Method for training film parameter optimization model, method for determining film preparation parameters, electronic device and system

Abstract

The application provides a film parameter optimization model training method, a film preparation parameter determination method, an electronic device and a system. The training comprises: linearly encoding sample bacterial cellulose composite film sample preparation parameters to obtain sample encoding data; obtaining sample film performance evaluation parameters according to updated latent state parameters obtained by performing a latent state update operation using state space parameters; and training a deep learning model according to the sample film performance evaluation parameters and sample film performance evaluation labels to obtain a film parameter optimization model.

Description

Technical Field

[0001] This application relates to the fields of biomaterials and computers, and in particular to a method for training a membrane parameter optimization model, a method for determining membrane preparation parameters, electronic devices, and systems. Background Technology

[0002] Bacterial cellulose (BC) is a natural nanofiber polymer material synthesized by microorganisms such as Acetobacter xylinum. It possesses unique advantages such as high purity, high crystallinity, excellent mechanical strength, good biocompatibility, and biodegradability. By combining bacterial cellulose with other functional materials, bacterial cellulose composite membranes that combine the intrinsic properties of bacterial cellulose with the functionality of the composite components can be prepared.

[0003] However, the preparation process of bacterial cellulose composite membranes involves a variety of parameters such as culture conditions, the method and timing of adding functional components, and composite technology, and its performance is greatly affected by the preparation process.

[0004] Therefore, there is a need for an optimization method and system for the preparation parameters of bacterial cellulose composite membranes that can reduce the subjectivity of manual recording of membrane growth status and minimize errors. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method for training a membrane parameter optimization model, a method for determining membrane preparation parameters, an apparatus, an electronic device, a storage medium, a program product, and a system.

[0006] According to one embodiment of this application, a training method for a membrane parameter optimization model is provided, comprising: linearly encoding sample preparation parameters of a bacterial cellulose composite membrane to obtain sample coded data, wherein the sample preparation parameters include at least one of sample culture time parameters, sample membrane thickness, sample acetic acid bacteria culture medium parameters, sample functional component parameters for composite modification, or sample spraying time parameters; obtaining sample membrane performance evaluation parameters based on the updated latent state parameters obtained by performing a latent state update operation using state space parameters, wherein the state space parameters are determined based on the sample coded data, and the latent state parameters are used to fuse and memorize the sample coded data to form an internal state representation of evolution, the internal state representation reflecting the inherent laws of the growth process of the first bacterial cellulose membrane used to prepare the bacterial cellulose composite membrane; and training a deep learning model based on the sample membrane performance evaluation parameters and sample membrane performance evaluation labels to obtain a membrane parameter optimization model.

[0007] According to another embodiment of this application, a method for determining membrane preparation parameters is provided, comprising: processing multiple preparation parameters of a bacterial cellulose composite membrane using a membrane parameter optimization model to obtain membrane performance evaluation parameters corresponding to each of the multiple preparation parameters; and determining a target preparation parameter from the multiple preparation parameters based on the membrane performance evaluation parameters corresponding to each of the multiple preparation parameters; wherein the membrane parameter optimization model is trained using the above-described training method.

[0008] According to another embodiment of this application, a training device for a membrane parameter optimization model is provided, comprising: an encoding module for linearly encoding sample preparation parameters of a bacterial cellulose composite membrane to obtain sample encoding data, wherein the sample preparation parameters include at least one of sample culture time parameters, sample membrane thickness, sample acetic acid bacteria culture medium parameters, sample functional component parameters for composite modification, or sample spraying time parameters; an updating module for obtaining sample membrane performance evaluation parameters based on updated latent state parameters obtained by performing a latent state update operation using state space parameters, wherein the state space parameters are determined based on the sample encoding data, and the latent state parameters are used to fuse and memorize the sample encoding data to form an internal state representation of evolution, the internal state representation reflecting the inherent laws of the growth process of the first sample bacterial cellulose membrane used to prepare the bacterial cellulose composite membrane; and a training module for training a deep learning model based on the sample membrane performance evaluation parameters and sample membrane performance evaluation labels to obtain a membrane parameter optimization model.

[0009] According to another embodiment of this application, a membrane preparation parameter determination device is provided, comprising: an acquisition module, configured to process multiple preparation parameters of a bacterial cellulose composite membrane using a membrane parameter optimization model to obtain membrane performance evaluation parameters corresponding to each of the multiple preparation parameters; and a determination module, configured to determine a target preparation parameter from the multiple preparation parameters based on the membrane performance evaluation parameters corresponding to each of the multiple preparation parameters; wherein the membrane parameter optimization model is trained using the above-described training method.

[0010] According to another embodiment of this application, an electronic device is provided, including one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above-described training method or prediction method.

[0011] According to another embodiment of this application, a bacterial cellulose composite membrane preparation parameter optimization system is provided, comprising: the electronic device described in the embodiments of this application.

[0012] Another embodiment of this application provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the method described above.

[0013] Another embodiment of this application provides a computer program product including computer-executable instructions that, when executed, implement the method described above.

[0014] According to embodiments of this application, by linearly encoding the sample preparation parameters of the bacterial cellulose composite membrane, the standardization and vectorization problems of multi-source heterogeneous process parameters (such as time, thickness, concentration, spray sequence, etc.) are solved. The set physical parameters are transformed into sample encoded data that can be directly processed by a deep learning model, providing a unified digital interface for subsequent calculations. Furthermore, by performing a latent state update operation using state-space parameters, the sample membrane performance evaluation parameters are obtained. This introduces a state-space model mechanism, enabling the model to dynamically fuse and memorize temporally sequenced sample encoded data through latent state parameters, forming a time-evolving... The internal state characterization is not simply a record of data, but reflects the intrinsic laws of the growth process of the first sample bacterial cellulose membrane. This transforms the complex biological growth dynamics into a computable and captureable mathematical model, realizing a description of the dynamic essence of the preparation process. Subsequently, a membrane parameter optimization model is trained. By utilizing the mapping relationship between evaluation parameters including process dynamic information and real performance labels, the trained membrane parameter optimization model can associate static parameters with results. This enables the obtained membrane parameter optimization model to accurately recommend the optimal combination of preparation parameters, including timing operations, from the target performance, realizing a fundamental shift from traditional experience-based trial and error to data-driven reverse design. Attached Figure Description

[0015] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0016] Figure 1 A schematic diagram of the structure of an image acquisition device according to an embodiment of the present disclosure is shown.

[0017] Figure 2 A schematic diagram of a spraying device structure according to an embodiment of the present disclosure is shown.

[0018] Figure 3 A schematic diagram of a membrane separation performance measuring device according to an embodiment of the present disclosure is shown.

[0019] Figure 4 A flowchart illustrating a method for training a membrane preparation parameter optimization model according to an embodiment of the present disclosure is shown.

[0020] Figure 5A schematic diagram illustrates the structure of a training apparatus for a membrane preparation parameter optimization model according to an embodiment of the present disclosure;

[0021] Figure 6 A schematic block diagram of a membrane preparation parameter determination apparatus according to an embodiment of the present disclosure is shown; and

[0022] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a training method for a membrane preparation parameter optimization model and / or a membrane preparation parameter determination method according to embodiments of the present disclosure;

[0023] Figure 8 A flowchart illustrating a method for determining membrane preparation parameters according to an embodiment of the present disclosure is shown. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "comprising" as used herein indicates the presence of features, steps, or operations, but does not exclude the presence or addition of one or more other features.

[0026] When using expressions such as "at least one of A, B, and C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.). When using expressions such as "at least one of A, B, or C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, or C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0027] The preparation parameters of bacterial cellulose composite membranes determine their physical properties, and these parameters can be optimized through manual experiments. For example, the growth state of the membrane (e.g., logarithmic growth phase, membrane thickness) can be manually observed and recorded. Then, nanoparticles are manually sprayed onto the bacterial cellulose membrane during the logarithmic growth phase using a handheld sprayer. Finally, the membrane is dried, and its performance and morphology are tested. However, the entire bacterial cellulose composite membrane preparation process suffers from problems such as excessive manual intervention, inaccurate recording of growth state changes, and uncontrollable nanoparticle spraying time and concentration.

[0028] Figure 1 A schematic diagram of the structure of an image acquisition device according to an embodiment of the present disclosure is shown.

[0029] like Figure 1 As shown, it includes: electronic devices and image acquisition devices;

[0030] An image acquisition device is configured to acquire a time-series image of a first sample bacterial cellulose membrane during its growth process.

[0031] The electronic device is configured to extract first sample pixel statistical value change data from the first sample time series image, and use a membrane thickness prediction model that characterizes the mapping relationship between pixel statistical value change data and membrane thickness change data to process the first sample pixel statistical value change data to obtain first sample membrane thickness change data. The first sample membrane thickness change data is used to determine the sample logarithmic growth phase of the first sample bacterial cellulose membrane, the start time of the sample logarithmic growth phase is used to determine the sample spray start time, the duration of the sample logarithmic growth phase is used to determine the number of sample sprays, and the sample spray start time, sample spray interval duration, and sample spray order are used to determine the sample spray time in the sample spray time sequence. The sample spray order is an integer greater than or equal to 1 and less than or equal to the number of sample sprays.

[0032] According to embodiments of this application, the image acquisition device can transform the membrane biological growth process, which is difficult to measure in real time, into continuously acquireable pixel time-series data and accurately calculateable membrane thickness time-series data. This makes the dynamic changes in membrane thickness a clear signal that can be processed and analyzed by electronic devices in real time. At the same time, the accurate membrane thickness data also makes the parameter optimization process more efficient and accurate.

[0033] According to embodiments of this application, the image acquisition device may include an infrared light source, a glass culture flask, a diffuse reflector, and a near-infrared camera.

[0034] Specifically, in some specific embodiments of this application, the infrared light source is near-infrared LED light (850~640nm). By using infrared light for film thickness detection, interference from visible light can be avoided, and the accuracy of film thickness detection can be improved.

[0035] In some specific embodiments of this application, a diffuse reflector is used to convert the direct directional beam emitted by the infrared LED light strip into a uniform and soft surface light source, thereby ensuring stable measurement conditions for the subsequent film thickness calculation process.

[0036] In some specific embodiments of this application, the infrared camera is a bandpass filtered near-infrared camera, which only allows a specific narrow band of infrared light matching the wavelength emitted by the infrared light source (such as an 850nm near-infrared LED) to pass through, while blocking all other light outside this band (such as visible light, ambient light, and thermal radiation noise). This ensures that the light signal received by the infrared camera comes only from the preset detection light source, rather than other interference, thus improving the accuracy of film thickness detection. The infrared camera can be connected to electronic devices via cables or other means to transmit data.

[0037] In some specific embodiments of this application, the process of the image acquisition device acquiring a time-series image of the first sample bacterial cellulose membrane during its growth process includes:

[0038] Based on the growth characteristics of bacterial cellulose membranes, near-infrared LED light (850-640nm) is obliquely incident on the side of a glass culture flask, and a diffuse reflector is placed opposite the glass culture flask. A bandpass filtered near-infrared camera is aligned with the diffuse reflector. Then, the infrared camera transmits pixel data to an electronic device to obtain the first sample time-series image of the bacterial cellulose membrane during its growth process.

[0039] According to embodiments of this application, the first sample time-series image can be time-series data showing the change of infrared light pixels transmitted through the bacterial cellulose membrane over time. The first sample membrane thickness change data refers to the data showing the change of the thickness of the bacterial cellulose membrane over time during its growth process. The first sample membrane thickness change data can be obtained using a membrane thickness prediction model that characterizes the mapping relationship between pixel statistical value change data and membrane thickness change data.

[0040] In some specific embodiments of this application, the mapping relationship between pixel statistical value change data and film thickness change data is calculated by the attenuation of light in the bacterial cellulose composite film according to the Beer-Lambert law, that is, when a beam of monochromatic light passes through the absorbing medium, the intensity of the transmitted light will decrease as the concentration and thickness of the medium increase. That is, near-infrared light penetrates the bacterial cellulose film to reach the diffuse reflector, and then is reflected back to the camera from the diffuse reflector.

[0041] According to embodiments of this application, the method for calculating film thickness by light intensity (i.e., pixel value) includes calculations using equations (1-1) to (1-8) in (1) to (8):

[0042] (1) The absorption of light by the bacterial cellulose membrane satisfies ,(1-1;In equation (1-1), Infrared incident light intensity, The thickness of the bacterial cellulose composite membrane, The absorption coefficient of the bacterial cellulose composite membrane in the 640nm-850nm range (determined experimentally). This is the measured light intensity after passing through the membrane.

[0043] (2) Since the LED is incident at an oblique angle, the effective optical path of the film for light is ,(1-2;In equation (1-2), For effective optical path.

[0044] (3) Therefore, the intensity of the transmitted light is ,(1-3;In equation (1-3), This is the measured light intensity after passing through the membrane.

[0045] (4) The camera pixel value and the received light intensity satisfy the linear imaging model. ,(1-4;In equation(1-4), P is the pixel gray value (0-255). denoted as camera system gain, and b as the level bias coefficient.

[0046] (5) can be done The average pixel value is calculated in (1-5). ).

[0047] (6) The relationship between film thickness and pixel brightness can be obtained by taking the logarithm:

[0048] , (1-6).

[0049] (7) Constant calibration method: using a known thickness ( The bacterial cellulose composite film and the average pixel value obtained from the photograph. ,pass The results can be obtained by combining (1-7) with linear fitting: , .

[0050] (8) The obtained film thickness prediction formula: , (1-8).

[0051] In some specific embodiments of this application, the film thickness prediction model is obtained through the following operations:

[0052] Extract the pixel statistical value change data of the second sample from the second sample time series image that reflects the growth process of the bacterial cellulose membrane of the second sample.

[0053] By processing the statistical value change data of the second sample pixels using a predetermined model, the thickness change data of the second sample membrane reflecting the growth process of the bacterial cellulose membrane in the second sample is obtained. The model structure of the predetermined model is constructed based on the physical law of light attenuation.

[0054] Based on the numerical optimization method, a predetermined model is trained using the actual membrane thickness change data of the second sample and the membrane thickness change data of the second sample, which reflect the growth process of the bacterial cellulose membrane of the second sample, to obtain a membrane thickness prediction model. The numerical optimization method includes at least one of the following: gradient descent method, least squares method, or ensemble learning regression method based on decision tree.

[0055] Specifically, the second sample time series image can be acquired by the aforementioned image acquisition device, thereby obtaining the second sample pixel statistical value change data.

[0056] Specifically, the actual membrane thickness change data of the second sample, reflecting the bacterial cellulose membrane growth process, refers to the actual membrane thickness change over time, which can be obtained through physical measurement. Numerical optimization methods refer to a series of algorithms used to train the model. By adjusting the parameters of the predetermined model, the predicted results (second sample membrane thickness change data) continuously approximate the actual measured data (second sample actual membrane thickness change data). These methods specifically include gradient descent, least squares, or decision tree-based ensemble learning regression methods.

[0057] According to embodiments of this application, data is automatically extracted from images, and a preliminary, physically interpretable prediction is made using a predetermined model constructed based on the physical law of light attenuation. Then, combined with experimentally measured actual thickness data, a numerical optimization method is used to train and calibrate the physical model in a data-driven manner. By combining prior physical data with machine learning, a highly adapted and accurate membrane thickness prediction model is obtained, thus enabling quantitative, non-contact, dynamic monitoring and prediction of bacterial cellulose membrane thickness growth simply by analyzing image sequences.

[0058] Further, according to embodiments of this application, the logarithmic growth phase of the sample refers to the rapid growth stage of the bacterial cellulose membrane. The start time of the sample spray is determined by the start time of the logarithmic growth phase and is used to trigger automatic spraying. The number of sample sprays is calculated based on the duration of the logarithmic growth phase, and the spraying frequency is determined by dividing time periods (e.g., 10-minute intervals). The sample spray interval is the time difference between sprays, such as 10 minutes or more subdivided intervals. The sample spray sequence represents the sequential number of the sprays, which can be from 1 to the number of sprays. The sample spray time sequence is a series of time points generated by combining the start time, interval length, and sequence. The sample spray time is a specific time point in the sequence, and is executed by the automatic spraying system controlled by electronic equipment.

[0059] According to embodiments of this application, the membrane parameter optimization system further includes a spraying device, such as... Figure 2 As shown, the spraying device also includes a drive controller and a spray actuator; the aforementioned electronic device is further configured to generate a spray control command based on the sample spraying time parameters, including the sample spraying time sequence and the duration of a single sample spray; the spraying device includes: a drive controller, communicatively connected to the electronic device, configured to generate a spray drive signal in response to the spray control command; and a spray actuator, coupled to the drive controller, configured to perform a spraying operation in response to the spray drive signal to add sample functional components to the first sample bacterial cellulose membrane to obtain a sample bacterial cellulose composite membrane.

[0060] In some specific embodiments of this application, the spraying device may further include a liquid inlet, which includes an electromagnetic stirrer and a solution for storing the composite modified nanoparticles. The composite modified nanoparticle solution includes a rotor that works in conjunction with the electromagnetic stirrer to ensure that a uniformly supplied composite modified nanoparticle solution is always available.

[0061] Specifically, the drive controller can connect the inlet and the spray actuator through a peristaltic pump, thereby controlling the working state of the sprayer by controlling the switch of the peristaltic pump, and then executing the spray operation according to the spray control command to add sample functional components to the first sample bacterial cellulose membrane to obtain the sample bacterial cellulose composite membrane.

[0062] In some specific embodiments of this application, after generating or receiving a spray control command, the electronic device in the membrane parameter optimization system can control the working state of the peristaltic pump through a Raspberry Pi microcomputer, that is, make the Raspberry Pi interface generate a high-low level switching signal, and realize the precise control of the on / off state of the peristaltic pump through a relay, thereby adding sample functional components to the first sample bacterial cellulose membrane to obtain a sample bacterial cellulose composite membrane.

[0063] According to embodiments of this application, the spraying device can convert spray control commands into precise physical actions through electronic devices, and control the spray actuator to automatically perform spraying operations through a drive controller, thereby realizing intelligent closed-loop control of the process. While improving the accuracy of process parameters, it can also ensure that multiple parallel experiments are carried out under relatively uniform experimental conditions, thereby enabling precise and efficient optimization of preparation parameters based on membrane physical properties.

[0064] According to the embodiments of this application, Figure 1 and Figure 2 The electronic devices mentioned can be the same electronic device, or two or more electronic devices capable of transmitting data; this application does not limit this.

[0065] According to the embodiments of this application, the membrane preparation parameter optimization system provided by this application constructs a closed-loop strategy of real-time perception, immediate decision-making, and precise execution. It can detect the real-time membrane thickness through image acquisition equipment and electronic equipment, and then determine the growth status (such as whether it has entered the logarithmic growth phase) in real time. At the optimal time window (logarithmic growth phase), it triggers the spraying equipment to spray through electronic equipment, realizing the dynamic and adaptive adjustment of the process parameters for preparing bacterial cellulose composite membranes during the growth process, ensuring the optimal timing and effect of composite modification. At the same time, it can realize the independent and precise control of each preparation parameter, thereby better and more accurately optimizing the preparation parameters of bacterial cellulose composite membranes.

[0066] Furthermore, according to some embodiments of this application, the membrane performance evaluation parameters obtained include at least one of the pure water flux or molecular weight cutoff of the bacterial cellulose composite membrane.

[0067] Specifically, pure water flux can reflect the permeation efficiency of the bacterial cellulose composite membrane; the molecular weight cutoff characterizes the separation precision of the membrane and clarifies its applicable separation range.

[0068] In some specific embodiments of this application, the bacterial cellulose composite membrane preparation parameter optimization system may further include a testing device for testing the physical properties of the prepared bacterial cellulose composite membrane, such as... Figure 3 As shown.

[0069] Specifically, through Figure 3 The measuring device shown measures pure water flux including:

[0070] The feed solution (200 ppm PEG / water solution) stored in the ultrafiltration cup under nitrogen pressure is permeated through the bacterial cellulose membrane to the other side. During this process, the magnetic stirring speed is controlled at 300-500 rpm to prevent contaminant deposition on the membrane surface. The filtered solution is then collected. The volume of the permeate is recorded, and the flux data is obtained using equation (3-1).

[0071] (3-1);

[0072] In equation (3-1), J w Pure water flux (L·m) -2 ·h -1 Q is the effective area A (m²) of the membrane. 2 The amount of solution that passes through within time t (h).

[0073] Specifically, through Figure 3 The measuring device shown measures the polyethylene glycol (PEG) rejection, including:

[0074] By filtering a series of PEGs with different molecular weights (10, 35, 60, 100, 200 and 300 kDa), the concentrations of feed liquid and permeate were tested using a total organic carbon analyzer. The PEG rejection rate data were obtained by equation (3-2). The molecular weight of PEG corresponding to a rejection rate of 90% was defined as the molecular weight cut-off (MWCO) of the membrane.

[0075] (3-2);

[0076] In equation (3-2), R is the PEG retention rate (%), and C p PEG concentration of the permeate, C f This represents the PEG concentration of the feed solution.

[0077] Furthermore, based on the electronic equipment, image acquisition equipment, spraying equipment, and membrane physical property measurement device in the above membrane preparation parameter optimization system, a series of preparation parameters can be obtained: culture time (t), membrane thickness (D), spraying time (tp), acetic acid bacteria culture medium ratio (r), nanoparticle concentration (p), and the pure water flux (J) and MWCO data corresponding to each set of preparation parameters, as shown in Table 1 below.

[0078] Table 1

[0079]

[0080] Furthermore, based on the preparation parameters and the measured membrane physical properties, these can be input into a computer model to obtain an optimization model for the preparation parameters of bacterial cellulose composite membranes.

[0081] According to another embodiment of this application, a method for training a membrane parameter optimization model is provided, such as... Figure 4 As shown, this includes operations S410 to S430.

[0082] In operation S410, the sample preparation parameters of the bacterial cellulose composite membrane were linearly encoded to obtain sample coded data.

[0083] In operation S420, the sample membrane performance evaluation parameters are obtained based on the updated latent state parameters obtained by performing a latent state update operation using the state space parameters.

[0084] In operation S430, a deep learning model is trained based on the sample membrane performance evaluation parameters and sample membrane performance evaluation labels to obtain the membrane parameter optimization model.

[0085] Sample preparation parameters may include at least one of the following: sample culture time parameters (duration of the culture process), sample membrane thickness, sample acetic acid bacteria culture medium parameters, sample functional component parameters for composite modification, or sample spraying time parameters (spraying duration and start time).

[0086] The state-space parameters can be determined based on the sample encoded data. The latent state parameters can be used to fuse and memorize the sample encoded data to form an internal state characterization of the evolution. The internal state characterization reflects the intrinsic laws of the growth process of the first sample bacterial cellulose membrane used to prepare the sample bacterial cellulose composite membrane.

[0087] According to embodiments of this application, by linearly encoding the sample preparation parameters of the bacterial cellulose composite membrane, the standardization and vectorization problems of multi-source heterogeneous process parameters (such as time, thickness, concentration, spray sequence, etc.) are solved. The set physical parameters are transformed into sample encoded data that can be directly processed by a deep learning model, providing a unified digital interface for subsequent calculations. Furthermore, by performing a latent state update operation using state-space parameters, the sample membrane performance evaluation parameters are obtained. This introduces a state-space model mechanism, enabling the model to dynamically fuse and memorize temporally sequenced sample encoded data through latent state parameters, forming a time-evolving... The internal state characterization is not simply a record of data, but reflects the intrinsic laws of the growth process of the first sample bacterial cellulose membrane. This transforms the complex biological growth dynamics into a computable and captureable mathematical model, realizing a description of the dynamic essence of the preparation process. Subsequently, a membrane parameter optimization model is trained. By utilizing the mapping relationship between evaluation parameters including process dynamic information and real performance labels, the trained membrane parameter optimization model can associate static parameters with results. This enables the obtained membrane parameter optimization model to accurately recommend the optimal combination of preparation parameters, including timing operations, from the target performance, realizing a fundamental shift from traditional experience-based trial and error to data-driven reverse design.

[0088] According to an embodiment of this application, in operation S41, the bacterial cellulose composite membrane (or simply composite membrane) is a high-performance thin film composed of a natural nanocellulose network produced by microbial fermentation (such as *Acetobacter xylinum*) and nanoparticles used for composite modification. Preparation parameters refer to key process condition variables that are artificially set or measured during the preparation of the bacterial cellulose composite membrane; these variables collectively determine the physical properties of the final composite membrane. Linear encoding refers to mapping the original preparation parameters to a higher-dimensional, dimensionally unified latent space vector through linear transformation, thereby enhancing the data feature representation capability of the preparation parameters to make the data conform to the input requirements of the model. Sample encoded data refers to the numerical matrix or vector generated after linear encoding of the sample preparation parameters, which is more suitable for model input and understanding.

[0089] Specifically, the cultivation time parameter refers to the total time for the bacterial cellulose membrane to undergo static or dynamic fermentation in the culture medium. Membrane thickness refers to the physical thickness of the prepared bacterial cellulose composite membrane, with units such as cm, mm, or μm. Acetobacter culture medium parameters refer to the proportions of various components (such as carbon source, nitrogen source, and inorganic salts) in the acetic acid bacteria culture medium, which affect the nutrient environment during the bacterial cellulose composite membrane cultivation process. Optimized proportions provide the best nutrient environment for the membrane-producing bacteria, thereby obtaining high-yield and high-quality bacterial cellulose composite membranes. The sample functional component parameter used for composite modification refers to the concentration of nanoparticles used for composite modification. This refers to the content of nanomaterials added to the reaction system to impart specific functions (such as antibacterial properties, enhanced mechanical properties, and enhanced separation performance) to the bacterial cellulose composite membrane. The concentration directly affects functionality (such as antibacterial properties) and final performance. The spray time parameter refers to the duration of the spraying process when adding nanoparticles via spraying for composite modification, such as 1 min or 5 min. This parameter affects the uniformity of distribution and loading of modified nanoparticles in the composite membrane.

[0090] In some specific embodiments of this application, the nanoparticles used for composite modification can be metal and metal oxide nanoparticles, such as nano silver (Ag), nano zinc oxide (ZnO), and nano titanium dioxide (TiO2); they can also be inorganic reinforced nanoparticles, such as silicon dioxide (SiO2), montmorillonite, and nano hydroxyapatite; they can also be carbon-based nanomaterials, such as graphene, graphene oxide, and carbon nanotubes; and they can also be organic polymer nanoparticles, such as chitosan nanoparticles and polylactic acid nanoparticles.

[0091] In some specific embodiments of this application, the acetic acid bacteria culture medium ratio refers to the ratio of xylitol bacterial culture medium to nutrient solution. The nutrient solution includes water, black tea and glucose, and the mass ratio of the three is 10:1:0.1.

[0092] According to an embodiment of this application, in operation S42, the state space parameters refer to a set of mathematical parameters used to define the dynamic model of the system, determined based on sample encoded data. The latent state parameters refer to variables that dynamically change within the model and are used to fuse and remember the input sample encoded data. Performing a latent state update operation refers to calculating and refreshing the latent state parameters according to the rules defined by the state space parameters, combined with new input data. Specifically, the updated latent state parameters refer to the new internal state variables obtained after performing a latent state update operation. The internal state representation refers to the system dynamic information specifically carried and expressed by the latent state parameters or the updated latent state parameters, specifically referring to the inherent laws of the growth process of the first sample bacterial cellulose membrane, such as the inherent dynamic growth characteristics in the biosynthesis stage of the membrane, for example, typical microbial growth curves can be divided into a lag phase, a logarithmic growth phase, and a stationary phase. Obtaining sample membrane performance evaluation parameters refers to calculating based on the updated latent state parameters, thereby outputting predicted values ​​for membrane performance (such as flux and rejection rate). Sample membrane performance evaluation parameters refer to the specific performance prediction data generated by the operation of obtaining sample membrane performance evaluation parameters.

[0093] According to an embodiment of this application, in operation S43, the sample membrane performance evaluation label refers to the actual performance data of the membrane sample corresponding to the sample membrane performance evaluation parameters, obtained through actual measurement using laboratory instruments. For example, it could be the measured pure water flux value, or the membrane flux (L / (m²)). 2 Data such as (·h) and polyethylene glycol rejection (kDa) can characterize the physical properties of bacterial cellulose composite membranes and can be adjusted according to actual needs; this application does not impose any limitations on this. Training a deep learning model refers to the process of automatically adjusting the internal weights of the model using sample membrane performance evaluation parameters and sample membrane performance evaluation labels as paired data through an optimization algorithm.

[0094] Specifically, in step S41, the sample preparation parameters of the bacterial cellulose composite membrane are linearly encoded to obtain sample encoded data, including operations S411 to S412:

[0095] Operation S411: Standardize the sample preparation parameters of the bacterial cellulose composite membrane to obtain the characteristic data of the sample preparation parameters;

[0096] Operation S412: Perform a linear transformation on the sample preparation parameter feature data using linear transformation parameters to obtain sample encoded data. The linear transformation data includes weight parameters and bias parameters. The weight parameters are used to determine the linear mapping relationship from the input feature space to the high-dimensional encoding space, and the bias parameters are used to provide a reference offset for the dimension of the high-dimensional encoding space.

[0097] According to an embodiment of this application, in operation S411, standardization processing refers to a data preprocessing operation used to eliminate interference caused by different units (dimensions) (e.g., cultivation time in hours, nanoparticle concentration in g / L) and differences in numerical ranges of different parameters on model calculations. For example, the processed data is converted into a form with a mean of 0, a standard deviation of 1, or other uniform scale. For instance, the average value of all cultivation time parameters can be subtracted from the average value and then divided by the standard deviation, so that the data are distributed around 0 and have a consistent scale.

[0098] According to an embodiment of this application, in operation S412, the linear transformation parameters refer to a specific set of mathematical coefficients used to perform the linear transformation, including weight parameters and bias parameters. The weight parameters are coefficients in the linear transformation parameters used to scale and combine the sample preparation parameter feature data; for example, they can be a matrix where the magnitude of a certain value determines the importance of the culture time feature in generating sample coding data. The bias parameters are coefficients in the linear transformation parameters used to add a fixed reference value to each dimension after the transformation; for example, they can be a vector, whose function is to ensure that the generated sample coding data is not all crowded near the origin. Linear transformation refers to a mathematical operation performed on the sample preparation parameter feature data using linear transformation parameters (weight parameters and bias parameters).

[0099] In some specific embodiments of this application, the state space parameters include state transition parameters and input projection parameters; in step S42, the sample membrane performance evaluation parameters are obtained based on the updated latent state parameters obtained by performing a latent state update operation using the state space parameters, including operations S421 to S424:

[0100] Operation S421: Based on the sample encoded data at the current time and the input projection parameters at the current time, obtain the input driving component at the current time, wherein the input driving component at the current time represents the new part of the sample encoded data at the current time to the latent state parameters at the current time that are being formed.

[0101] Operation S422: Based on the latent state parameters of the previous time step and the state transition parameters of the current time step, obtain the state recursion component of the current time step, wherein the state recursion component of the current time step represents the retained part of the historical information carried by the latent state parameters of the previous time step after attenuation and linear transformation by the state transition parameters of the current time step;

[0102] Operation S423: Based on the input driving component and the state recursion component at the current time, obtain the potential state parameters at the current time as updated potential state parameters;

[0103] Operation S424: Obtain the sample membrane performance evaluation parameters based on the updated potential state parameters.

[0104] According to an embodiment of this application, in operation S421, the input projection parameter refers to the weight parameters in the state space parameters specifically used to process the current input data. For example, it can be a weight matrix, which maps the sample encoded data to the state space. The input driving component refers to the result calculated from the sample encoded data at the current moment through the input projection parameter, representing the new influence of the current input information on the state. For example, it can be a vector representing the new changes brought about by external stimuli (such as a spray event) at the current moment. The latent state parameter refers to the dynamic variables within the state space model used to remember and fuse historical information. For example, it can be a hidden state vector whose value is updated recursively over time, carrying the memory of the membrane growth process.

[0105] According to an embodiment of this application, in operation S422, the input projection parameter is a parameter in the state space specifically used to map external input data to the state space. For example, it is a matrix responsible for converting the current sample data into a driving signal for the internal state. The sample encoded data at the current time is the preparation parameter data after preprocessing and encoding at the current time point. For example, it is a numerical vector representing parameters such as culture time and concentration at t=48 hours.

[0106] According to an embodiment of this application, in operation S423, the latent state update operation is the core step of iteratively calculating the internal state based on the state space parameters. For example, it is the mathematical process of refreshing the model's internal memory at each time point by combining new inputs and historical memories.

[0107] According to embodiments of this application, by explicitly distinguishing state-space parameters into state transition parameters and input projection parameters, decoupling and fine control of the influence of historical memory and current input during membrane growth are achieved. Specifically, the current sample coding data and input projection parameters are used to generate input driving components, accurately characterizing the immediate influence of current process conditions on the membrane state. At the same time, the potential state parameters and state transition parameters from the previous moment are used to generate state recursive components, effectively simulating and preserving the decay and evolution laws of membrane growth history information. Finally, by integrating the input driving components and state recursive components, updated potential state parameters are obtained, enabling the membrane parameter optimization model to more accurately and dynamically reflect the intrinsic continuity of the bacterial cellulose membrane growth process. Thus, the sample membrane performance evaluation parameters obtained based on these parameters have higher prediction accuracy and physical interpretability, providing a reliable state basis for the reverse optimization of subsequent preparation parameters.

[0108] In some specific embodiments of this application, the training method for the membrane preparation parameter optimization model can be as follows:

[0109] (1) Data collection

[0110] Suppose there are n sets of experimental data, numbered 1, 2, ..., n.

[0111] The input vector is defined by the following equation (4-1):

[0112] ,(4-1;

[0113] In equation (4-1), d is the input dimension.

[0114] The output is defined as a 2D vector as shown in equation (4-2):

[0115] (4-2);

[0116] In equation (4-2), The pure water flux of the composite membrane prepared with the i-th set of parameters is given. Let be the amount of polyethylene glycol retained in the composite membrane prepared using the i-th set of parameters.

[0117] To ensure numerical stability, the input and output of each dimension are first standardized as follows (4-3):

[0118] (4-3);

[0119] In equation (4-3), The mean of the input data. The standard deviation of the input data. The mean of the output data. This represents the standard deviation of the output data.

[0120] (2) The selective state-space model is used as the basic model to model the temporal characteristics and performance indicators of the film-making process.

[0121] The basic building block of Mamba is the selective state space layer, which is mathematically expressed as (4-4) and (4-5).

[0122] ,(4-4;

[0123] (4-5);

[0124] In equations (4-4) and (4-5), Given a one-dimensional input sequence, As a potential state, This is the output. This is the state transition matrix. Given the input transformation matrix, The output transformation matrix consists of learnable parameters.

[0125] (3) Linear encoding of multidimensional inputs

[0126] After numerical standardization, the multidimensional inputs are linearly encoded to fit Mamba's selective state-space model structure. For each input... Define a linear mapping to change the input dimension from The system is elevated to the latent space dimension N to drive the state-space dynamic system, as shown in equation (4-6):

[0127] (4-6);

[0128] In equation (4-6), This is the weight matrix. For bias vectors, For the input vector, This is the output after linear transformation.

[0129] The final encoded sequence { This will be used as input to the selective state space module for state updates.

[0130] (4) Output computation layer

[0131] The final output obtained by passing the data into the Mamba algorithm is denoted as: .

[0132] (5) Define the loss function

[0133] Using MSE loss: All parameters are trained through backpropagation.

[0134] (6) Destandardized output results

[0135] The model predicts a standardized output. Then, the true value is restored through destandardization.

[0136] (7) Obtain the membrane parameter optimization model

[0137] Using the trained Mamba model For any input recipe parameters: .

[0138] The results obtained from the standardized input model Then, by inverse standardization, we can obtain the mapping model from membrane preparation parameters to membrane performance indicators: .

[0139] According to another embodiment of this application, a method for determining membrane preparation parameters, wherein, as Figure 8 As shown, operations S210 to S220 are included:

[0140] Operation S210: Multiple preparation parameters of the bacterial cellulose composite membrane are processed using a membrane parameter optimization model to obtain membrane performance evaluation parameters corresponding to each of the multiple preparation parameters; and

[0141] Operation S220: Determine the target preparation parameter from the multiple preparation parameters based on the membrane performance evaluation parameters corresponding to each of the multiple preparation parameters.

[0142] According to embodiments of this disclosure, a membrane preparation parameter optimization model can be used to obtain predicted membrane physical property data of bacterial cellulose composite membranes based on any combination of input parameters. This reduces the time and economic costs of manual experiments, provides quantitative and predictable decision support, changes the model that relies on experience and intuition, and makes the process development process more precise, scientific, and efficient.

[0143] Based on the training method of the membrane preparation parameter optimization model described above, this disclosure also provides a training device for the membrane preparation parameter optimization model. The following will be combined with... Figure 5 The device is described in detail.

[0144] Figure 5 A schematic block diagram of a training apparatus for a parameter optimization model for bacterial cellulose composite membrane preparation according to an embodiment of the present disclosure is shown.

[0145] like Figure 5 As shown, the training device 500 for the bacterial cellulose composite membrane preparation parameter optimization model in this embodiment includes an encoding module 510, an update module 520, and a training module 530.

[0146] The encoding module 510 is used to linearly encode the sample preparation parameters of the bacterial cellulose composite membrane to obtain sample coded data. The sample preparation parameters include at least one of the following: sample culture time parameters, sample membrane thickness, sample acetic acid bacteria culture medium parameters, sample functional component parameters for composite modification, or sample spraying time parameters. In one embodiment, the encoding module 510 can be used to perform the operation S41 described above, which will not be repeated here.

[0147] The update module 520 is used to obtain sample membrane performance evaluation parameters based on the updated latent state parameters obtained by performing a latent state update operation using the state space parameters. The state space parameters are determined based on the sample coding data, and the latent state parameters are used to fuse and memorize the sample coding data to form an internal state representation of the evolution. This internal state representation reflects the inherent laws governing the growth process of the first sample bacterial cellulose membrane used to prepare the sample bacterial cellulose composite membrane. In one embodiment, the update module 520 can be used to perform the operation S42 described above, which will not be repeated here.

[0148] The training module 530 is used to train a deep learning model based on the sample membrane performance evaluation parameters and sample membrane performance evaluation labels to obtain a membrane parameter optimization model. In one embodiment, the training module 530 can be used to perform the operation S43 described above, which will not be repeated here.

[0149] According to embodiments of this disclosure, any plurality of modules among the encoding module 510, update module 520, and training module 530 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of this disclosure, at least one of the encoding module 510, update module 520, and training module 530 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the encoding module 510, update module 520, and training module 530 may be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0150] Based on the above-mentioned method for predicting the membrane performance indicators of bacterial cellulose composite membranes, this disclosure also provides a device for predicting the membrane performance indicators of bacterial cellulose composite membranes. The following will be combined with... Figure 6 The device is described in detail.

[0151] Figure 6 A schematic block diagram of a membrane performance index prediction device for bacterial cellulose composite membranes according to an embodiment of the present disclosure is shown.

[0152] like Figure 6 As shown, the membrane performance index prediction device 600 for bacterial cellulose composite membrane in this embodiment includes an acquisition module 610 and a determination module 620.

[0153] The acquisition module 610 is used to process multiple preparation parameters of the bacterial cellulose composite membrane using a membrane parameter optimization model to obtain membrane performance evaluation parameters corresponding to each of the multiple preparation parameters. In one embodiment, the input module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0154] The determining module 620 is used to determine the target preparation parameter from multiple preparation parameters based on the membrane performance evaluation parameters corresponding to each of the multiple preparation parameters. In one embodiment, the processing module 620 can be used to perform the operation S220 described above, which will not be repeated here.

[0155] According to embodiments of this disclosure, a membrane preparation parameter optimization model can be used to obtain predicted membrane physical property data of bacterial cellulose composite membranes based on any combination of input parameters. This reduces the time and economic costs of manual experiments, provides quantitative and predictable decision support, changes the model that relies on experience and intuition, and makes the process development process more precise, scientific, and efficient.

[0156] According to embodiments of this disclosure, any plurality of modules in the obtaining module 610 and the determining module 620 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of this disclosure, at least one of the obtaining module 610 and the determining module 620 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the input module 610 and the processing module 620 may be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0157] Figure 7 A block diagram schematically illustrates an electronic device suitable for implementing a training method for a membrane parameter optimization model or a method for determining membrane preparation parameters according to embodiments of the present disclosure.

[0158] like Figure 7As shown, an electronic device 700 according to an embodiment of the present disclosure includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. The processor 701 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 701 may also include onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0159] RAM 703 stores various programs and data required for the operation of electronic device 700. Processor 701, ROM 702, and RAM 703 are interconnected via bus 704. Processor 701 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 702 and / or RAM 703. It should be noted that programs may also be stored in one or more memories other than ROM 702 and RAM 703. Processor 701 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in one or more memories.

[0160] According to embodiments of this disclosure, the electronic device 700 may further include an input / output (I / O) interface 705, which is also connected to a bus 704. The electronic device 700 may also include one or more of the following components connected to the input / output (I / O) interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 707 including a network interface card such as a LAN card, modem, etc. The communication section 707 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output (I / O) interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0161] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A training method for a membrane parameter optimization model, comprising: The sample preparation parameters of the bacterial cellulose composite membrane are linearly encoded to obtain sample coded data. The sample preparation parameters include at least one of the following: sample culture time parameter, sample membrane thickness, sample acetic acid bacteria culture medium parameter, sample functional component parameter for composite modification, or sample spraying time parameter. Based on the updated latent state parameters obtained by performing a latent state update operation using state space parameters, sample membrane performance evaluation parameters are obtained. The state space parameters are determined based on the sample encoded data. The latent state parameters are used to fuse and memorize the sample encoded data to form an internal state characterization of its evolution. This internal state characterization reflects the intrinsic laws governing the growth process of the first sample bacterial cellulose membrane used to prepare the sample bacterial cellulose composite membrane. A deep learning model is trained based on the sample membrane performance evaluation parameters and sample membrane performance evaluation labels to obtain a membrane parameter optimization model.

2. The method according to claim 1, wherein, The sample preparation parameters of the bacterial cellulose composite membrane are linearly encoded to obtain sample coded data, including: The sample preparation parameters of the bacterial cellulose composite membrane were standardized to obtain characteristic data of the sample preparation parameters; and The sample preparation parameter feature data is linearly transformed using linear transformation parameters to obtain the sample encoding data. The linear transformation data includes weight parameters and bias parameters. The weight parameters are used to determine the linear mapping relationship from the input feature space to the high-dimensional encoding space, and the bias parameters are used to provide a reference offset for the dimension of the high-dimensional encoding space.

3. The method according to claim 1 or 2, wherein, The state space parameters include state transition parameters and input projection parameters; The step of obtaining sample membrane performance evaluation parameters based on the updated latent state parameters obtained by performing a latent state update operation using state space parameters includes: Based on the sample encoded data and the input projection parameters at the current moment, the input driving component at the current moment is obtained, wherein the input driving component at the current moment represents the new part of the sample encoded data at the current moment to the potential state parameters at the current moment that are being formed. Based on the potential state parameters of the previous time step and the state transition parameters of the current time step, the state recursive component of the current time step is obtained. The state recursive component of the current time step represents the retained part of the historical information carried by the potential state parameters of the previous time step after attenuation and linear transformation by the state transition parameters of the current time step. Based on the input driving component and the state recursion component at the current moment, the potential state parameters at the current moment, which are then updated as potential state parameters, are obtained; and The sample membrane performance evaluation parameters are obtained based on the updated potential state parameters.

4. The method according to any one of claims 1 to 3, wherein, The sample spraying time parameters include a sample spraying time sequence and the duration of a single sample spraying. The sample spraying time in the sample spraying time sequence is determined based on the sample spraying start time, the sample spraying interval duration, and the sample spraying order. The sample spraying order is an integer greater than or equal to 1 and less than or equal to the number of sample sprays. The sample spraying start time is determined based on the start time of the logarithmic growth phase of the first sample bacterial cellulose membrane. The number of sample sprays is determined based on the duration of the logarithmic growth phase and the sample spraying interval duration. The bacterial cellulose composite membrane was prepared by adding the sample functional components to the first bacterial cellulose membrane by spraying during the logarithmic growth phase of the sample, according to the sample spraying time sequence and the duration of a single sample spraying. The logarithmic growth phase of the sample is determined based on the first sample membrane thickness change data of the first sample bacterial cellulose membrane. The first sample membrane thickness change data is obtained by processing the first sample pixel statistical value change data using a membrane thickness prediction model that characterizes the mapping relationship between pixel statistical value change data and membrane thickness change data. The first sample pixel statistical value change data is extracted from the first sample time series image that reflects the growth process of the first sample bacterial cellulose membrane.

5. The method according to claim 4, wherein, The film thickness prediction model was obtained in the following way: Extract the pixel statistical value change data of the second sample from the time series image of the second sample, which reflects the growth process of the bacterial cellulose membrane in the second sample; The pixel statistical value change data of the second sample is processed using a predetermined model to obtain the membrane thickness change data of the second sample, reflecting the bacterial cellulose membrane growth process of the second sample. The predetermined model's structure is constructed based on the physical law of light attenuation. Based on the numerical optimization method, the predetermined model is trained using the actual membrane thickness change data of the second sample and the membrane thickness change data of the second sample, which reflect the growth process of the bacterial cellulose membrane of the second sample, to obtain the membrane thickness prediction model. The numerical optimization method includes at least one of the following: gradient descent method, least squares method, or decision tree-based ensemble learning regression method.

6. A method for determining membrane preparation parameters, comprising: Multiple preparation parameters of bacterial cellulose composite membranes were processed using a membrane parameter optimization model to obtain membrane performance evaluation parameters corresponding to each of the multiple preparation parameters. as well as The target preparation parameter is determined from the plurality of preparation parameters based on the membrane performance evaluation parameters corresponding to each of the plurality of preparation parameters; The membrane parameter optimization model is trained using the method described in any one of claims 1 to 5.

7. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 5.

8. A membrane parameter optimization system, comprising the electronic device according to claim 7.

9. The system according to claim 8, wherein, The electronic device is further configured to generate spray control commands based on the sample spray time parameters, including the sample spray time sequence and the duration of a single sample spray. The system also includes: Spraying equipment, including: A drive controller, communicatively connected to the electronic device, is configured to generate a spray drive signal in response to the spray control command; and A spray actuator, coupled to the drive controller, is configured to perform a spraying operation in response to the spray drive signal to add sample functional components to a first sample bacterial cellulose membrane, thereby obtaining a sample bacterial cellulose composite membrane.

10. The system according to claim 9, further comprising: An image acquisition device is configured to acquire a time-series image of a first sample bacterial cellulose membrane during its growth process. The electronic device is further configured to extract first sample pixel statistical value change data from the first sample time series image, and process the first sample pixel statistical value change data using a membrane thickness prediction model that characterizes the mapping relationship between pixel statistical value change data and membrane thickness change data to obtain first sample membrane thickness change data. The first sample membrane thickness change data is used to determine the sample logarithmic growth phase of the first sample bacterial cellulose membrane. The start time of the sample logarithmic growth phase is used to determine the sample spray start time. The duration of the sample logarithmic growth phase is used to determine the number of sample sprays. The sample spray start time, sample spray interval duration, and sample spray order are used to determine the sample spray time in the sample spray time sequence. The sample spray order is an integer greater than or equal to 1 and less than or equal to the number of sample sprays.

Images