PVA film gas permeability intelligent detection method based on machine learning

Abstract

The invention discloses a PVA film gas permeability intelligent detection method based on machine learning, and relates to the technical field of material testing. The method comprises the following steps: acquiring experimental environment data, first detection data, first diffusion theory characteristic data and second detection data, and constructing and training a neural network model; setting application environment parameters, obtaining third detection data, and calculating second diffusion theory characteristic data according to the application environment parameters and the third detection data; and inputting the application environment parameters, the third detection data and the second diffusion theory characteristic data into a neural network model to obtain a gas permeability prediction value. According to the method, the neural network model is constructed and trained, and the gas permeability is predicted through the neural network model, the application environment parameters, the third detection data and the second diffusion theory characteristic data, so that the dependence on a differential pressure method gas permeability experiment is reduced, the time cost of material testing and experiment is reduced, and the test efficiency is improved. And meanwhile, online monitoring and adjustment of the production process are facilitated, and production benefits are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of material testing, and in particular to an intelligent detection method for gas permeability of a PVA film based on machine learning. Background Art

[0002] PVA is a water-soluble polymer with good film-forming properties, transparency and mechanical properties. It is also biocompatible and biodegradable. Based on ordinary PVA film, a PVA porous film can be made by forming a large number of pores inside it through a specific process. The presence of pores in the PVA porous film gives it a higher specific surface area and better air permeability, making it suitable for some application scenarios with specific requirements. For example, PVA porous film can be used for food preservation packaging. The selective gas permeability of the porous film can be used to control the proportion of gas components in the package and extend the shelf life of the food; or, in industrial emissions, it can effectively separate components such as carbon dioxide and nitrogen from flue gas.

[0003] Different application scenarios have different requirements for the gas permeability of PVA porous membranes. Through testing, the performance of the membrane in actual applications can be verified and data support can be provided for product improvement. In order to ensure that the gas permeability of the product meets the needs of actual use scenarios, it is often necessary to sample and test each production batch of products. The differential pressure method is a common method for determining gas permeability, but the current differential pressure method detection equipment has a long detection cycle each time. For production environments with high-frequency detection requirements, the detection efficiency of existing technologies is difficult to meet the needs. Summary of the invention

[0004] The present invention provides a method for intelligently detecting gas permeability of a PVA film based on machine learning to solve at least one of the problems mentioned in the above background technology.

[0005] The specific technical solutions provided by this application are as follows: A method for intelligently detecting gas permeability of a PVA film based on machine learning, comprising the steps of: Acquire experimental environment data, first detection data, first diffusion theory characteristic data and second detection data, and construct and train a neural network model; the input layer of the neural network model is the experimental environment data, the first detection data and the first diffusion theory characteristic data, and the output layer is the second detection data; the first diffusion theory characteristic data includes the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow rate; the experimental environment data includes gas concentration, high-pressure side pressure, low-pressure side pressure, ambient temperature and ambient humidity; the first detection data includes membrane thickness, air permeability area, pore parameter group, first wave peak intensity and surface roughness; the second detection data includes a gas permeability parameter group; Setting application environment parameters, acquiring third detection data, and calculating second diffusion theory characteristic data according to the application environment parameters and the third detection data; The application environment parameters, the third detection data and the second diffusion theory characteristic data are input into the neural network model to obtain the gas permeability prediction value.

[0006] As a preferred embodiment, the pore parameter group includes porosity, maximum pore size, first average pore size, second average pore size and third average pore size; the first average pore size represents the average diameter of all pores in the membrane; the second average pore size represents the average diameter of all pores in the mold that are less than or equal to the mean free path of the test gas; the third average pore size represents the average diameter of all pores in the mold that are greater than the mean free path of the test gas.

[0007] As a preferred solution, the theoretical value of the Knudsen diffusion coefficient is expressed as:

[0008] in, is the second average pore size, R is the gas constant, T is the ambient temperature, and M is the gas molecular weight.

[0009] As a preferred solution, the theoretical volume flow rate is expressed as:

[0010] in, is the third average pore size, T is the ambient temperature, is the set standard ambient temperature, is the high pressure side pressure of the pressure difference method, is the low-pressure side pressure of the pressure difference method, is the set standard ambient pressure; It is the gas molecular viscosity, which is set according to the specific test gas.

[0011] As a preferred solution, the training of the neural network model comprises the steps of: The experimental environment data, the first detection data, the first diffusion theory characteristic data and the second detection data are set as a sample data set; Set the initial weights and biases of the neural network model using a sample data set and a particle swarm optimization algorithm; Dividing the sample data set into a training set and a validation set; using the training set to train the neural network model through a back propagation algorithm, and minimizing the prediction error by adjusting weights and biases; Use the validation set data to obtain the performance evaluation index of the neural network model. If the performance evaluation index is higher than the set threshold, the training is terminated.

[0012] As a preferred solution, the initial weights and biases of the neural network model are set by using a sample data set and a particle swarm optimization algorithm, including the steps of: A group of particles are randomly generated, and the position and speed of each particle are initialized; the position of the particle represents the weight and bias of a group of neural network models; the speed of the particle represents the application environment parameters of each batch of products, the third detection data and the second diffusion theory characteristic data moving direction and step length; Construct the loss function of the neural network and set the loss function as the fitness function; Get the fitness function value corresponding to each particle; Update the speed and position of each particle based on the individual best position and the group best position; Iterate until the maximum number of iterations or the fitness function value meets the output condition, and output the optimal position of the particle swarm as the initial weight and bias of the neural network model.

[0013] As a preferred solution, the velocity of the particle is expressed as: , The position of a particle is expressed as: , in, is the inertia weight; and They are individual learning factor and group learning factor respectively; is the individual best position of a particle, and g represents the group best position of all particles. and are all random numbers in the interval [0,1], used to provide randomness for each iteration; i is the number of the particle, represents the position of the i-th particle at time t, represents the velocity of the i-th particle at time t.

[0014] As a preferred solution, the obtaining of the third detection data includes: Use a thickness gauge to measure the actual thickness of the PVA porous film; Set the ventilation area according to the testing equipment; Using a scanning electron microscope to obtain the porosity, the maximum pore size, the first average pore size, the second average pore size, and the third average pore size; A Fourier transform infrared spectrometer is used to measure the first peak intensity at a set wavelength; The surface roughness of the samples was measured using an atomic force microscope.

[0015] As a preferred solution, the steps are also included: A differential pressure gas permeability experiment is set up according to the application environment parameters to obtain the actual value of the gas permeability; the neural network model is optimized according to the predicted value of the gas permeability and the actual value of the gas permeability.

[0016] As a preferred embodiment, the pressure difference method gas permeability test comprises the following steps: Cut and fix the sample to be tested between the high-pressure sealed cavity and the low-pressure sealed cavity to make the sample to be tested sealed; Adjust the temperature of the sample to be tested, the high-pressure sealed chamber, the low-pressure sealed chamber and the detection gas to the ambient temperature of the application environment parameters; The low-pressure sealed cavity and the high-pressure sealed cavity are evacuated in sequence by a vacuum pump; Introduce the test gas with set gas concentration and humidity into the high-pressure sealed cavity, so that the pressure of the high-pressure sealed cavity is the high-pressure side pressure of the application environment parameters; The detection gas penetrates from the high-pressure sealed cavity to the low-pressure sealed cavity through the sample to be tested, and the pressure in the low-pressure sealed cavity is made to be the low-pressure side pressure of the application environment parameter through the vacuum pump; after a preset time, the gas volume of each gas that penetrates into the low-pressure sealed cavity through the sample to be tested within the preset time is obtained through the detection instrument; The actual value of the gas permeability is calculated based on the gas amount of each gas.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present application constructs and trains a neural network model through experimental environment data, first detection data, first diffusion theory characteristic data and second detection data; calculates the second diffusion theory characteristic data according to the application environment parameters and the third detection data; and predicts the gas permeability through the application environment parameters, the third detection data and the second diffusion theory characteristic data of each batch of products, so as to reduce the dependence on the pressure difference method gas permeability experiment, reduce the time cost of material testing and experiments, and facilitate online monitoring and adjustment of the production process to improve the economic benefits of production.

[0018] The embodiment of the present application uses the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow rate calculated by the experimental environment data and the first detection data as the input features of the neural network model, representing the diffusion characteristics of the gas at smaller pore sizes and larger pore sizes (compared with the mean free path of gas molecules), so as to help the neural network model learn the characteristics of gas permeability more comprehensively, accelerate model training and improve model robustness.

[0019] The embodiment of the present application sets the initial weights and biases through the particle swarm optimization algorithm, which can quickly find the initial parameters with lower losses, significantly shorten the model training time, and improve the training efficiency; after setting the initial weights and biases, combined with the evaluation of the validation set and setting the threshold, the training process is made more stable, the model performance is adaptively adjusted, and over-training is avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0022] Figure 1 A schematic diagram of a process flow of a PVA film gas permeability intelligent detection method based on machine learning provided in an embodiment of the present invention; Figure 2 A schematic diagram of a process for training a neural network provided by an embodiment of the present invention; Figure 3 A schematic diagram of a process for setting initial weights and biases of a neural network model using a sample data set and a particle swarm optimization algorithm provided by an embodiment of the present invention; Figure 4 A schematic diagram of the flow chart of a pressure difference method gas permeability experiment provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0025] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0026] PVA is a water-soluble polymer with good film-forming properties, transparency and mechanical properties. It is also biocompatible and biodegradable. Based on ordinary PVA film, a PVA porous film can be made by forming a large number of pores inside it through a specific process. The presence of pores in the PVA porous film gives it a higher specific surface area and better air permeability, making it suitable for some application scenarios with specific requirements. For example, PVA porous film can be used for food preservation packaging. The selective gas permeability of the porous film can be used to control the proportion of gas components in the package and extend the shelf life of the food; or, in industrial emissions, it can effectively separate components such as carbon dioxide and nitrogen from flue gas.

[0027] Different application scenarios have different requirements for the gas permeability of PVA porous membranes. Through testing, the performance of the membrane in actual applications can be verified and data support can be provided for product improvement. In order to ensure that the gas permeability of the product meets the needs of actual use scenarios, it is often necessary to sample and test each production batch of products. The differential pressure method is a common method for determining gas permeability, but the current differential pressure method detection equipment has a long detection cycle each time. For production environments with high-frequency detection requirements, the detection efficiency of existing technologies is difficult to meet the needs.

[0028] See also Figure 1 The present invention provides a method for intelligently detecting gas permeability of a PVA film based on machine learning, comprising the steps of: S1, obtaining experimental environment data, first detection data, first diffusion theory characteristic data and second detection data, constructing and training a neural network model; the input layer of the neural network model is the experimental environment data, the first diffusion theory characteristic data and the first detection data, and the output layer is the second detection data; The experimental environment data, the first detection data and the second detection data are historical data obtained through a pressure difference method gas permeability experiment on the test sample; the first diffusion theory characteristic data is data calculated based on the experimental environment data and the first detection data.

[0029] The experimental environment data includes gas concentration, high-pressure side pressure, low-pressure side pressure, ambient temperature and ambient humidity; the first detection data includes membrane thickness, air permeability area, pore parameter group, first peak intensity and surface roughness; the second detection data includes a gas permeability parameter group.

[0030] Among them, temperature can affect the movement rate of gas molecules and the physical properties of membrane materials; humidity can change the hygroscopicity of PVA membranes, thereby affecting the structure and performance of the membrane. For the movement speed of gas molecules, under normal conditions, when the temperature rises, the movement of gas molecules intensifies, and the permeability of PVA porous membranes increases, thus causing changes in gas permeability; for the physical properties of membrane materials, the increase in temperature will reduce the cohesion of the material, increase the distance between parallel molecular chains, and increase the gas permeability.

[0031] For the film thickness, pore parameter group, first wave peak intensity and surface roughness of the first detection data, the thicker the film, the greater the resistance to gas transmission and the lower the transmittance. The first wave peak intensity characterizes the absorption peak intensity of the spectrum of the film material at a specific wavelength, which is related to the composition structure of the film material; for example, in one embodiment, the test sample is a PVA-PVAc film (polyvinyl alcohol-polyvinyl acetate film), and the first wave peak intensity is set to 1740cm -1 The peak intensity at is the absorption peak intensity of carbonyl stretching vibration, which is highly correlated with the PVAc content in the PVA-PVAc membrane, and further with the pore parameter group and gas permeability. Surface roughness is used to describe the flatness of the membrane surface, which affects the movement path of gas on the membrane surface, thereby affecting the permeability.

[0032] The gas permeability parameter group includes the gas permeability of several gases; for the test sample of the present application, ie, the PVA film, the permeability of different gases (such as oxygen, carbon dioxide, nitrogen, etc.) in the same film is different.

[0033] Furthermore, the pore parameter group includes porosity, maximum pore size, first average pore size, second average pore size and third average pore size. Among them, porosity refers to the proportion of the total volume of pores in the membrane to the total volume of the membrane, and a higher porosity usually means a higher gas permeability; the maximum pore size refers to the diameter of the largest pore in the membrane, and a larger maximum pore size may significantly increase the gas permeability and reduce the gas selectivity; the first average pore size refers to the average diameter of all pores in the membrane, reflecting the overall size characteristics of the pores; the second average pore size refers to the average diameter of all pores in the membrane that are less than or equal to the mean free path of the test gas; the third average pore size refers to the average diameter of all pores in the membrane that are greater than the mean free path of the test gas.

[0034] Furthermore, the first diffusion theory characteristic data includes a theoretical value of a Knudsen diffusion coefficient and a theoretical value of a volume flow rate; The theoretical value of the Knudsen diffusion coefficient is expressed as:

[0035] in, is the second average pore size, R is the gas constant, T is the ambient temperature, and M is the gas molecular weight.

[0036] The theoretical volume flow rate is expressed as:

[0037] in, is the third average pore size, T is the ambient temperature, is the set standard ambient temperature, is the high pressure side pressure of the pressure difference method, is the low pressure side pressure of the pressure difference method, is the set standard ambient pressure; It is the gas molecular viscosity, which is set according to the specific test gas.

[0038] In the PVA porous membrane tested in this application, the transfer of gas involves molecular sieving, Knudsen diffusion and viscous flow; among them, molecular sieving is a mechanism for separating gas molecules based on pore size. Molecules smaller than the pore size can pass through, and molecules larger than the pore size are blocked. In gas separation and purification, the pore size design of the porous membrane allows molecules of a specific size to selectively permeate, thereby achieving separation. For example, it is used for separation between molecules of different sizes such as hydrogen and methane. Knudsen diffusion refers to the situation where when the pore size of gas molecules is smaller than its mean free path, the collision of molecules mainly occurs on the pore wall, rather than between molecules. At this time, the transmission speed of gas molecules is determined by their molecular weight, and lighter molecules diffuse faster. Viscous flow refers to the flow of gas under the action of pressure gradient when the pore size is large and the frequency of collisions between gas molecules is much higher than the collision with the pore wall.

[0039] In theory, the gas permeability of each gas can be calculated based on the principles of molecular sieving, Knudsen diffusion and viscous flow. However, since the test data has errors and the calculation model is difficult to accurately fit, the present application chooses to predict the gas permeability through a neural network model. It can be understood that the principle of molecular sieving is only related to the pore size and molecular size, and the neural network model can easily learn the relevant features of the molecular sieving principle; even if the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow provided in this embodiment still have errors with the actual values, the neural network can adjust the weights to a large extent by learning and adjusting the complex relationship between the input features through the ability of the neural network to learn and adjust the complex relationship between the input features. Therefore, in this embodiment, the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow calculated by the experimental environment data and the first detection data are used as the input features of the neural network model, indicating the diffusion characteristics of the gas at a smaller pore size and a larger pore size (compared with the mean free path of the gas molecules), so as to help the neural network model learn the characteristics of the gas permeability more comprehensively, accelerate the model training and improve the model robustness.

[0040] As a preferred embodiment, the neural network model is a BP neural network model, and the weights and biases are adjusted by the back propagation algorithm to minimize the error. The input data of the neural network model includes experimental environment data, first diffusion theory feature data and first detection data, and the number of input nodes is the number of features directly included in the input data; the number of hidden layers is 3, and the number of nodes in each layer can be set to 64 to 256 nodes. The specific number of nodes in each layer can be adjusted according to the complexity of the model and actual needs; the activation function uses the ReLU function to increase the nonlinear representation ability of the model; the output data is the second detection data (gas permeability parameter group), the number of output nodes is consistent with the number of gas types in the gas permeability parameter group, and the activation function uses a linear activation function.

[0041] See also Figure 2 Based on the above embodiment, the training neural network model includes the steps of: S11, setting the experimental environment data, the first detection data, the first diffusion theory characteristic data and the second detection data as a sample data set; S12, setting the initial weights and biases of the neural network model through a sample data set and a particle swarm optimization algorithm; Dividing the sample data set into a training set and a validation set ensures that the model learns through the training set, while the validation set can be used for performance evaluation, which helps reduce the risk of overfitting. Setting the initial weights and biases of the neural network model through the particle swarm optimization algorithm can effectively explore the parameter space, improve the rationality of the initial parameters, and enhance the efficiency of model training.

[0042] S13, dividing the sample data set into a training set and a validation set; using the training set to train the neural network model through a back propagation algorithm, and minimizing the prediction error by adjusting weights and biases; The neural network model is trained using the back-propagation algorithm, which enables the model to gradually learn the underlying patterns and relationships in the sample data by continuously adjusting the weights and biases to minimize the prediction error.

[0043] S14. Use the validation set data to obtain the performance evaluation index of the neural network model. If the performance evaluation index is higher than the set threshold, the training is terminated.

[0044] The performance evaluation indicators of the model are obtained through the validation set data to ensure the performance of the model on unseen data, so as to timely identify whether the model needs to continue training, adjust the neural network structure or optimize by adjusting hyperparameters.

[0045] See also Figure 3 Based on the above embodiment, the initial weights and biases of the neural network model are set by using a sample data set and a particle swarm optimization algorithm, including the steps of: S121, randomly generate a group of particles, and initialize the position and speed of each particle; the position of the particle represents a group of weights and biases of the neural network model; the speed of the particle represents the application environment parameters of each batch of products, the third detection data and the second diffusion theory characteristic data moving direction and step length; The position vector of a particle can be expressed as P = [w1, w2, ..., w n , b1, b2, ..., b m ], where [w1,w2, ..., w n ] and [b1, b2, ..., b m ] are weights and biases respectively.

[0046] S122, constructing a loss function of a neural network, and setting the loss function as a fitness function; Among them, the loss function can use loss functions such as mean square error and cross entropy to measure the gap between the model prediction value and the true value. The loss function is set as the fitness function to reflect the performance of each particle. The lower the fitness, the better the performance.

[0047] S123, obtaining the fitness function value corresponding to each particle; inputting the position parameters (weight and bias) of each particle into the neural network, performing forward propagation, calculating the output of the model, and calculating the fitness function value based on the output of the model.

[0048] S124, updating the speed and position of each particle according to the individual best position and the group best position; Calculate the fitness function value corresponding to each particle, and record the personal best position (the best historical position of each particle) and the group best position (the best historical position of all particles).

[0049] The velocity of the particle is expressed as: , The position of a particle is expressed as: , in, is the inertia weight, which represents the influence of the velocity of the previous generation of particles on the velocity of the current generation of particles; and They are individual learning factor and group learning factor respectively; is the individual best position of a particle, and g represents the group best position of all particles. and are all random numbers in the interval [0,1], used to provide randomness for each iteration; i is the number of the particle, represents the position of the i-th particle at time t, represents the velocity of the i-th particle at time t.

[0050] S125, iterating until the maximum number of iterations or the fitness function value meets the output condition, outputting the optimal position of the particle swarm as the initial weight and bias of the neural network model.

[0051] The particle speed and position are continuously updated iteratively until the maximum number of iterations is reached or the fitness function value meets the preset output condition. After the iteration, the optimal position of the particle swarm is output as the initial weight and bias of the neural network model, providing a good starting point for the subsequent training stage.

[0052] This embodiment sets the initial weights and biases through the particle swarm optimization algorithm, which can quickly find the initial parameters with lower losses, significantly shorten the model training time, and improve the training efficiency; after setting the initial weights and biases, combined with the evaluation of the validation set and setting the threshold, the training process is more stable, and the model performance is adaptively adjusted to avoid unnecessary overtraining. Among them, the particle swarm optimization algorithm has a strong global search capability by updating the position and velocity of the particles, reducing the risk of falling into the local optimal solution; the parameters of the particle swarm optimization algorithm (such as learning factor, inertia weight, etc.) are adjustable, have good adaptability, and can be optimized for different data sets and network structures.

[0053] S2, setting application environment parameters, obtaining third detection data, and calculating second diffusion theory characteristic data according to the application environment parameters and the third detection data; Among them, the application environment parameters correspond to the experimental environment data, including gas concentration, high-pressure side pressure, low-pressure side pressure, ambient temperature and ambient humidity, and the specific values ​​are set according to the actual application scenario of the PVA porous membrane. The third test data corresponds to the first test data, including membrane thickness, air permeability area, pore parameter group, first peak intensity and surface roughness; the second diffusion theory characteristic data corresponds to the first diffusion theory characteristic data, including the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow rate.

[0054] The obtaining of the third detection data comprises: S21, using a thickness gauge to measure the actual thickness of the PVA porous film; Prepare the PVA porous membrane sample to be tested and cut it into a suitable size; adjust the thickness gauge to zero to ensure correct calibration at the beginning of the measurement; place the PVA porous membrane in the measurement area of ​​the thickness gauge; measure the thickness of the sample at different positions multiple times and take the average value as the final measurement result.

[0055] S22. Setting the ventilation area according to the testing equipment; In this embodiment, the gas permeability area refers to the surface area of ​​the sample membrane that effectively allows gas to pass through during the test process of the pressure difference method gas permeability detection device.

[0056] S23, using a scanning electron microscope (SEM) to obtain a porosity, a maximum pore size, a first average pore size, a second average pore size, and a third average pore size; Take a small piece of PVA porous membrane sample for pretreatment, such as metal spraying, to improve conductivity; put the treated membrane sample into the SEM sample chamber; adjust the SEM parameters (such as acceleration voltage, working distance) to obtain a clear pore structure image; measure the diameter of different pores one by one, calculate and record the porosity, maximum pore size, first average pore size, second average pore size and third average pore size. Among them, the porosity is the total volume percentage of pores in the membrane calculated by image analysis software; the maximum pore size is the maximum pore diameter found and measured in multiple fields of view; the first average pore size is the average value of all pores measured and calculated multiple times. The second average pore size represents the average diameter of all pores in the mold that are less than or equal to the mean free path of the test gas; the third average pore size represents the average diameter of all pores in the mold that are greater than the mean free path of the test gas.

[0057] S24, measuring the first peak intensity at a set wavelength using a Fourier transform infrared spectrometer (FT-IR); Take a small amount of PVA porous membrane sample and cut it into a suitable size; fix the sample on the FT-IR sample holder to ensure that the sample is flat; set the scanning parameters of the FT-IR instrument (such as wavelength range, resolution); scan the sample, obtain the infrared spectrum, and determine the first peak intensity at the set wavelength.

[0058] S25. Measure the surface roughness of the sample using an atomic force microscope.

[0059] Take the sample and fix it on the AFM sample stage; adjust the probe to contact the sample surface and ensure correct operation; use AFM to scan the membrane surface and set the appropriate scanning range and resolution; obtain and analyze the dimensional morphology of the sample surface and calculate the surface roughness.

[0060] Through the above detailed measurement and calculation operations, various physical and structural properties of the PVA porous membrane can be accurately obtained. These data are used as input parameters of the neural network model to provide a data basis for predicting gas permeability.

[0061] S3, inputting the application environment parameter, the third detection data and the second diffusion theory characteristic data into the neural network model to obtain the gas permeability prediction value; The application environment parameters, the third test data and the second diffusion theory characteristic data are integrated and input into the neural network model to predict the gas permeability according to the current application conditions of the PVA film to be tested and the third test data.

[0062] This step can predict the gas permeability by real-time input of the application environment parameters, the third test data and the second diffusion theory characteristic data of each batch of products, so as to reduce the dependence on the pressure difference method gas permeability experiment, reduce the time cost of material testing and experiments, and facilitate online monitoring and adjustment of the production process to improve production economic benefits.

[0063] As a preferred embodiment, the present application provides a method for intelligently detecting gas permeability of a PVA film based on machine learning, further comprising the steps of: S4. Setting a differential pressure method gas permeability experiment according to application environment parameters to obtain an actual value of gas permeability; optimizing a neural network model according to the predicted value of gas permeability and the actual value of gas permeability.

[0064] After the neural network model is deployed, the differential pressure gas permeability test is used to identify potential failure modes or inapplicable situations of the neural network model, help avoid risks, reduce uncertainties in product development and application, and ensure material safety and effectiveness.

[0065] The predicted gas permeability value obtained in step S3 is compared with the actual gas permeability value obtained in step S4, and the prediction accuracy of the model is evaluated by calculating the error (such as root mean square error RMSE or mean absolute error MAE). Based on the error analysis, the deficiencies of the model under specific conditions can be discovered, thereby guiding the improvement or retraining of the subsequent model to form a feedback mechanism, so that the model can continuously learn new data and gradually improve its prediction ability.

[0066] For further information, see Figure 4In this embodiment, the pressure difference method gas permeability experiment includes the following steps: Cut and fix the sample to be tested between the high-pressure sealed cavity and the low-pressure sealed cavity to make the sample to be tested sealed; Adjust the temperature of the sample to be tested, the high-pressure sealed chamber, the low-pressure sealed chamber and the detection gas to the ambient temperature of the application environment parameters; The low-pressure sealed cavity and the high-pressure sealed cavity are evacuated in sequence by a vacuum pump; The test gas with set gas concentration and humidity is introduced into the high-pressure sealed cavity, so that the pressure of the high-pressure sealed cavity is the high-pressure side pressure of the application environment parameters; the gas molecules have higher concentration and kinetic energy on the high-pressure side, so they will diffuse to the low-pressure side. The pressure difference provides a driving force, allowing more gas molecules to pass through the sample, and measuring its flow rate can determine the permeability of the sample.

[0067] The detection gas penetrates from the high-pressure closed cavity to the low-pressure closed cavity through the sample to be tested, and the pressure in the low-pressure closed cavity is made the low-pressure side pressure of the application environment parameter by a vacuum pump; after a preset time, the gas amount of each gas that penetrates into the low-pressure closed cavity through the sample to be tested within the preset time is obtained by the detection instrument; wherein, the detection instrument can be a pressure sensor, a gas chromatograph, etc.

[0068] The actual value of the gas permeability is calculated based on the gas amount of each gas.

[0069] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the modules can be electrical, mechanical or other forms.

[0070] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0071] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0072] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

[0073] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An intelligent detection method for gas permeability of PVA film based on machine learning, Features: Includes steps: Acquire experimental environment data, first detection data, first diffusion theory characteristic data and second detection data, and construct and train a neural network model; the input layer of the neural network model is the experimental environment data, the first detection data and the first diffusion theory characteristic data, and the output layer is the second detection data; the first diffusion theory characteristic data includes the theoretical value of the Knudsen diffusion coefficient and the theoretical value of the volume flow rate; the experimental environment data includes gas concentration, high-pressure side pressure, low-pressure side pressure, ambient temperature and ambient humidity; the first detection data includes membrane thickness, air permeability area, pore parameter group, first wave peak intensity and surface roughness; the second detection data includes a gas permeability parameter group; Setting application environment parameters, acquiring third detection data, and calculating second diffusion theory characteristic data according to the application environment parameters and the third detection data; The application environment parameters, the third detection data and the second diffusion theory characteristic data are input into the neural network model to obtain the gas permeability prediction value.

2. According to claim 1, a method for intelligent detection of gas permeability of PVA film based on machine learning, Features: The pore parameter group includes porosity, maximum pore size, first average pore size, second average pore size and third average pore size; the first average pore size represents the average diameter of all pores in the membrane; the second average pore size represents the average diameter of all pores in the mold that are less than or equal to the mean free path of the test gas; the third average pore size represents the average diameter of all pores in the mold that are greater than the mean free path of the test gas.

3. According to claim 2, a method for intelligent detection of gas permeability of PVA film based on machine learning, Features: The theoretical value of the Knudsen diffusion coefficient is expressed as: ; in, is the second average pore size, R is the gas constant, T is the ambient temperature, and M is the gas molecular weight.

4. According to claim 2, a method for intelligent detection of gas permeability of PVA film based on machine learning, Features: The theoretical volume flow rate is expressed as: ; in, is the third average pore size, T is the ambient temperature, is the set standard ambient temperature, is the high pressure side pressure of the pressure difference method, is the low-pressure side pressure of the pressure difference method, is the set standard ambient pressure; It is the gas molecular viscosity, which is set according to the specific test gas.

5. According to claim 1, a method for intelligent detection of gas permeability of PVA film based on machine learning, Features: The training neural network model comprises the steps of: The experimental environment data, the first detection data, the first diffusion theory characteristic data and the second detection data are set as a sample data set; Set the initial weights and biases of the neural network model using a sample data set and a particle swarm optimization algorithm; Dividing the sample data set into a training set and a validation set; using the training set to train the neural network model through a back propagation algorithm, and minimizing the prediction error by adjusting weights and biases; Use the validation set data to obtain the performance evaluation index of the neural network model. If the performance evaluation index is higher than the set threshold, the training is terminated.

6. The intelligent detection method for gas permeability of PVA film based on machine learning according to claim 5, Features: The initial weights and biases of the neural network model are set by using a sample data set and a particle swarm optimization algorithm, including the steps of: A group of particles are randomly generated, and the position and speed of each particle are initialized; the position of the particle represents the weight and bias of a group of neural network models; the speed of the particle represents the application environment parameters of each batch of products, the third detection data and the second diffusion theory characteristic data moving direction and step length; Construct the loss function of the neural network and set the loss function as the fitness function; Get the fitness function value corresponding to each particle; Update the speed and position of each particle based on the individual best position and the group best position; Iterate until the maximum number of iterations or the fitness function value meets the output condition, and output the optimal position of the particle swarm as the initial weight and bias of the neural network model.

7. The intelligent detection method for gas permeability of PVA film based on machine learning according to claim 6, Features: The velocity of the particle is expressed as: ， The position of a particle is expressed as: ， in, is the inertia weight; and They are individual learning factor and group learning factor respectively; is the individual best position of a particle, and g represents the group best position of all particles and are all random numbers in the interval [0,1], used to provide randomness for each iteration; i is the number of the particle, represents the position of the i-th particle at time t, represents the velocity of the i-th particle at time t.

8. The intelligent detection method for gas permeability of PVA film based on machine learning according to claim 1, Features: The obtaining of the third detection data comprises: Use a thickness gauge to measure the actual thickness of the PVA porous film; Set the ventilation area according to the testing equipment; Using a scanning electron microscope to obtain the porosity, the maximum pore size, the first average pore size, the second average pore size, and the third average pore size; A Fourier transform infrared spectrometer is used to measure the first peak intensity at a set wavelength; The surface roughness of the samples was measured using an atomic force microscope.

9. The intelligent detection method for gas permeability of PVA film based on machine learning according to claim 1, Features: Also includes the steps: A differential pressure gas permeability experiment is set up according to the application environment parameters to obtain the actual value of the gas permeability; the neural network model is optimized according to the predicted value of the gas permeability and the actual value of the gas permeability.

10. The intelligent detection method of gas permeability of PVA film based on machine learning according to claim 9, Features: The pressure difference method gas permeability experiment comprises the following steps: Cut and fix the sample to be tested between the high-pressure sealed cavity and the low-pressure sealed cavity to make the sample to be tested sealed; Adjust the temperature of the sample to be tested, the high-pressure sealed chamber, the low-pressure sealed chamber and the detection gas to the ambient temperature of the application environment parameters; The low-pressure sealed cavity and the high-pressure sealed cavity are evacuated in sequence by a vacuum pump; Introduce the test gas with set gas concentration and humidity into the high-pressure sealed cavity, so that the pressure of the high-pressure sealed cavity is the high-pressure side pressure of the application environment parameters; The detection gas penetrates from the high-pressure sealed cavity to the low-pressure sealed cavity through the sample to be tested, and the pressure in the low-pressure sealed cavity is made to be the low-pressure side pressure of the application environment parameter through the vacuum pump; after a preset time, the gas volume of each gas that penetrates into the low-pressure sealed cavity through the sample to be tested within the preset time is obtained through the detection instrument; The actual value of the gas permeability is calculated based on the gas amount of each gas.

Images