response.

[0116] Preferably, environmental drift compensation can be achieved using lookup tables, linear correction, polynomial correction, or correction based on a calibration model. For high humidity conditions, an anti-condensation strategy is preferred to reduce the risk of condensation and maintain the relative stability of the volatile components in the headspace as much as possible.

[0117] In one embodiment of the present invention, the steps for calculating the deviation include:

[0118] Principal component analysis was performed on the steady-state sample set to obtain the steady-state mean vector. and principal component projection matrix ,in , The feature dimension, The principal component dimension after dimensionality reduction and ;

[0119] Establishing a steady-state reference center in low-dimensional space and regularized covariance matrix , will be the moment The high-dimensional feature vector is mapped to a low-dimensional representation vector, and the deviation is calculated:

[0120]

[0121]

[0122]

[0123] in, Indicates time The low-dimensional representation vector, Indicates time The high-dimensional feature vector, Represents the principal component projection matrix transpose, Indicates time The current deviation of the odor fingerprint from the steady-state reference domain. It is the inverse of the regularized covariance matrix. The covariance matrix of the steady-state sample in low-dimensional space. The regularization coefficient is and , It is an identity matrix.

[0124] Preferably, regularization is used. The value is preferably 0.001 to 0.1, more preferably 0.01, or can be adaptively selected through cross-validation / leave-one-out method.

[0125] like Figure 2 As shown, the construction process of the steady-state odor fingerprint reference model is as follows:

[0126] S1. Perform steady-state sampling during stable system operation;

[0127] S2, by sampling period Continuously acquire multi-channel sensor signals, extract window features of each channel within a sliding time window W, and concatenate them to form a steady-state feature vector set {x}. n};

[0128] S3, regarding the steady-state eigenvector set {x} n Principal component analysis is performed to train and obtain the steady-state mean vector. and principal component projection matrix Calculate the steady-state reference center in low-dimensional space The regularized covariance matrix is ​​obtained by applying a regularization term to the low-dimensional covariance matrix. , The complete parameter set of the steady-state reference model is output and stored in the processing module.

[0129] S4. During the online operation phase, the processing module continuously acquires data from the electronic nose array (MOS / CP type cross-sensitive sensor) and the hazardous gas sensor at the same sampling period. Multi-channel output of NH3 specific electrochemical sensor and temperature and humidity sensor;

[0130] S5. Preprocessed by baseline correction, temperature and humidity compensation and drift suppression;

[0131] S6. Form a high-dimensional feature vector within the sliding window W. ;

[0132] S7, Projected by principal component matrix Will Mapped to low-dimensional representation vectors ;

[0133] S8. Calculate the deviation using the squared regularized Mahalanobis distance. and output synchronously. and For use in subsequent status determination;

[0134] S9. The system determines in each cycle whether it is in a normal state and the process is stable. If so, the parameters are adjusted in a restricted manner. Perform small-weight updates to slowly track long-term baseline drift caused by factors such as batch-to-batch variability in culture media;

[0135] S10. If the system is in a drift state or a protection state, updates will be suspended to protect the reference model from being contaminated by abnormal samples.

[0136] like Figure 3 As shown, in one embodiment of the present invention, the states of the state machine include a normal state, a drift state, a protect state, and a recover state. The state machine is configured to satisfy... The three-level deviation threshold, among which, This indicates the recovery threshold. This indicates the drift detection threshold. The system uses a hysteresis mechanism for state machine transitions to indicate the protection threshold, avoiding false alarms caused by single-point instantaneous fluctuations. The transition rules between states are as follows:

[0137] When the system is in a normal state and the deviation is continuous More than one time window At that time, it enters a drift state;

[0138] When deviation Exceed or electrochemical The sensor detected that the hydrogen sulfide concentration exceeded the preset high-level alarm threshold, or an electrochemical... When the sensor detects that the ammonia concentration exceeds the preset high ammonia concentration alarm threshold, or the pH of the bionic colon culture module exceeds the preset safe range, it immediately switches to protection mode.

[0139] When the system is in a drift state or protection state, and the deviation is... continuous A time window below When all the concentrations of the risk gases are below their respective recovery thresholds, the system enters a recovery state.

[0140] When the system is in recovery state, and continuously Deviation of each time window Persistently below Then, it returns to normal.

[0141] Understandable, if any state Reaching or exceeding High alarm threshold or Reaching or exceeding High alarm threshold If the pH exceeds the threshold, it will directly enter a protective state; after adjustment, it will enter a biological observation period. ,exist Internal suppression of repetition adjustment. The length of the time window can be set according to actual usage requirements, and this invention does not impose a specific limitation on it.

[0142] This invention determines the state machine state based on deviation, drift trend indicators, and anomaly type determination results, and combines this with a biological observation period to reduce over-regulation and control oscillations caused by microecological lag, thereby improving the stability and automation of the long-term operation of the continuous flow biomimetic colon system.

[0143] In the drift state, anomaly types are preferably identified based on the overall offset direction of the low-dimensional odor fingerprint, and further verified by combining risk gas signals and process variables. Preferably, a steady-state reference center is used. Based on this, reference direction vectors for acidification drift, putrefaction drift, and other abnormal modes are pre-established. Once the system meets the drift entry conditions, the matching degree between the current sample's offset direction and smooth drift direction in the low-dimensional space and the reference direction of each abnormal mode is calculated. The abnormal mode with the highest matching score and that meets the confidence threshold is selected as the current drift type.

[0144] Preferably, if the current drift direction has the highest matching score with the acidification reference direction, it is determined to be an acidification drift; in this case, the control strategy includes reducing the carbon source feed rate and appropriately increasing the dilution rate, while adjusting the pH control parameters. If the current drift direction has the highest matching score with the putrefaction reference direction, it is determined to be a putrefaction drift; in this case, the control strategy includes reducing the nitrogen source feed rate and maintaining or slightly adjusting the supply of easily fermentable carbon sources according to the overall metabolic state, while appropriately increasing the dilution rate.

[0145] Among them, pH change trend, NDIR Output and electrochemistry Sensor output is preferred as a consistency verification condition for drift type identification; electrochemical Sensor output, electrochemical Safety events such as sensor output, pH exceeding limits and condensation risk, electronic nose array poisoning risk, sensor cavity abnormality or sampling gas path abnormality can all serve as hard trigger conditions for entering the protection state.

[0146] Understandably, the safe pH range can be set based on the simulated colon segment.

[0147] Preferably, the safe pH range is 5.8 to 7.4.

[0148] More preferably, the safe pH range is 6.0 to 7.2.

[0149] Understandably, in a protected state, the system preferentially cuts off or significantly reduces the substrate feed currently posing a risk, increases the dilution rate to reduce the accumulation of risky metabolites, and temporarily adjusts the pH control target range, titration dead zone, or titration trigger threshold according to the type of anomaly to reduce secondary disturbances caused by vigorous titration; simultaneously, when electrochemical... When the concentration detected by the sensor reaches the protection threshold of the electronic nose array, the control unit switches the gas path, isolates the electronic nose array sensing cavity, and initiates inert gas backflushing. Preferably, a dedicated risk channel and NDIR are retained in the protection state. Online monitoring of channels and pH signals is necessary to prevent the system from becoming unresponsive.

[0150] Preferably, , The optimal duration is 5 to 60 minutes, depending on the sampling period. Configure in conjunction with the sliding window W.

[0151] Preferably, after any adjustment action is performed, the system enters a biological observation period of a preset duration. During the observation period, the system continues to collect and update monitoring data, but suppresses repeated adjustments to similar mild to moderate anomalies, in order to wait for the microbial community to generate a metabolic response to the previous control action; if a risk gas exceeds the threshold, pH exceeds the threshold, or other high-risk events occur during the observation period, the observation period is allowed to be interrupted and the system directly jumps to the protection state.

[0152] In one embodiment of the present invention, drift trajectory evaluation specifically includes:

[0153] Based on the current position offset vector and the current smooth drift trend vector The similarity components of location and drift trend are calculated by matching them with the reference directions of each anomaly type. The drift anomaly types include acidification drift, putrefaction drift, and other anomalous drifts.

[0154]

[0155]

[0156]

[0157]

[0158]

[0159]

[0160] in, This indicates the direction of the current sample's position offset relative to the steady-state center and the first... Cosine similarity of the reference direction for anomalies. Indicates the direction from the steady-state reference center to the first The reference direction vector of the mean center of the anomaly type. Indicates the current smooth drift trend direction relative to the first... Cosine similarity of the reference direction for anomalies. This represents the mean center of the j-th type of abnormal pattern sample in the low-dimensional space. Indicates the steady-state reference center. This represents the position offset vector of the current low-dimensional representation vector relative to the steady-state reference center. Represents a low-dimensional representation vector. This represents the difference vector between the current low-dimensional representation vector and the low-dimensional representation vectors from multiple time windows ago. f represents the position offset vector of the low-dimensional representation vector relative to the steady-state reference center before step f, where f represents the smoothing step size;

[0161] Preferably, f is 3 to 10 sampling steps or proportional to the window length.

[0162] The comprehensive orientation matching score is calculated based on the position orientation similarity component and the drift trend orientation similarity component:

[0163]

[0164]

[0165] in, Indicates the first The drift direction matching score corresponding to the reference drift direction of the anomaly type. Weighting coefficients representing positional and directional similarity. Weighting coefficients representing the similarity of drift trend directions.

[0166] Understandably, if the maximum value among all abnormal patterns exceeds a preset matching threshold, the current drift can be determined to belong to the j-th type of abnormal pattern; otherwise, it is preferably determined to be a mixed or uncertain drift, and a conservative adjustment strategy is implemented, that is, only small adjustments are made to each control variable to avoid excessive intervention. The preset matching threshold and confidence threshold can be set based on historical data of steady-state samples and labeled abnormal samples through cross-validation, leave-one-out method, or ROC analysis. The preset matching threshold is used to determine candidate abnormal patterns, and the confidence threshold is used to determine the final output type after combining with consistency verification.

[0167] Furthermore, drift type identification also incorporates consistency verification with specific sensors: a continuously decreasing pH trend serves as supporting evidence for acidification-type drift; C NH3 ( The continued upward trend serves as further evidence of corruption-related drift; Changes in this value serve as a reference for overall metabolic activity. When the direction matching score... When the highest type matches the above consistency verification conditions, that type is output as the final anomaly type identification result; if the direction matching score conflicts with the consistency verification, the confidence of the corresponding type is reduced, and the hard trigger condition is given priority.

[0168] In one embodiment of the present invention, the differentiated control command is generated as follows:

[0169] Under normal conditions, the central control unit maintains the current feed flow rate, liquid dilution rate and pH target value unchanged, and does not output control intervention commands.

[0170] During drift, the central control unit determines the control strategy based on the type of drift anomaly:

[0171] When the abnormal drift type is acidification drift, reduce the carbon source feed flow rate and increase the dilution rate, and adjust it in conjunction with the pH control parameters;

[0172] When the abnormal drift type is putrefactive drift, reduce the nitrogen source feed flow rate and maintain or adjust the supply of easily fermentable carbon sources according to the overall metabolic state, while increasing the dilution rate.

[0173] If the abnormal drift type is other abnormal drift, the carbon source feed flow rate and nitrogen source feed flow rate will be adjusted.

[0174] In the protection state, the central control unit adjusts the feed flow rate and liquid dilution rate, and suspends routine microbial intervention operations;

[0175] In the recovery state, the central control unit gradually adjusts the feed flow rate, liquid dilution rate, and pH target value according to preset step sizes until the state machine returns to the normal state:

[0176]

[0177] in, Indicates the current regulatory amount. Indicates time The control vector, This represents the slope regression coefficient, preventing secondary disturbances caused by overly rapid recovery, and , This represents the baseline value for normal operation.

[0178] In one embodiment of the present invention, the control and execution module includes a feed pump, a dilution and liquid discharge control module, a pH titration module, and a waste liquid collection container, wherein,

[0179] The feed pump is used to adjust the feed flow rates of carbon source and nitrogen source according to differentiated control instructions, so as to control the substrate supply of the biomimetic colon culture module.

[0180] The dilution and dispensing control module is used to adjust the dispensing pump speed to control the dispensing volume per unit time and adjust the liquid dilution rate to the value specified by the command.

[0181] The pH titration module is used to add acid or alkali to the biomimetic colon culture module according to differentiated control instructions in order to adjust the pH to the target value.

[0182] The waste liquid collection container is connected to the liquid outlet of the bionic colon culture module and is used to collect the waste liquid generated during the culture process.

[0183] Understandably, the feed pump uses a multi-channel peristaltic pump, with independent carbon and nitrogen source pipelines that can operate at the flow rates set by the control commands to precisely control the substrate supply. The dilution and dispensing control module controls the dispensing volume per unit time by adjusting the dispensing pump speed, thereby adjusting the system dilution rate to the specified value. The pH titration module is equipped with acid (hydrochloric acid solution) and alkali (sodium hydroxide solution) interfaces, which can add acid or alkali to the culture chamber according to differentiated control commands to adjust the pH to the target value. At the same time, it can dynamically adjust the pH control target range, titration dead zone range, and titration trigger threshold according to commands to reduce secondary disturbances caused by vigorous titration. The waste liquid collection container is connected to the dispensing end of the culture chamber to collect the waste liquid generated during the culture process.

[0184] To verify the effectiveness of the system of the present invention, two sets of disturbance experiments were carried out using a continuous flow biomimetic colon reactor inoculated with human fecal bacteria samples. During the stable operation phase, a steady-state reference model was first established, and corresponding deviation thresholds, recovery thresholds, and risk gas thresholds were set:

[0185] The reactor operating volume was 1L, the temperature was 37℃, and the pH target was 6.6–6.9; the sampling period was [not specified]. 1 minute, sliding window The sampling period was 30 min; during the steady-state modeling phase, continuous data collection lasted 48 hours, with the number of sampling points being [number missing]. The number of window feature vectors extracted based on the sliding window is PCA dimensionality reduction with a dimension k of 3 explains approximately 90% of the cumulative variance. Deviation is measured using the squared Mahalanobis distance. To perform state recognition, three deviation thresholds are set. , , Preferably, the threshold is taken as... , , ; High alarm threshold 20ppm Array protection threshold 50ppm recovery threshold The threshold is 5 ppm; the NH3 threshold can be set using the same strategy (e.g., High alarm threshold 50 ppm Array protection threshold 100ppm recovery threshold (10 ppm); biological observation period The time frame is set to 4h. Based on the above parameters, a significant separation of the principal component distributions under normal / drift / protection states can be observed in low-dimensional space.

[0186] Example 1: Simulating acidification drift induced by high carbohydrate load.

[0187] During stable operation, the carbon source feed load was artificially increased. Following the disturbance, the system detected NDIR. Output enhancement, deviation The pH level continued to rise, while the pH showed a downward trend. Based on the joint judgment logic, the system identified it as an acidification-type drift and entered a drift state, automatically reducing the carbon source feed flow rate and increasing the dilution rate. After entering the biological observation period, the pH downward trend was suppressed and gradually recovered in the following hours. The results show that the present invention can promptly detect microecological drift and implement stable correction without relying on frequent offline sampling.

[0188] Example 2: Simulating putrefactive drift and sensor protection scenarios caused by high protein or sulfur-containing substrate loads.

[0189] During stable operation, the loading of nitrogen- or sulfur-containing substrates was artificially increased. Electrochemical detection was performed on the system. Sensor output and / or electrochemical The sensor output increases, and the deviation also increases. When the concentration reaches the preset protection threshold, the system directly enters the protection state without being limited to the observation period. It switches the gas path to isolate the electronic nose array sensing cavity and initiates inert gas backflushing, while simultaneously reducing the feed of the risky substrate and increasing the dilution rate. After both the risky gas concentration and deviation decrease, the system enters the recovery state and gradually restores the control quantity to the baseline setting using a ramp-up mechanism. The results show that this invention balances microecological anomaly control and sensor protection, making it suitable for long-term online operation.

[0190] As can be seen from Examples 1 and 2, the present invention can transform the microecological evolution trend that cannot be directly observed into a quantifiable deviation indicator, and combine the risk gas and biological observation period gating mechanism to realize the online monitoring and adaptive closed-loop regulation of the continuous flow biomimetic colon system.

[0191] This invention also provides a method for monitoring and regulating the biomimetic colonic microecology based on an electronic nose, applied to the biomimetic colonic microecology monitoring and regulation system based on an electronic nose as described above, comprising the following steps:

[0192] In the biomimetic colon culture module, an anaerobic continuous flow fermentation environment is provided for the intestinal flora, and headspace gas generated in the gas phase space of the biomimetic colon culture module is continuously collected.

[0193] The headspace gas is detected by a multi-channel sensor array, and a multi-dimensional response signal reflecting the current odor fingerprint is output.

[0194] The multidimensional response signal is compensated and its features are extracted to obtain the characteristic representation of the current odor fingerprint;

[0195] The feature representation is compared with a pre-established steady-state odor fingerprint reference model to calculate the deviation of the current micro-ecological state from the steady state, and the drift direction is comprehensively evaluated to determine the abnormal mode of the current micro-ecological state.

[0196] Based on the identified abnormal patterns, differentiated control instructions are generated to adjust the substrate supply, liquid dilution rate, and pH of the biomimetic colon culture module.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A biomimetic colonic microecological monitoring and regulation system based on an electronic nose, characterized in that, Comprising: A bionic colon culture module, used to provide an anaerobic continuous-flow fermentation environment for the intestinal flora and generate headspace gas reflecting the metabolic state of the flora; A gas sensing module, connected to the gas phase space of the bionic colon culture module, including a sampling pretreatment unit and a multi-channel sensor array. The sampling pretreatment unit preprocesses the headspace gas and then introduces it into the multi-channel sensor array, and the multi-channel sensor array outputs a multi-dimensional response signal reflecting the current odor fingerprint; A processing module, electrically connected to the gas sensing module, used to perform compensation processing and feature extraction on the multi-dimensional response signal, obtain the feature representation of the current odor fingerprint, compare the feature representation with a pre-established steady-state odor fingerprint reference model, obtain the deviation degree and drift direction of the current microecological state relative to the steady state, and generate a differential regulation instruction; A regulation execution module, electrically connected to the processing module and connected to the bionic colon culture module, and adjusts the substrate supply amount, liquid dilution rate and pH of the bionic colon culture module respectively according to the differential regulation instruction.

2. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 1, characterized in that: The bionic colon culture module includes a culture cavity and a thermostatic control unit, an anaerobic maintenance unit and a continuous-flow inlet and outlet assembly configured with the culture cavity. Among them, The thermostatic control unit is used to maintain the internal temperature of the culture cavity within a preset temperature range; The anaerobic maintenance unit is used to maintain the anaerobic state in the culture cavity; The continuous-flow inlet and outlet assembly is used to input the substrate into the culture cavity and discharge the waste liquid to maintain the continuous-flow fermentation state. Among them, a gas phase space is provided at the top of the culture cavity, and the gas phase space is connected to the gas sensing module through a sampling pipeline.

3. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 1, characterized in that: The multi-channel sensor array includes an electronic nose sensor array, a gas sensor group and a temperature and humidity sensor; among them, The electronic nose sensor array, including one or more of MOS sensors, CP sensors and electrochemical cross-sensitive sensors, is used to detect metabolic gases; The gas sensor group includes electrochemical Sensors, electrochemistry Sensors, NDIR sensor; The temperature and humidity sensor is arranged in the sampling pipeline and is used to collect the gas path temperature and relative humidity.

4. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 3, characterized in that: The processing module includes an edge computing unit and a central control unit. Among them, the edge computing unit is electrically connected to the multi-channel sensor array and is used to perform signal preprocessing and feature extraction. The central control unit is communicatively connected to the edge computing unit and is used to perform odor fingerprint analysis, anomaly judgment and instruction generation. The steps of the processing module specifically include: Sampling each sensor channel at a preset sampling period, normalizing the current output signal of each sensor channel with respect to the baseline reference value established during its steady-state operation, and obtaining the relative response value of each channel; and performing environmental drift compensation on the relative response value of each channel according to the temperature and relative humidity collected by the temperature and humidity sensor, and obtaining the compensated signal of each channel; Within a sliding time window of a preset length, extract the window features of the compensated signal of each channel, and splice the window features of all channels to form a high-dimensional feature vector; where the window features include one or more of mean value, peak value, integral area, maximum slope, recovery slope, fluctuation degree and inter-channel ratio; The high-dimensional feature vector is projected to the low-dimensional space through a low-dimensional mapping transformation to obtain a low-dimensional representation vector. The deviation of the low-dimensional representation vector from the pre-established steady-state odor fingerprint reference model is calculated. The deviation reflects the magnitude of the current odor fingerprint's deviation from the steady-state reference domain. The drift trajectory of the time series of low-dimensional representation vectors is evaluated to obtain a drift trend index that reflects the dynamic evolution trend of the micro-ecological state. The drift trend index includes a positional direction similarity component and a drift trend direction similarity component. Based on the offset direction and drift trend direction of the current sample in the low-dimensional space, the reference directions of each anomaly type are matched and calculated to obtain the comprehensive direction matching score of each anomaly type. Based on this, adaptive anomaly pattern judgment is performed, and the anomaly type identification result of the current drift is output. Based on the deviation, drift trend indicators and anomaly type judgment results, state transition judgment is performed in the state machine to determine the current state of the system; Based on the current status and the result of the anomaly type determination, differentiated control instructions are generated for the feed pump, dilution and liquid discharge control module, and pH titration module.

5. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 4, characterized in that: The steps for calculating the deviation include: Principal component analysis was performed on the steady-state sample set to obtain the steady-state mean vector. and principal component projection matrix ,in , The feature dimension, The principal component dimension after dimensionality reduction and ; Establishing a steady-state reference center in low-dimensional space and regularized covariance matrix , will be the moment The high-dimensional feature vector is mapped to a low-dimensional representation vector, and the deviation is calculated: ； ； ； in, Indicates time The low-dimensional representation vector, Indicates time High-dimensional feature vectors, Represents the principal component projection matrix transpose, Indicates time The current deviation of the odor fingerprint from the steady-state reference domain. Let be the inverse of the regularized covariance matrix. The covariance matrix of the steady-state sample in low-dimensional space. The regularization coefficient is and , It is an identity matrix.

6. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 5, characterized in that: The states of the state machine include a normal state, a drift state, a protection state, and a recovery state. The finite state machine is provided with conditions satisfying... The three-level deviation threshold, among which To restore the judgment threshold, This indicates the drift detection threshold. To protect the judgment threshold, the system as a whole adopts a hysteresis mechanism for state machine transitions, and the transition rules between each state are as follows: When the system is in a normal state and the deviation is continuous More than one time window At that time, it enters a drift state; When deviation Exceed or electrochemical The sensor detected that the hydrogen sulfide concentration exceeded the preset high-level alarm threshold, or an electrochemical... When the sensor detects that the ammonia concentration exceeds the preset high ammonia concentration alarm threshold, or the pH of the bionic colon culture module exceeds the preset safe range, it immediately switches to protection mode. When the system is in a drift state or protection state, and the deviation is... continuous A time window below When all the concentrations of the risk gases are below their respective recovery thresholds, the system enters a recovery state. When the system is in recovery state, and continuously Deviation of each time window Persistently below Then, it returns to normal.

7. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 6, characterized in that: The drift trajectory evaluation specifically includes: Based on the current position offset vector and the current smooth drift trend vector The similarity components of location and drift trend are calculated by matching them with the reference directions of each anomaly type. The drift anomaly types include acidification drift, putrefaction drift, and other anomalous drifts. ； ； in, This indicates the direction of the current sample's position offset relative to the steady-state center and the first... Cosine similarity of the reference direction for anomalies. Indicates the direction from the steady-state reference center to the first The reference direction vector of the mean center of the anomaly type. Indicates the current smooth drift trend direction relative to the first... Cosine similarity of the reference direction for anomalies. This represents the offset vector of the current low-dimensional representation vector relative to the steady-state reference center. This represents the difference vector between the current low-dimensional representation vector and the low-dimensional representation vectors from multiple time windows prior. The comprehensive orientation matching score is calculated based on the position orientation similarity component and the drift trend orientation similarity component: ； in, Indicates the first The drift direction matching score corresponding to the reference drift direction of the anomaly type. Weighting coefficients representing positional and directional similarity. The weighting coefficients represent the similarity of the drift trend directions, and satisfy the following conditions: .

8. The biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 7, characterized in that: The differentiated control instructions are generated in the following way: Under normal conditions, the central control unit maintains the current feed flow rate, liquid dilution rate and pH target value unchanged, and does not output control intervention commands. During drift, the central control unit determines the control strategy based on the type of drift anomaly: When the abnormal drift type is acidification drift, reduce the carbon source feed flow rate and increase the dilution rate, and adjust it in conjunction with the pH control parameters; When the abnormal drift type is putrefactive drift, reduce the nitrogen source feed flow rate and maintain or adjust the supply of easily fermentable carbon sources according to the overall metabolic state, while increasing the dilution rate. If the abnormal drift type is other abnormal drift, the carbon source feed flow rate and nitrogen source feed flow rate will be adjusted. In the protection state, the central control unit adjusts the feed flow rate and liquid dilution rate, and suspends routine microbial intervention operations; In the recovery state, the central control unit gradually adjusts the feed flow rate, liquid dilution rate and pH target value according to the preset step size until the state machine returns to the normal state.

9. A biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in claim 1, characterized in that: The control and execution module includes a feed pump, a dilution and liquid discharge control module, a pH titration module, and a waste liquid collection container. The feed pump is used to adjust the feed flow rates of carbon source and nitrogen source according to differentiated control instructions, so as to control the substrate supply of the biomimetic colon culture module. The dilution and dispensing control module is used to adjust the dispensing pump speed to control the dispensing volume per unit time and adjust the liquid dilution rate to the value specified by the command. The pH titration module is used to add acid or alkali to the biomimetic colon culture module according to differentiated control instructions in order to adjust the pH to the target value. The waste liquid collection container is connected to the liquid outlet of the bionic colon culture module and is used to collect the waste liquid generated during the culture process.

10. A biomimetic colonic microecological monitoring and regulation method based on an electronic nose, characterized in that, The application of the biomimetic colonic microecological monitoring and regulation system based on an electronic nose as described in any one of claims 1-9 includes the following steps: In the biomimetic colon culture module, an anaerobic continuous flow fermentation environment is provided for the intestinal flora, and headspace gas generated in the gas phase space of the biomimetic colon culture module is continuously collected. The headspace gas is detected by a multi-channel sensor array, and a multi-dimensional response signal reflecting the current odor fingerprint is output. The multidimensional response signal is compensated and its features are extracted to obtain the characteristic representation of the current odor fingerprint; The feature representation is compared with a pre-established steady-state odor fingerprint reference model to calculate the deviation of the current micro-ecological state from the steady state, and the drift direction is comprehensively evaluated to determine the abnormal mode of the current micro-ecological state. Based on the identified abnormal patterns, differentiated control instructions are generated to adjust the substrate supply, liquid dilution rate, and pH of the biomimetic colon culture module.

Images