Coupling neural network environment DNA enrichment system based on multi-element sensing

Abstract

The invention discloses a coupled neural network environment DNA enrichment system based on multi-element sensing, and belongs to the field of ecological environment. The system integrates a multi-element sensing module and a multi-stage filtering unit, and dynamically monitors key parameters such as pressure, temperature, pH value, flow velocity, dissolved oxygen concentration and turbidity of a water sample. In order to optimize the enrichment process, an artificial intelligence algorithm-heterogeneous data fusion technology is introduced, deep learning and fusion analysis are performed on multi-element sensor data through a back propagation and long and short term memory coupled neural network, and DNA sampling efficiency of different water qualities, water areas and species is predicted. And capturing the sampling efficiency change rule of the DNA species in different years and seasons. Through heterogeneous data fusion, the system can dynamically adjust a sampling strategy, and the efficiency and reliability of DNA sampling are significantly improved. The method has remarkable technical superiority in the aspects of adapting to complex water quality conditions, reducing the filter membrane blocking risk and dynamically monitoring key parameters in the enrichment process.

Description

Technical Field

[0001] The present invention relates to the field of ecological environment, and in particular to a coupled neural network environmental DNA enrichment system based on multi-element sensing. Background Art

[0002] In recent years, with the increasing demand for ecological monitoring and environmental protection, environmental DNA (eDNA) technology has rapidly developed and become a key tool for freshwater ecological damage assessment. eDNA technology can provide high-resolution information on ecosystem status through water sample enrichment and analysis. However, the core technical bottleneck facing current practical applications of water ecological damage assessment lies in the enrichment system. On the one hand, existing enrichment systems are prone to filter clogging, especially in water samples with high suspended particle concentrations or severe pollution. This leads to slow filtration rates and low flux, which not only affects enrichment efficiency but can also affect the lifespan of the enrichment system and even cause it to crash. On the other hand, traditional enrichment systems lack real-time monitoring capabilities, making it impossible to obtain and adjust key parameters (such as pressure, temperature, flow rate, pH, oxygen concentration, and turbidity), thereby increasing operational uncertainty. These issues not only limit the applicability of existing systems to complex water samples but also hinder the further development of water ecological damage assessment technology.

[0003] The problem of filter membrane clogging has long been the main technical difficulty of environmental DNA enrichment systems, especially in water bodies with high particle concentrations. For example, for water bodies with high turbidity, high algae density, and severe pollution, the device has the following problems in the filtration and enrichment process of samples: slow filtration speed and low flux; easy occurrence of filter membrane clogging and affecting the accuracy of plankton collection in the receiving water body, which in turn affects the reliability of the test results. On the other hand, existing enrichment systems usually lack multi-sensor monitoring functions, making it difficult to obtain important parameters in the filtration process in a timely manner. For example, in the processing of complex water samples, due to the inability to monitor the changes in water sample pressure in real time, the dynamic adjustment ability of the filter membrane load state is significantly insufficient, further exacerbating the instability of the system. Therefore, combining multi-sensor technology with a multi-stage filtration structure to form an integrated system that can be adjusted in real time is the key to solving the existing bottleneck.

[0004] Based on the current technical bottleneck, the improvement direction of the environmental DNA enrichment system focuses on the deep integration of multivariate sensing technology and multi-stage filtration structure. The multi-stage filtration structure works together through multiple filter membranes of different pore sizes to achieve division of labor and cooperation. It can effectively reduce the filter membrane clogging phenomenon when processing water samples with high suspended particle concentrations, while ensuring the stability of the enrichment efficiency. The integrated application of multivariate sensing technology can also provide multivariate parameters. However, the traditional enrichment system does not make full use of artificial intelligence algorithms, lacks the ability to deeply integrate multidimensional heterogeneous data, and cannot dynamically adjust the sampling strategy, resulting in insufficient adaptability of the system when facing complex water bodies. The present invention introduces artificial intelligence algorithm-heterogeneous data fusion (BPANN and LSTM network) technology, which can effectively process static data and time series data collected by multivariate sensors, dynamically optimize the sampling strategy, and significantly improve the efficiency and accuracy of environmental DNA sampling. This environmental DNA enrichment system based on the deep integration of multivariate sensing, multi-stage filtration and artificial intelligence algorithms not only solves the clogging and monitoring problems of traditional enrichment systems, but also provides technical support for the efficient enrichment of complex water bodies, laying the foundation for the precision and scientific nature of water ecological damage assessment. Summary of the Invention

[0005] To address the above problems, the present invention proposes a coupled neural network environmental DNA enrichment system based on multi-element sensing. The system can dynamically optimize the working parameters of the multi-stage filtration device, significantly improve the efficiency and robustness of environmental DNA sampling, while taking into account energy consumption and equipment stability.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A coupled neural network environmental DNA enrichment system based on multi-sensor technology, comprising:

[0008] The multi-stage filtration device is obtained by connecting multiple filter modules in series, which is used to filter impurities in the environmental water body step by step. The pipes connecting the filter modules at each level are equipped with pumps and electric valves;

[0009] The multi-sensor unit is composed of sensors installed on the connecting pipes of each level of filtration module, which is used to collect the working parameters of the multi-stage filtration device and the environmental water parameters; the environmental water parameters include water quality parameters, fluid parameters and DNA stability parameters;

[0010] The distributed intelligent control module is composed of a sensor layer, a data transmission layer, a central control layer, and an execution layer. The multi-sensor unit serves as the sensor layer. The data collected by the sensor layer is uploaded to the central control layer via the data transmission layer. The central control layer issues instructions for controlling the power of the pump and the opening of the electric valve, and the execution layer executes the instructions.

[0011] The central control layer includes a data analysis unit and a decision-making unit. The data analysis unit is used to obtain the operating parameters of the multi-stage filtration device and the environmental water parameters, predict the DNA sampling efficiency, and use reinforcement learning to optimize the control parameters of the multi-stage filtration device to determine the sampling strategy. The decision-making unit generates instructions based on the results of the data analysis unit.

[0012] The method of predicting DNA sampling efficiency and optimizing control parameters of a multi-stage filtration device using reinforcement learning to determine a sampling strategy includes:

[0013] BPANN is used to predict the sampling efficiency with water quality parameters, fluid parameters, and DNA stability parameters as input; a reinforcement learning method is used to train a strategy model, and the sampling efficiency is introduced into the reward function. The trained strategy model can output the pump speed adjustment amplitude and the opening degree of each valve based on the working parameters of the multi-stage filtration device, the environmental water parameters and their parameter change rate, and the sampling efficiency as input; the LSTM network is used to predict the sampling efficiency in the future period with historical time series test data as input, and environmental DNA enrichment is carried out during the time period with the highest sampling efficiency in the future.

[0014] As a preferred embodiment of the present invention, the multi-stage filtration device includes three-stage filtration modules, and the last-stage filtration module is integrated with a microporous filter membrane for enriching environmental DNA; the filter mesh diameter of the first-stage filtration module is 1-2 mm, the filter mesh diameter of the second-stage filtration module is 50-100 μm, and the filter mesh diameter of the third-stage filtration module is 5-10 μm, and the lower layer of the filter mesh of the third-stage filtration module is provided with a microporous filter membrane.

[0015] As a preferred embodiment of the present invention, the multi-sensor unit arrangement scheme is specifically as follows:

[0016] The first turbidity sensor and pH sensor are arranged at the water inlet of the first-stage filtration module;

[0017] A first pressure sensor, a temperature sensor, and a second turbidity sensor are arranged on the pipeline between the first stage and the second stage;

[0018] A first flow sensor is arranged on the pipeline between the second stage and the third stage;

[0019] The water outlet of the third-stage filter module is arranged with a second flow sensor, a second pressure sensor and an oxygen concentration sensor.

[0020] As a preferred embodiment of the present invention, the water quality parameters include data measured by the first turbidity sensor, the second turbidity sensor, and the pH sensor; the fluid parameters include data measured by the first pressure sensor, the second pressure sensor, the first flow sensor, and the second flow sensor; the DNA stability parameters include data measured by the oxygen concentration sensor and the temperature sensor; and an alarm is triggered when the measured value of the second flow sensor at the water outlet of the third-stage filtration module is lower than a threshold value.

[0021] As the preferred predicted DNA sampling efficiency of the present invention, specifically:

[0022] Define sampling efficiency, which refers to the number of intact DNA fragments obtained per unit time based on the synergistic effects of water quality parameters, fluid parameters, and DNA stability parameters;

[0023] A multi-stage filtration device was used to enrich environmental DNA in different water areas and water quality conditions, and the number of intact DNA fragments obtained per unit time was calculated. Under the same water area and water quality conditions, fluid parameters were actively adjusted by adjusting the pump power and the opening of the electric valve. BPANN was trained to fit the nonlinear relationship between water quality parameters, fluid parameters, DNA stability parameters, and sampling efficiency.

[0024] The trained BPANN was used to predict the sampling efficiency with water quality parameters, fluid parameters, and DNA stability parameters as input.

[0025] As a preferred embodiment of the present invention, the control parameters of the multi-stage filtering device optimized by reinforcement learning include:

[0026] A reinforcement learning method was used to train a policy model. The environment consisted of a multi-stage filtration device and the water body to be enriched with DNA. The state space consisted of the operating parameters of the multi-stage filtration device, the parameters of the water body, and their parameter change rates. The action space consisted of the pump speed adjustment range and the opening degree of each valve.

[0027] By introducing sampling efficiency into the reward function, the trained strategy model can output the pump speed adjustment range and the opening degree of each valve based on the working parameters of the multi-stage filtration device, the environmental water parameters and their parameter change rate, and the sampling efficiency as input.

[0028] As a preferred embodiment of the present invention, the reward function is:

[0029] r t =α·ΔE-β·‖a t ‖2-γ·Ec+P

[0030] Among them, r t represents the reward, ΔE represents the improvement in sampling efficiency; ‖a t‖2 represents the action amplitude, that is, the L2 norm of the pump speed adjustment amplitude and the valve opening adjustment amount; Ec represents the pump energy consumption, P represents the over-limit penalty, and α, β, and γ are weight coefficients.

[0031] As a preferred embodiment of the present invention, the sampling strategy is determined as follows:

[0032] Obtaining time series test data of historical years, seasons, and months, and training an LSTM network to learn the time series features of the time series test data; the time series test data is environmental water parameters and sampling efficiency of target DNA species obtained in chronological order;

[0033] The trained LSTM network can use historical time series test data as input to predict the sampling efficiency in the future and enrich environmental DNA in the time period with the highest sampling efficiency in the future.

[0034] The beneficial effects of the present invention are:

[0035] The present invention designs a distributed intelligent control module combined with a multi-sensor unit to intelligently control a multi-stage filtration device. By obtaining the working parameters of the multi-stage filtration device and the environmental water parameters, the DNA sampling efficiency is predicted and reinforcement learning is used to optimize the control parameters of the multi-stage filtration device. The sampling strategy is determined, thereby achieving accurate prediction and optimization of the environmental DNA sampling efficiency. The system can adapt to complex water bodies.

[0036] More specifically, the present invention has at least the following three advantages:

[0037] Advantage 1: Compared to traditional single-stage enrichment systems, the multi-stage pre-filtration system achieves the dual advantages of high-efficiency filtration and system stability through a scientifically designed, staged filtration system and the coordinated use of precision components. In the primary 1mm pore pre-filtration stage, a larger pore size device achieves low-resistance filtration, allowing fluid flow with minimal pressure loss. This effectively removes silt and large particles from the water sample, protecting the subsequent filter units from particle impact and extending their service life. The 50μm pore size pre-filtration stage uses a retardation filtration mechanism triggered by the reduced pore size to further trap suspended particles, reducing the risk of particle clogging on the subsequent microporous membrane, improving the efficiency of the downstream enrichment module and significantly reducing the filtration burden on the microporous membrane. The filtration stage of the enrichment module consists of a 5μm filter pad and a 0.22μm organic PTFE microporous membrane. The filter pad provides final filtration of the water sample, effectively capturing any remaining larger particles, while the microporous membrane is specifically designed to enrich environmental DNA, ensuring sample purity and accuracy during DNA extraction. With its exceptional particle retention capacity, the filter pad not only effectively reduces the workload of the microporous filter membrane, but also further optimizes the overall throughput and operational stability of the filtration system, providing a crucial guarantee for the efficient operation of the enrichment module. Overall, the multi-stage pre-filtration system, through its progressively optimized filtration strategy, reduces system pressure loss and maintenance complexity, ensuring efficient and stable filtration performance and laying a solid foundation for environmental DNA enrichment and subsequent analysis.

[0038] Advantage 2: Compared to other enrichment systems, the integration of multi-sensor technology is another key feature of this invention. Based on the fundamental principles of multi-sensor technology and combining high-precision sensors with intelligent control algorithms, the multi-sensor control system demonstrates significant advantages in complex environmental monitoring and control. By integrating multiple sensor modules, including pressure, flow, temperature, pH, oxygen concentration, and turbidity, the system achieves dynamic and precise measurement of key parameters, effectively reducing measurement errors and ensuring data reliability. The built-in PID control algorithm intelligently adjusts pump flow and pressure valves based on real-time monitoring data, achieving multi-parameter coordinated optimization to ensure optimal system operation while improving filtration efficiency and system stability. In the event of abnormal parameters, the system quickly prompts user intervention through an alarm mechanism to avoid operational failures or efficiency degradation. Combined with data storage and visualization analysis modules, the system records and visually displays changing trends in key parameters, providing a scientific basis for optimizing filtration processes and tracing faults. Overall, this system, through its highly integrated precision measurement, intelligent control, and data analysis, provides strong technical support for the efficient operation and long-term reliability of environmental DNA enrichment systems.

[0039] Advantage 3: Through AI algorithms and heterogeneous data fusion technology, the system is able to deeply fuse and process multidimensional, heterogeneous data collected by multiple sensors. BPANN accurately predicts DNA sampling efficiency under different environmental conditions by learning nonlinear relationships between static data such as temperature, oxygen concentration, pH, and turbidity. The LSTM model, by processing multi-year or multi-season time series data, captures the variations in sampling efficiency of DNA species across years and seasons. By fusing the outputs of BPANN and LSTM models, the system can comprehensively consider static environmental data and time series trends, dynamically adjust sampling strategies, and significantly improve the efficiency and accuracy of environmental DNA sampling. Overall, through the high-level integration of precise measurement, intelligent control, data analysis, and AI algorithms, the system provides strong technical support for the efficient operation and long-term reliability of the environmental DNA enrichment system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart for realizing multi-stage filtering and multi-element sensors;

[0041] Figure 2 It is a working flow chart of a multi-stage filtration device;

[0042] Figure 3 It is a schematic diagram of the structure of the multi-stage filtering device and multi-sensor fusion;

[0043] Figure 4 It is a schematic diagram of a multi-stage filtration device;

[0044] Figure 5 Schematic diagram of a distributed intelligent control module;

[0045] Figure 6 This is the interface diagram of a highly integrated multi-sensor detection and control system designed based on LabVIEW software;

[0046] Figure 7 Schematic diagram of the BPANN structure;

[0047] Figure 8 Schematic diagram of LSTM network structure. DETAILED DESCRIPTION

[0048] The present invention will be further described and illustrated below in conjunction with specific embodiments. The embodiments are merely illustrative of the present disclosure and do not limit its scope. The technical features of the various embodiments of the present invention may be combined accordingly, provided that there is no conflict between them.

[0049] The accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Some of the blocks shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0050] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all steps. For example, some steps may be decomposed, while some steps may be combined or partially combined, so the actual execution order may change according to actual circumstances.

[0051] like Figure 1-3 As shown, the present invention proposes a coupled neural network environmental DNA enrichment system based on multi-element sensing, which mainly includes three parts: a multi-stage filtration device, a multi-element sensing unit, and a distributed intelligent control module. Among them, the multi-stage filtration device and the multi-element sensing unit serve as the hardware part of the system, and the distributed intelligent control module serves as the control part of the system, which work together to achieve accurate prediction and optimization of environmental DNA sampling efficiency. The system can adapt to complex water bodies.

[0052] The implementation process of this part is explained below.

[0053] (1) Multi-stage filtration device

[0054] like Figure 4 As shown, it is obtained by connecting multiple stages of filtering modules in series, which is used to filter impurities in the environmental water body step by step. Pumps and electric valves are provided on the pipelines connecting the filtering modules at each stage.

[0055] The multi-stage filtration device includes three-stage filtration modules, and the last-stage filtration module is integrated with a microporous filter membrane for enriching environmental DNA. The filter mesh diameters of each stage of filtration modules decrease successively. The filter mesh diameter of the first-stage filtration module is 1-2 mm, the filter mesh diameter of the second-stage filtration module is 50-100 μm, and the filter mesh diameter of the third-stage filtration module is 5-10 μm. A microporous filter membrane is provided at the lower layer of the filter mesh of the third-stage filtration module.

[0056] In a specific embodiment of the present invention, the design of the multi-stage filtration device is as follows:

[0057] Prepare a multi-stage filtration module consisting of a 1mm pre-filter, a 50μm pre-filter, and an enrichment module (comprising a filter pad and microporous membrane). Ensure all components are securely connected and that the equipment is leak-free. Regularly inspect the filter screen and membrane of each filtration unit to ensure they are free of damage and meet the expected filtration standards.

[0058] refer to Figure 1 , the level-by-level filtering process is as follows:

[0059] 1) 1mm pre-filtration stage: In this stage, the water sample passes through the 1mm pore pre-filtration device, which mainly relies on the mechanical screening effect of the pore size to remove larger particles such as sediment and plant debris. During the filtration process, the resistance of the fluid passing through the filter can be approximately described as:

[0060]

[0061] Where: ΔP is the pressure drop; μ is the fluid viscosity; Q is the flow rate; L is the filter thickness; d is the filter mesh aperture.

[0062] At this stage, the pore size is relatively large, and the fluid flow can basically be regarded as having no significant pressure loss (low-resistance screening filtration). It mainly removes large particles in the water sample and protects the subsequent filtration units.

[0063] 2) 50μm pre-filtration stage: The water sample is further filtered in a 50μm pore size pre-filter to intercept larger suspended matter. The filtration efficiency at this stage is generally affected by particle settling and screening. When the fluid flows through the filter, the flow state can be described by Darcy's law:

[0064]

[0065] Where: k is the permeability of the filter material; A is the filtration area; ΔP is the pressure drop; L is the thickness of the filter material.

[0066] The reduced pore size leads to increased pressure loss, and suspended solids are trapped by the filter. This retardation filtration improves the efficiency of downstream enrichment modules, ultimately reducing the burden on the microporous membrane when enriching DNA.

[0067] 3) Enrichment Module Filtration Stage: The enrichment module consists of a filter pad for absorbing and filtering 5μm particles and a 0.22μm organic PTFE microporous filter membrane. The filter pad is responsible for the final filtration of the water sample, removing residual larger particles, while the microporous filter membrane is specifically used to enrich environmental DNA, ensuring sample purity and accuracy during DNA extraction. The filter pad captures residual larger particles. Particle removal at this stage follows the filtration efficiency model:

[0068]

[0069] Where: η is the filtration efficiency; α is the adhesion coefficient of the filter material surface; C is the particle concentration; ρ is the fluid density; L is the filter material thickness.

[0070] The efficient particle retention capacity of the filter cotton not only significantly reduces the workload of the subsequent microporous filter membrane, but also optimizes the flux and stability of the entire filtration system, providing an important guarantee for the efficient operation of the enrichment module.

[0071] After the system sampling is complete, each filter module is cleaned and inspected. The 1mm pre-filter and 50μm pre-filter are backwashed to remove accumulated particles. The 5μm filter pad and 0.22μm organic microporous filter are removed and replaced after each enrichment process.

[0072] The present invention designs a highly efficient multi-stage filtration device, including a pre-filtration device with a coarse filter screen of 1 mm pore size to remove larger particulate matter such as mud and sand, a pre-filter with a fine filter screen of 50 μm pore size to effectively intercept larger suspended matter and other impurities, and a graded filtration module composed of filter cotton for adsorbing and filtering 5 μm tiny particles, which can effectively improve the enrichment flow rate and volume, so that the entire enrichment system can efficiently capture DNA fragments in the environment. Among them, the enrichment module adopts a replaceable coarse filter cotton sheet and a microporous filter membrane structure, which can effectively improve the filtration efficiency, reduce the risk of filter membrane clogging, and enhance the adaptability to water samples with high suspended particle concentrations. The above system can also be provided with a backwash function according to the requirements of those skilled in the art, which is not the focus of the present invention and will not be described here.

[0073] (2) Multi-sensor unit

[0074] It consists of sensors installed on the connecting pipes of each level of filtration modules, and is used to collect the working parameters of the multi-stage filtration device and the environmental water parameters; the environmental water parameters include water quality parameters, fluid parameters and DNA stability parameters.

[0075] The present invention introduces a multi-sensor structure, including a temperature sensor based on the principle of thermistor change, a pressure sensor based on the change of piezoresistive elements, a flow sensor based on the Doppler effect principle, an oxygen concentration sensor based on electrode redox reaction, a pH sensor based on glass electrode to measure hydrogen ion activity, and a turbidity sensor that detects the concentration of suspended particles in water through optical principles, etc., to dynamically monitor key parameters of water samples such as pressure, temperature, pH value, flow rate, dissolved oxygen concentration and turbidity.

[0076] The principles of each sensor are as follows: the temperature sensor measures the change in thermistor resistance and calculates the water temperature based on a calibration curve or formula; the pressure sensor causes the resistance change of the piezoresistive element on the diaphragm to change by measuring pressure and converts it into a pressure value; the flow sensor is based on the Doppler effect principle. The Doppler flowmeter uses the interaction between ultrasound and suspended particles in the water flow to accurately measure the water flow rate by calculating the frequency change; the oxygen concentration sensor generates current based on the redox reaction caused by the diffusion of oxygen to the electrode surface, and the sensor converts the current change into dissolved oxygen concentration; the pH sensor uses a glass electrode to measure the activity of hydrogen ions in the solution, and then measures the change in electrode potential to calculate the pH value of the water; the turbidity sensor detects the concentration of suspended particles in the water through optical principles and then measures the turbidity of the water.

[0077] In a specific implementation of the present invention, an optional sensor arrangement scheme is as follows:

[0078] The first turbidity sensor and pH sensor are arranged at the water inlet of the first-stage filtration module, wherein:

[0079] First turbidity sensor: monitors the turbidity of incoming water in real time, providing initial water quality data for subsequent filtration strategies. Its measurement range is 0-1000NTU, with an accuracy of ±5NTU.

[0080] PH sensor: detects the pH of the incoming water, with a measurement range of 0-14pH and an accuracy of ±0.05pH.

[0081] Turbidity sensors and pH sensors are installed near the water source in the water inlet pipe to measure the initial turbidity and pH value of the water sample entering the filtration system. The turbidity represents the turbidity of the external water area, and the pH value represents the acidity and alkalinity of the external water area. It is one of the important parameters for judging the water quality characteristics of different water areas.

[0082] A first pressure sensor, a temperature sensor and a second turbidity sensor are arranged on the pipeline between the first stage and the second stage, wherein:

[0083] The first pressure sensor measures the water pressure in this section of the pipeline, with a measuring range of 0-10 bar and an accuracy of ±0.05 bar.

[0084] Temperature sensor: monitors water temperature, with a measurement range of 0-50°C and an accuracy of ±0.5°C.

[0085] The second turbidity sensor measures turbidity again, comparing it to the inlet turbidity to assess the effectiveness of the first-stage filtration. Its measurement range and accuracy are the same as the inlet turbidity sensor. Installed in the pipe after the primary filtration structure, it measures the turbidity of the water flowing through the primary filter. By monitoring the turbidity after the primary filtration, we can initially assess the effectiveness of the primary filtration in removing suspended particles in the water. If the turbidity does not decrease significantly, it may indicate that the primary filter screen needs cleaning or replacement, or that the primary filter structure design needs optimization. It may also be related to valve size, providing a basis for subsequent valve adjustments.

[0086] The first flow sensor is arranged on the pipeline between the second and third stages, wherein:

[0087] First flow sensor: measures water flow, with a measurement range of 0-100L / min and an accuracy of ±1L / min.

[0088] The water outlet of the third-stage filter module is provided with a second flow sensor, a second pressure sensor and an oxygen concentration sensor, wherein:

[0089] Second flow sensor: monitors the final water outflow. When the measured value of the second flow sensor is lower than the threshold, blockage is likely to occur and the alarm is activated.

[0090] Second pressure sensor: accurately measures the outlet water pressure, with a measurement range of 0-5bar and an accuracy of ±0.01bar.

[0091] Oxygen concentration sensor: Detects dissolved oxygen concentration in water with a measurement range of 0-20 mg / L and an accuracy of ±0.1 mg / L. Installed near the microporous membrane sampling area after the tertiary filtration, the oxygen concentration here directly reflects the oxygen content in the water sample used for DNA collection. Oxygen concentration in water has a significant impact on the survival of aquatic organisms and the release of DNA. Real-time monitoring of oxygen concentration at this location provides critical data for analyzing the relationship between collected DNA and the aquatic environment.

[0092] Here, the water quality parameters in the environmental water body parameters include data measured by the first turbidity sensor, the second turbidity sensor, and the pH sensor; the fluid parameters include data measured by the first pressure sensor, the second pressure sensor, the first flow sensor, and the second flow sensor; the DNA stability parameters include data measured by the oxygen concentration sensor and the temperature sensor.

[0093] In a specific implementation of the present invention, the workflow of the multi-sensor is as follows:

[0094] Step 1: Before collecting water samples, initialize the multi-sensor module, including self-testing and calibration of temperature, pressure, flow rate, and other sensors. The system's built-in diagnostics ensure accurate sensor response and stable data output, providing a reliable monitoring foundation for subsequent enrichment processes.

[0095] Step 2: During the enrichment process, a multi-sensor module is used to monitor key parameters of the filtration unit in real time. For example, a pressure sensor based on the change in the piezoresistive element (ΔR) on the diaphragm caused by pressure P dynamically measures the pressure difference (ΔP) across the filter membrane, determining whether the filter is severely clogged and triggering an alarm mechanism. A flow sensor based on the Doppler effect utilizes the interaction between ultrasound and suspended particles in the water flow to calculate the frequency change (Δf) to monitor the flow rate (v) of the water sample through the enrichment unit in real time, ensuring operation within the designed range. A pH sensor based on the measurement of hydrogen ion activity using a glass electrode uses a glass electrode to measure the hydrogen ion activity in the solution, and then measures the electrode potential change (ΔE), ultimately achieving real-time monitoring of the water pH value to ensure the stability and extraction efficiency of the environmental DNA. Temperature, oxygen concentration, and turbidity sensors based on thermistor changes, redox reaction current detection, and optical scattering are used to measure water temperature, dissolved oxygen concentration, and suspended particle concentration, respectively.

[0096] Step 3: Real-time acquisition of sensor parameters. The system dynamically adjusts parameter changes based on preset thresholds. For example, when the pressure difference across the filter membrane exceeds a critical value, an alarm is triggered or filtration is suspended. Depending on flow rate changes, the sampling time is adjusted or the filter membrane is replaced to ensure enrichment.

[0097] Step 4: After the collection is completed, the system automatically saves the key parameter data monitored by the sensor, which facilitates the tracing and subsequent analysis of the sampling process and provides support for the extraction and detection of environmental DNA.

[0098] (3) Distributed intelligent control module based on artificial intelligence algorithm-heterogeneous data fusion technology

[0099] It consists of a sensor layer, a data transmission layer, a central control layer, and an execution layer. The sensor layer is the multi-sensor unit mentioned above. The collected data is uploaded to the central control layer through the data transmission layer. The central control layer gives instructions for controlling the power of the pump and the opening of the electric valve, and the execution layer executes the instructions. Figure 5 shown.

[0100] The central control layer includes:

[0101] Data analysis unit: used to obtain the working parameters of the multi-stage filtration device and the environmental water parameters, predict the DNA sampling efficiency and use reinforcement learning to optimize the control parameters of the multi-stage filtration device and determine the sampling strategy.

[0102] Decision-making unit: Generates instructions based on the results of the data analysis unit.

[0103] In the environmental DNA enrichment system, multiple sensors work together to provide comprehensive, multidimensional environmental data input to the AI algorithm. Temperature sensors monitor water temperature, whose changes directly affect DNA stability and species activity patterns. Oxygen sensors measure dissolved oxygen levels, reflecting the ecological health of the water. pH sensors detect water acidity and alkalinity, as changes in pH can accelerate or slow DNA degradation. Turbidity sensors assess water turbidity, which directly affects the efficiency of the enrichment system in capturing environmental DNA. Pressure sensors and flow sensors monitor water pressure and flow rate, respectively, to ensure stable operation of the sampling equipment. Data from these sensors are integrated through heterogeneous data fusion technology, enabling them to work together. For example, data from temperature and oxygen sensors can be combined to analyze water ecological activity, data from pH and turbidity sensors can be used to assess the impact of the water environment on DNA capture, and data from pressure and flow sensors ensure the stability of the enrichment system during the sampling process. Through deep learning and time series analysis using the BPANN and LSTM models, the system can dynamically adjust sampling strategies, significantly improving the efficiency and accuracy of environmental DNA sampling. The specific implementation details are as follows:

[0104] 1) Prediction of sampling efficiency based on BPANN

[0105] Sampling efficiency is defined as the number of intact DNA fragments obtained per unit time based on the synergistic effects of water quality parameters, fluid parameters, and DNA stability parameters.

[0106] Here, higher temperatures can accelerate DNA degradation, while changes in oxygen concentration and pH can affect species distribution. Turbidity directly impacts DNA capture efficiency. Through training, BPANN can output predicted values for species DNA sampling efficiency in different water qualities and water areas, providing a scientific basis for sampling strategies.

[0107] like Figure 7 As shown in the figure, the BPANN principle: The BP algorithm is a classic learning algorithm in artificial neural networks. Its structure generally includes an input layer, one or more hidden layers, and an output layer. Each layer consists of several neurons (nodes). The output value of each node is determined by the input value, activation function, and threshold. The network learning process is divided into two phases: forward propagation of information and backward propagation of error. In the forward propagation phase, input data is transmitted from the input layer through the hidden layer to the output layer. After being processed by the activation function, the output value is obtained and compared with the expected value. If there is an error, the backward propagation phase begins. The error is returned along the previous connection path, and the neuron weights are adjusted layer by layer to reduce the error. This process is repeated until the output meets the predetermined accuracy requirement.

[0108] Forward propagation stage: The output y of a neuron j in the hidden layer and output layer of BPANN is determined by the following formula:

[0109]

[0110] Where t is the neuron output, f is the activation function, ω i is the weight, x i is the input and b is the bias.

[0111] The storage information of BP network is mainly reflected in two aspects. One is the network architecture, that is, the number of nodes in the network input layer, hidden layer and output layer; the other is the connection weights between adjacent layer nodes. The main parameters affecting the network structure are the number of nodes in the hidden layer, the learning rate η and the system error ε. The number of nodes in the input layer and output layer is determined by the system application and is generally fixed, while the number of nodes in the hidden layer is determined by the user based on experience. Too few will affect the effectiveness of the network, and too many will greatly increase the network training time. The learning rate is usually between 0.01 and 0.9. Generally speaking, the smaller the learning rate, the more training times, but too large a learning rate will affect the stability of the network structure. The formulation error usually needs to be determined according to the output requirements. The lower ε is, the higher the required accuracy.

[0112] Backpropagation phase:

[0113] Error calculation: First, calculate the error between the actual output of the output layer neuron and the expected output. The error can be expressed by the square error formula:

[0114]

[0115] Among them, y j is the actual output of the output layer neurons, d j is the expected output, and m is the number of neurons in the output layer.

[0116] Error back propagation: The output layer error is back propagated to the hidden layer through the chain rule. For the weight update of the output layer neurons, the gradient is calculated as:

[0117] δ j =(y j -d j )f′(z j )

[0118] Δω ij =-ηδ j x i

[0119] Among them, δ j represents the error signal of the output layer neuron, f′(z j) is the derivative of the activation function, η is the learning rate, xix i Represents the output value of the neuron in the input layer or the previous layer.

[0120] Hidden layer error propagation: The error of the hidden layer neurons is back-propagated by the weighted error of the output layer, and the formula is:

[0121]

[0122] The corresponding weight update is:

[0123] Δω ik =-ηδ j x k

[0124] Weight update: Based on the calculated gradient information, the connection weights are updated according to the learning rate:

[0125]

[0126] Repeated iteration: This process is repeated until the error converges to a predetermined threshold or the maximum number of iterations is reached.

[0127] During the training process, a multi-stage filtration device was used to enrich environmental DNA under different water areas and water quality conditions, and the number of intact DNA fragments obtained per unit time was calculated. Under the same water area and water quality conditions, fluid parameters were actively adjusted by adjusting the pump power and the opening of the electric valve. BPANN was trained to fit the nonlinear relationship between water quality parameters, fluid parameters, DNA stability parameters, and sampling efficiency.

[0128] The trained BPANN was used to predict the sampling efficiency with water quality parameters, fluid parameters, and DNA stability parameters as input.

[0129] 2) Optimize the control parameters of the multi-stage filtration device

[0130] A reinforcement learning method is used to train a strategy model. The environment is composed of a multi-stage filtration device and the environmental water body to be enriched with DNA. The state space is composed of the operating parameters of the multi-stage filtration device, the environmental water body parameters and their parameter change rates. The action space is composed of the pump speed adjustment amplitude and the opening degree of each valve. The strategy model adopts a deep Q network.

[0131] Introducing sampling efficiency into the reward function, the reward function is designed as follows:

[0132] r t =α·ΔE-β·‖a t ‖2-γ·Ec+P

[0133] Among them, r t represents the reward, ΔE represents the improvement in sampling efficiency; ‖a t‖2 represents the action amplitude, that is, the L2 norm of the pump speed adjustment amplitude and the valve opening adjustment amount; Ec represents the pump energy consumption, P represents the over-limit penalty, and α, β, and γ are weight coefficients.

[0134] The trained strategy model can output the pump speed adjustment range and the opening of each valve based on the working parameters of the multi-stage filtration device, the environmental water parameters and their parameter change rate, and the sampling efficiency as input.

[0135] Secondly, the data analysis unit is used to determine the sampling strategy based on the LSTM network, specifically:

[0136] The hidden layer of the LSTM network is composed of long short-term memory blocks, primarily used to process time series data. By integrating sensor data for temperature, oxygen concentration, pH, turbidity, and other parameters over multiple years or seasons, it can capture the changing patterns in DNA species sampling efficiency across years and seasons. For example, time series data from temperature and oxygen sensors can reflect the impact of seasonal climate change on species activity, while changes in pH and turbidity can reveal long-term trends in the evolution of the aquatic environment. Through its memory cells, the LSTM effectively learns these time series features and predicts DNA species sampling efficiency over a specific time period in the future, providing support for long-term environmental DNA monitoring.

[0137] During the training process, time series test data of historical years, seasons, and months are obtained, and the LSTM network is trained to learn the time series features of the time series test data; the time series test data is the sampling efficiency of the target DNA species obtained in chronological order;

[0138] The trained LSTM network can use historical time series test data as input to predict the sampling efficiency in the future and enrich environmental DNA in the time period with the highest sampling efficiency in the future.

[0139] like Figure 8 As shown in Figure 1, the hidden layer structure of the LSTM network is a long short-term memory block, which consists of three control gates and a cell structure. In the figure: the rectangular box represents the memory cell, and the horizontal line on the cell conveys the cell state; f t ,i t ,o t They are forget gate, input gate and output gate respectively.

[0140] LSTM outputs h through the previous moment t-1 and the current input x t , together form the input vector [h t-1 ,x t ]Calculate the forget gate f t , to control the memory cell state:

[0141] f t=σ(W f [h t-1 ,x t ]+b f )

[0142] Among them, W f ,b f are the input layer weights and bias vectors respectively; σ(·) is the activation function, which generally uses the sigmoid function:

[0143] Next, new information that needs to be updated is generated: input gate i t , the sigmoid function is used to determine which values are used for update; the tanh layer generates new candidate values As the candidate value generated by the current layer, it is added to the memory cell:

[0144] i t =σ(W i [h t-1 ,x t ]+b i )

[0145]

[0146] Where: W C , b C are the weight and bias vector of the state update layer respectively; tanh(·) is the tanh activation function.

[0147] At the output layer, the network passes through the output gate o t Controls the output of update status:

[0148] o t =σ(W o [h t-1 ,x t ]+b o )

[0149] h t =o t tanh(C t )

[0150] Where: W o , b o are the weight and bias vector of the output layer respectively.

[0151] The Long Short-Term Memory (LSTM) neural network is trained using a time-based backpropagation algorithm, with errors propagated back through the time dimension. This training enables the network to extract features from time series data, reflecting the bearing's degradation process in the time domain. This results in more accurate predictions for time series beyond the current moment.

[0152] This invention introduces an artificial intelligence algorithm—heterogeneous data fusion technology—which integrates reinforcement learning, back-propagation neural networks (BPANN), and long short-term memory networks (LSTM) to perform deep learning and fusion analysis on multi-sensor data. BPANN uses data such as temperature, oxygen concentration, pH, and turbidity to predict DNA sampling efficiency for different water qualities, water areas, and species. Reinforcement learning optimizes the operating parameters of multi-stage filtration devices by incorporating sampling efficiency into the reward function. LSTM analyzes multi-year or multi-season time series data (such as temperature, oxygen concentration, pH, turbidity, pressure, flow, and sampling efficiency) to capture the changing patterns of DNA species sampling efficiency in different years and seasons. Through heterogeneous data fusion, the system can dynamically adjust sampling strategies, significantly improving the efficiency and reliability of DNA sampling.

[0153] This invention, based on LabVIEW-designed control logic, enables real-time acquisition and precise control of key parameters during the enrichment process, including pressure, temperature, flow rate, oxygen concentration, pH, and turbidity. This provides efficient monitoring and intelligent control technology for complex multivariable environments. The LabVIEW-based program enables real-time acquisition and precise control of key parameters such as pressure, temperature, flow rate, oxygen concentration, pH, and turbidity. The program establishes communication with various sensors and, during the initialization phase, completes interface configuration and self-calibration to ensure proper sensor operation and improve data acquisition accuracy. During operation, the system synchronously acquires multi-parameter data at optimized acquisition intervals and displays it in real time using sliders, dynamic curves, and other formats. The program incorporates built-in intelligent control and alarm logic, utilizing LabVIEW's PID control module to achieve coordinated control of pump flow and pressure valves. An indicator light illuminates when the system triggers an alarm. The program also boasts powerful data management capabilities, supporting historical data storage and real-time data trend display, providing a basis for system optimization and fault diagnosis. Overall, this program, through a modular design, integrates sensor signal acquisition, dynamic display, and intelligent control functions, providing an efficient solution for real-time monitoring and intelligent management of complex multivariable environments.

[0154] In a specific implementation of the present invention, Figure 6 As shown in the figure, this solution uses LabVIEW software to design a highly integrated multi-sensor detection and control system, designed to acquire and precisely control key parameters in the enrichment process in real time, including pressure, temperature, flow rate, oxygen concentration, pH, and turbidity. The system utilizes a modular design, integrating sensor signal acquisition, dynamic data display, and alarm linkage, providing efficient monitoring and intelligent control technology support for complex multivariable environments.

[0155] Step 1: Initialize and calibrate the multi-sensor

[0156] During program startup, LabVIEW is first used to initialize the multi-sensor module. This process includes configuring the communication interfaces for the pressure, temperature, flow, oxygen concentration, pH, and turbidity sensors and performing self-test calibration. Built-in diagnostic routines verify the operating status of each sensor to ensure proper operation. The system also performs preliminary signal calibration to ensure the accuracy and reliability of subsequent data acquisition, laying a solid foundation for subsequent real-time monitoring.

[0157] Step 2: Real-time acquisition and processing of multiple parameters

[0158] After confirming the sensors are functioning properly, the system begins real-time, synchronous acquisition and dynamic processing of the core parameters in the multi-sensor module. The acquisition interval and processing speed of each parameter are optimized based on system requirements to ensure rapid response to environmental changes.

[0159] Step 3: Real-time control and alarm logic

[0160] Using the built-in PID control module in LabVIEW, the system achieves intelligent, coordinated control of pump flow and valves. When a key parameter (such as pressure or flow) reaches or exceeds a preset threshold, the system automatically identifies it as out-of-tolerance and triggers an alarm. When an alarm occurs, a front-panel indicator lights up, allowing users to quickly identify and make adjustments, ensuring the system always operates optimally.

[0161] Step 4: Data storage and analysis

[0162] During data acquisition and control, the system records key data from multiple sensors in real time. Specific features include: Historical Data Storage: Storing pressure, temperature, flow rate, oxygen concentration, pH, and turbidity parameters as files for easy traceability and data analysis. Furthermore, Data Statistics and Visualization: The built-in statistics module analyzes collected data, displaying the changing trends of various parameters through dynamic curves, providing a basis for decision-making in subsequent filtration process optimization and troubleshooting.

[0163] Compared to traditional single-stage systems, the environmental DNA enrichment system proposed in this paper utilizes a multi-stage pre-filtration system that leverages a phased filtration design and precision components to achieve efficient filtration and stable operation. A 1mm pre-filtration filter removes large particles such as sediment, protecting subsequent units; a 50μm pre-filtration filter reduces clogging of the microporous filter membrane. In the enrichment module, a 5μm filter captures large particles, and a 0.22μm filter membrane enriches environmental DNA, ensuring purity and accuracy and optimizing system performance. This system utilizes step-by-step optimization to reduce pressure loss and maintenance difficulties, laying the foundation for DNA enrichment and analysis.

[0164] This invention integrates multi-sensor technology, fusing high-precision sensors with intelligent algorithms. The system integrates multiple sensing modules, such as pressure and flow, to accurately measure key parameters. This enables real-time optimization of pumps and pressure valves to ensure optimal operation, improving filtration efficiency and stability. Alarms are generated when parameters are abnormal, and data storage and visualization modules facilitate process optimization and fault tracing, supporting efficient and reliable system operation.

[0165] The present invention demonstrates significant technical superiority in its ability to adapt to complex water quality conditions, reduce the risk of filter membrane clogging, and dynamically monitor key parameters during the enrichment process. The present invention overcomes the limitations of traditional systems in terms of processing efficiency, sample throughput, and water quality adaptability, significantly improving the stability and reliability of the enrichment process. At the same time, the dynamic monitoring function provides support for the real-time optimization of key process parameters, ensuring efficient and accurate environmental DNA enrichment under complex environmental conditions. This technological breakthrough not only provides stronger support for ecological monitoring and environmental assessment, but also demonstrates broad application potential in the fields of environmental science and biodiversity research, opening up new directions for the development and practical application of related technologies.

[0166] Embodiments of the system of the present invention can be applied to any device with data processing capabilities, such as a computer or other device. System embodiments can be implemented through software, hardware, or a combination of software and hardware. For example, a software implementation, as a logical device, is implemented by a processor of any device with data processing capabilities, reading corresponding computer program instructions from non-volatile memory into internal memory and executing them.

[0167] The above examples are merely specific embodiments of the present invention. Obviously, the present invention is not limited to the above examples, and many variations are possible. All variations that can be directly derived or imagined by a person skilled in the art from the disclosure of the present invention should be considered to be within the scope of protection of the present invention.

Claims

1. A coupled neural network environmental DNA enrichment system based on multi-sensor, characterized in that: include: The multi-stage filtration device is obtained by connecting multiple filter modules in series, which is used to filter impurities in the environmental water body step by step. The pipes connecting the filter modules at each level are equipped with pumps and electric valves; The multi-sensor unit is composed of sensors installed on the connecting pipes of each level of filtration module, which is used to collect the working parameters of the multi-stage filtration device and the environmental water parameters; the environmental water parameters include water quality parameters, fluid parameters and DNA stability parameters; The distributed intelligent control module is composed of a sensor layer, a data transmission layer, a central control layer, and an execution layer. The multi-sensor unit serves as the sensor layer. The data collected by the sensor layer is uploaded to the central control layer via the data transmission layer. The central control layer issues instructions for controlling the power of the pump and the opening of the electric valve, and the execution layer executes the instructions. The central control layer includes a data analysis unit and a decision-making unit. The data analysis unit is used to obtain the operating parameters of the multi-stage filtration device and the environmental water parameters, predict the DNA sampling efficiency, and use reinforcement learning to optimize the control parameters of the multi-stage filtration device to determine the sampling strategy. The decision-making unit generates instructions based on the results of the data analysis unit. The method of predicting DNA sampling efficiency and optimizing control parameters of a multi-stage filtration device using reinforcement learning to determine a sampling strategy includes: Back-propagation artificial neural network was used to predict sampling efficiency with water quality parameters, fluid parameters, and DNA stability parameters as input; A reinforcement learning method is used to train a strategy model, and sampling efficiency is introduced into the reward function. The trained strategy model can output the pump speed adjustment amplitude and the opening degree of each valve based on the working parameters of the multi-stage filtration device, the environmental water parameters and their parameter change rate, and the sampling efficiency. The long short-term memory network is used to take the time series test data of historical time as input to predict the sampling efficiency in the future, and environmental DNA enrichment is carried out during the time period with the highest sampling efficiency in the future.

2. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 1, characterized in that: The multi-stage filtration device includes three-stage filtration modules, and the last-stage filtration module is integrated with a microporous filter membrane for enriching environmental DNA; the filter mesh diameter of the first-stage filtration module is 1-2 mm, the filter mesh diameter of the second-stage filtration module is 50-100 μm, and the filter mesh diameter of the third-stage filtration module is 5-10 μm, and the lower layer of the filter mesh of the third-stage filtration module is provided with a microporous filter membrane.

3. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 2, characterized in that: The multi-sensor unit arrangement scheme is specifically as follows: The first turbidity sensor and pH sensor are arranged at the water inlet of the first-stage filtration module; A first pressure sensor, a temperature sensor, and a second turbidity sensor are arranged on the pipeline between the first stage and the second stage; A first flow sensor is arranged on the pipeline between the second stage and the third stage; The water outlet of the third-stage filter module is arranged with a second flow sensor, a second pressure sensor and an oxygen concentration sensor.

4. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 3, characterized in that: The water quality parameters include data measured by the first turbidity sensor, the second turbidity sensor, and the pH sensor; the fluid parameters include data measured by the first pressure sensor, the second pressure sensor, the first flow sensor, and the second flow sensor; the DNA stability parameters include data measured by the oxygen concentration sensor and the temperature sensor; an alarm is triggered when the measurement value of the second flow sensor at the water outlet of the third-stage filtration module is lower than the threshold.

5. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 1, characterized in that: The predicted DNA sampling efficiency is specifically: Define sampling efficiency, which refers to the number of intact DNA fragments obtained per unit time based on the synergistic effects of water quality parameters, fluid parameters, and DNA stability parameters; A multi-stage filtration device was used to enrich environmental DNA in different water areas and water quality conditions, and the number of intact DNA fragments obtained per unit time was calculated. Under the same water area and water quality conditions, fluid parameters were actively adjusted by adjusting the pump power and the opening of the electric valve. BPANN was trained to fit the nonlinear relationship between water quality parameters, fluid parameters, DNA stability parameters, and sampling efficiency. The trained BPANN was used to predict the sampling efficiency with water quality parameters, fluid parameters, and DNA stability parameters as input.

6. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 1 or 5, characterized in that: The control parameters of the multi-stage filtering device optimized by reinforcement learning include: A reinforcement learning method was used to train a policy model. The environment consisted of a multi-stage filtration device and the water body to be enriched with DNA. The state space consisted of the operating parameters of the multi-stage filtration device, the parameters of the water body, and their parameter change rates. The action space consisted of the pump speed adjustment range and the opening degree of each valve. By introducing sampling efficiency into the reward function, the trained strategy model can output the pump speed adjustment range and the opening degree of each valve based on the working parameters of the multi-stage filtration device, the environmental water parameters and their parameter change rate, and the sampling efficiency as input.

7. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 6, characterized in that: The reward function is: r t =α·ΔE-β·‖a t ‖2-γ·Ec+P Among them, r t represents the reward, ΔE represents the improvement in sampling efficiency; ‖a t ‖2 represents the action amplitude, that is, the L2 norm of the pump speed adjustment amplitude and the valve opening adjustment amount; Ec represents the pump energy consumption, P represents the over-limit penalty, and α, β, and γ are weight coefficients.

8. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 1, characterized in that: The sampling strategy is determined as follows: Obtaining time series test data of historical years, seasons, and months, and training an LSTM network to learn the time series features of the time series test data; the time series test data is environmental water parameters and sampling efficiency of target DNA species obtained in chronological order; The trained LSTM network can use historical time series test data as input to predict the sampling efficiency in the future and enrich environmental DNA in the time period with the highest sampling efficiency in the future.

9. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 1, characterized in that: The LabVIEW program is used to collect and precisely control key parameters and workflows during the enrichment process in real time, including: Step 1: Initialize and calibrate the multi-sensor During the program startup phase, LabVIEW is first used to initialize the multi-sensor unit. This process includes configuring the sensor's communication interface and performing self-test calibration. The built-in diagnostic program verifies the working status of each sensor to ensure its normal operation. Step 2: Real-time acquisition and processing of multiple parameters After confirming that the sensor is operating normally, the system begins to synchronously collect and dynamically process the core parameters of the multi-sensor module in real time. The collection interval and processing speed of each parameter are optimized according to system requirements to ensure rapid response to environmental changes. Step 3: Real-time control and alarm logic The system performs intelligent linkage control on pump flow and electric valves. When the parameters reach or exceed the preset threshold, the system automatically determines that it is out of tolerance and triggers the alarm mechanism; Step 4: Data storage and analysis During the data acquisition and control process, the system records key data from multiple sensors in real time, including historical data storage, data statistics, and visualization.

10. The multi-sensor coupled neural network environmental DNA enrichment system according to claim 9, characterized in that: The LabVIEW program has a visual interface for displaying real-time sensor parameters and indicator light alarm status.

Images