Intelligent evaluation method for VOC (volatile organic compound) recovery effect in diaphragm printing and electronic equipment diaphragm

Abstract

The invention relates to the technical field of artificial intelligence, intelligent processing of printing data, electronic equipment diaphragms and the like, and provides an intelligent evaluation method for a VOC recovery effect in diaphragm printing and an electronic equipment diaphragm. Inlet VOC concentration detection data and outlet VOC concentration detection data of a VOC recovery device in diaphragm printing and operation parameters of the VOC recovery device are collected; the VOC recovery rate of the VOC recovery device is predicted by using the trained neural network model according to the collected printing data according to the environmental parameters during the operation of the VOC recovery device and the historical VOC recovery rate in the operation record of the VOC recovery device, and the VOC recovery rate of the VOC recovery device is predicted according to the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device according to the neural network model. According to the method, intelligent evaluation and optimization suggestions are provided for the recovery effect of the VOC recovery device, so that the recovery effect of VOC is effectively evaluated in the printing process of the electronic equipment diaphragm, fine management is performed on VOC recovery, and printing safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence, intelligent processing of printing data, and electronic device membranes, and in particular to an intelligent evaluation method for VOC recovery effect in membrane printing and an electronic device membrane. Background Art

[0002] In printing practice, solvent-based printing inks are often used to print electronic device membranes. Due to the characteristics of solvent-based printing inks, printing machines will produce VOC emissions during the printing process of electronic device membranes. High concentrations of VOCs have significant health hazards to operators, such as causing headaches, respiratory irritation, and even chronic poisoning. Long-term exposure to high-VOC environments may lead to serious occupational diseases. In order to solve this technical problem in printing, solvent recovery devices are usually installed on printing machines using solvent-based printing inks to recycle VOCs. However, in the long-term printing process, the recovery effect of solvent recovery devices on VOCs has not been evaluated, the management of VOC recovery is not sophisticated enough, and printing safety needs to be further improved.

[0003] In summary, in the prior art, there are technical problems in the printing process of electronic device films, such as the VOC recovery effect has not been evaluated, the management of VOC recovery is not sophisticated enough, and printing safety needs to be improved. Summary of the invention

[0004] In view of the deficiencies in the above-mentioned prior art, the present invention provides an intelligent evaluation method for VOC recovery effect in film printing and an electronic device film, so as to effectively evaluate the VOC recovery effect during the printing process of the electronic device film, carry out refined management of VOC recovery, and ensure printing safety.

[0005] In a first aspect, the present invention provides an intelligent evaluation method for VOC recovery effect in film printing, comprising: Collect the inlet VOC concentration detection data of the VOC recovery device during film printing, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters when the VOC recovery device is running, and the historical VOC recovery rate in the VOC recovery device operation record; According to the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters when the VOC recovery device is running, and the historical VOC recovery rate, the VOC recovery rate of the VOC recovery device is predicted using the trained neural network model; According to the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device using the neural network model, an intelligent evaluation and optimization suggestion for the recovery effect of the VOC recovery device are provided.

[0006] In a second aspect, the present invention provides an electronic device film. When printing the electronic device film, the intelligent evaluation method for VOC recovery effect in film printing is used to evaluate the VOC recovery effect of the VOC recovery device.

[0007] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides an intelligent evaluation method for VOC recovery effect in membrane printing and an electronic device membrane. The method collects inlet VOC concentration detection data, outlet VOC concentration detection data, operating parameters of the VOC recovery device, environmental parameters when the VOC recovery device is running, and historical VOC recovery rates in the operation records of the VOC recovery device in membrane printing. According to the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, environmental parameters when the VOC recovery device is running, and the historical VOC recovery rates, a trained neural network model is used to predict the VOC recovery rate of the VOC recovery device. According to the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device by the neural network model, an intelligent evaluation and optimization suggestion for the recovery effect of the VOC recovery device are provided, thereby achieving effective evaluation of the VOC recovery effect in the printing process of the electronic device membrane, performing refined management of VOC recovery, and ensuring printing safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The drawings described herein are used to provide further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0009] Figure 1 It is a flow chart of an intelligent evaluation method for VOC recovery effect in film printing according to an embodiment of the present invention; Figure 2 It is a schematic diagram of a process for training a neural network model in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only an embodiment of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0011] Embodiment 1 The present embodiment provides an intelligent evaluation method for VOC recovery effect in membrane printing, by collecting inlet VOC concentration detection data, outlet VOC concentration detection data, operating parameters of the VOC recovery device, environmental parameters during operation of the VOC recovery device, and historical VOC recovery rates in the operation records of the VOC recovery device in membrane printing, and predicting the VOC recovery rate of the VOC recovery device using a trained neural network model based on the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, environmental parameters during operation of the VOC recovery device, and the historical VOC recovery rates, and providing intelligent evaluation and optimization suggestions for the recovery effect of the VOC recovery device based on the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device using the neural network model, thereby achieving effective evaluation of the VOC recovery effect in the printing process of electronic equipment membranes, fine management of VOC recovery, and ensuring printing safety.

[0012] See also Figure 1-Figure 2 The intelligent evaluation method for VOC recovery effect in film printing provided in this embodiment may include the following steps: S101, collecting inlet VOC concentration detection data of the VOC recovery device during film printing, outlet VOC concentration detection data, operation parameters of the VOC recovery device, environmental parameters during the operation of the VOC recovery device, and historical VOC recovery rates in the operation records of the VOC recovery device; S102, predicting the VOC recovery rate of the VOC recovery device using the trained neural network model according to the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters when the VOC recovery device is running, and the historical VOC recovery rate; S103, providing intelligent evaluation and optimization suggestions on the recovery effect of the VOC recovery device based on the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device according to the neural network model.

[0013] It should be noted that in step S101, printing data related to VOC recovery evaluation in membrane printing are collected, and these data include the inlet VOC concentration detection data of the VOC recovery device in membrane printing, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters during the operation of the VOC recovery device, and the historical VOC recovery rate in the operation record of the VOC recovery device. Among them, the inlet VOC concentration detection data and the outlet VOC concentration detection data directly reflect the change in VOC concentration and can be used to calculate the treatment effect of the recovery device. The operating parameters of the VOC recovery device (such as air volume, temperature, adsorption material state, etc.) reflect the working state of the recovery device and determine the recovery efficiency. The environmental parameters (such as humidity, air pressure, etc.) during the operation of the VOC recovery device are external conditions that affect the VOC adsorption process of the VOC recovery device. The historical recovery rate can provide a benchmark for model training and trend analysis. In step S101, by collecting the inlet and outlet VOC concentrations, operating parameters and environmental parameters, the key factors affecting the VOC recovery effect are fully covered to ensure the comprehensiveness and accuracy of the model prediction. The operation record of the VOC recovery device reflects the long-term operation trend of the equipment and can provide a reliable data source for neural network model training. The multidimensionality of the data can provide a data basis for subsequent intelligent evaluation and optimization. In step S102, the VOC recovery rate is predicted using a neural network model. The neural network model can learn the complex nonlinear relationship between the input features, and the prediction results are more accurate compared to traditional methods (such as simple linear regression). Through high-precision model prediction, the current VOC recovery efficiency can be quickly obtained, providing a core basis for subsequent intelligent evaluation. In step S103, based on the comparison of the predicted VOC recovery rate with the target recovery rate, the current working effect of the VOC recovery device can be evaluated, specific problems (such as abnormal operating parameters or equipment aging) can be located, and targeted optimization suggestions (such as adjusting the air volume, controlling temperature and humidity, and replacing adsorption materials) can be put forward to achieve refined management of VOC recovery. Moreover, by promptly identifying situations where the recovery effect is poor, the risk of exceeding the VOC emission standard can be reduced and printing safety can be guaranteed.

[0014] In some preferred embodiments, based on the prediction results obtained by predicting the VOC recovery rate of the VOC recovery device by the neural network model, an intelligent evaluation and optimization suggestions for the recovery effect of the VOC recovery device are provided, including: comparing the VOC recovery rate predicted by the neural network model with the target recovery rate, judging whether the current working state of the VOC recovery device meets expectations, so as to provide intelligent evaluation and optimization suggestions for the recovery effect of the VOC recovery device. It should be noted that in this embodiment, through the benchmark value of the target recovery rate, it is possible to quickly judge whether the actual operating effect of the VOC recovery device reaches the expected level. If the actual predicted VOC recovery rate is lower than the target value, it means that the recovery effect is not good and corresponding optimization measures need to be taken; if it is higher than or close to the target value, it means that the device is working normally and the recovery effect is ideal.

[0015] In some further embodiments, when the actual predicted VOC recovery rate is lower than the target recovery rate, the characteristic weights of the neural network model are used to analyze the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters during the operation of the VOC recovery device, and the key parameters that have a greater impact on the VOC recovery rate in the historical VOC recovery rate to obtain negative key parameters that lead to poor recovery effects; and optimization suggestions are provided based on the negative key parameters. It should be noted that when the actual recovery rate is lower than the target recovery rate, the cause needs to be promptly investigated. By extracting the characteristic weights of the neural network model, the degree of influence of each feature (such as inlet VOC concentration, ambient temperature and humidity, air volume, etc.) on the recovery efficiency can be intuitively quantified, thereby quickly locating the negative key parameters that lead to a decrease in recovery efficiency. Only after finding the negative key parameters can more targeted improvement measures be given, for example, if the humidity is too high, it is recommended to reduce the humidity, if the adsorption material state is aged, it is recommended to replace the material, and if the air volume is insufficient, it is recommended to increase the air volume.

[0016] In some further embodiments, when using the feature weights of the neural network model to analyze the key parameters that have a greater impact on the VOC recovery rate, it includes: based on the trained neural network model, extracting the weights corresponding to each input feature, the weights corresponding to each input feature include the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters during the operation of the VOC recovery device, and the feature weights corresponding to the historical VOC recovery rate; sorting the extracted feature weights from large to small, identifying the features within the set threshold range as high-impact features; analyzing the high-impact features according to the current operating state of the VOC recovery device, and obtaining the negative key parameters that lead to poor recovery effect. It should be noted that although the neural network model can accurately predict the VOC recovery rate, it is difficult to directly explain why the recovery rate is low or which factors lead to a decrease in recovery efficiency. In this embodiment, by extracting feature weights + sorting, the importance of each feature (such as inlet VOC concentration, temperature and humidity, air volume, historical recovery rate trend, etc.) in the model prediction process can be clearly quantified, so that managers can quickly determine which features have a key impact on the current recovery efficiency. When the actual recovery rate is lower than the expected value, if the specific reason is not clear, it may be necessary to conduct a comprehensive investigation of all input features (including equipment parameters and environmental parameters, etc.), which is costly and time-consuming. In this embodiment, by sorting the feature weights and screening out the high-impact features, we can quickly focus on the few parameters that are most likely to cause poor recovery (such as aging adsorption materials, high humidity environment, abnormal air volume, etc.), greatly improving the efficiency of diagnosis and rectification. It should be noted that even if some features have higher weights in the global scope, if the actual operating parameters of the current VOC recovery device are in the normal range, they will not trigger a significant decrease in the recovery rate; on the contrary, some higher weight features are more likely to become negative key parameters if the current parameters exceed the normal range. In this embodiment, by combining the actual operating status of the current VOC recovery device (for example: abnormal temperature increase, air volume significantly lower than the standard value), the real negative key parameters that lead to the current low recovery efficiency can be accurately screened out, thereby avoiding unnecessary adjustments to high-weight features that are not abnormal.

[0017] In some further embodiments, when providing optimization suggestions based on the negative key parameters, it includes: based on the historical operation data of the VOC recovery device, setting a corresponding safety threshold value that allows fluctuations for each negative key parameter (such as air volume, temperature, humidity or adsorption material usage time, etc.); obtaining the actual operating parameters of the current VOC recovery device, comparing them one by one with the safety threshold value of each set negative key parameter, and calculating the deviation between each negative key parameter and the actual operating parameters of the current VOC recovery device; according to the deviation between each negative key parameter and the actual operating parameters of the current VOC recovery device, proposing corresponding optimization measures (for example: increasing or decreasing air volume, temperature; replacing adsorption materials; controlling environmental humidity or air pressure; adjusting intake and exhaust ratio, etc.). It should be noted that in this embodiment, by mining the historical operation data of the VOC recovery device, the numerical range or critical value of each parameter during normal and efficient operation can be determined. Compared with setting fixed values ​​based on experience, the safety threshold formed by using historical data is more scientific and applicable, and can effectively reflect the actual situation of the device under different environments and loads. In addition, by comparing the current actual operating parameters with the safety thresholds one by one, it is possible to quickly identify which parameters have large deviations. Compared with the traditional comprehensive investigation, this method is more targeted and can significantly improve diagnostic efficiency, avoiding wasting time and resources on irrelevant or normal parameters. After the specific deviations of each negative key parameter (such as air volume, temperature, humidity, adsorption material usage time, etc.) from the safety threshold are clarified, differentiated optimization measures can be proposed, such as: if the air volume is too low, increase the air volume; if the air volume is too high, reduce it appropriately; if the temperature and humidity exceed the standard, use dehumidification or cooling measures; if the adsorption material is aged, replace it in time; if the intake and exhaust ratio is abnormal, adjust the ratio scheme, etc.

[0018] In some preferred embodiments, when the VOC recovery rate of the VOC recovery device is predicted using the trained neural network model, the method includes: standardizing the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters during the operation of the VOC recovery device, and the historical VOC recovery rate to obtain standardized multidimensional feature data with a unified scale, reduced relative weight of abnormal values, and consistent numerical range; integrating the standardized multidimensional feature data into a multidimensional input feature vector for inputting the model; and inputting the multidimensional input feature vector into the trained neural network model to predict the VOC recovery rate of the VOC recovery device. It should be noted that the numerical ranges of the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters, the environmental parameters, and the historical VOC recovery rate may vary greatly. If standardized processing is not performed, the features with a larger numerical range will dominate during model training, resulting in unstable training or slow convergence. In this embodiment, by standardization (such as conversion to a form with a mean of 0 and a standard deviation of 1), all features can be within a relatively uniform scale range, ensuring that the model pays more balanced attention to each feature. In addition, the input data of the above model may have sudden peaks or abnormal points. If these abnormal values ​​are directly input into the model without any processing, it is easy to cause problems such as gradient explosion and prediction bias during training. Standardization can reduce the impact of abnormal values ​​on the overall distribution to a certain extent, making the model more robust during the learning process.

[0019] In some preferred embodiments, when the standardized multidimensional feature data is integrated into a multidimensional input feature vector for inputting a model, it includes: classifying the standardized multidimensional feature data to obtain continuous feature data, classified feature data and time series feature data; sorting the continuous feature data to obtain continuous feature data arranged in a fixed order; sorting the classified feature data to obtain numerically encoded classified feature data; sorting the time series feature data to obtain complete time series feature data; combining all the sorted continuous feature data, classified feature data and time series feature data in a fixed order to form a feature set with a unified scale and no redundancy between features; converting the combined multidimensional feature data into a vector form to form the multidimensional input feature vector for inputting the neural network model, wherein each feature corresponds to a component of the feature vector. It should be noted that in this embodiment, the standardized multidimensional feature data is first classified and sorted (continuous, classified features, time series), and then combined in a fixed order and converted into a neural network model input feature vector, which can maximize the matching of the attributes and processing requirements of different features, and ensure that the data structure of the final input model is orderly, non-redundant and easy to expand, thereby improving the model training efficiency and prediction stability, and providing a solid data foundation for the subsequent accurate evaluation of VOC recovery effect. Among them, continuous features (such as VOC concentration, temperature and humidity, air volume, pressure, etc.) and classified features (such as adsorption material state) and time series features (such as historical recovery rate) have different data forms and processing requirements. The standardized features are first classified and sorted, and the processing methods that best match their feature attributes can be used respectively (such as continuous features are arranged in order, classified features are numerically encoded, and time series features retain time order or extract statistical values) to avoid mixed processing in the same process and cause data conflicts or redundancy. In this embodiment, the three types of features (continuous, categorical, and time series) are integrated in segments, which means that if certain features are to be added in the future (such as new sensor data, more types of categorical information, or additional time series), they only need to be added in the corresponding feature category, and the overall data structure must be kept consistent with the model input specification, thereby reducing duplication of work and the risk of errors. In this embodiment, classification is performed first and then sorting can allow features of different properties (values, categories, and time series) to maintain their correct positions and data forms, avoiding being mistakenly regarded as the same type of data in the model (for example, time series features are treated as ordinary continuous features).

[0020] In some preferred embodiments, after forming the multidimensional input feature vector input to the neural network model, check whether the multidimensional input feature vector formed contains all the standardized feature data, and ensure that the feature order is consistent with the input requirements of the neural network model to ensure the integrity of the multidimensional input feature vector. It should be noted that in the process of sorting and combining continuous features, classification features, and time series features, if omissions or duplications occur, some features may not be included in the final vector, or additional irrelevant data may be mistakenly included. In this embodiment, by clarifying the inspection steps, it is ensured that all standardized features (such as inlet VOC concentration, outlet VOC concentration, environmental parameters, operating parameters, historical VOC recovery rate, etc.) are correctly included in the feature vector without missing or duplication. In addition, the input layer of the neural network model usually requires that the features correspond one by one in a specific order. Any change of order will cause confusion in the model during training and reasoning, resulting in inaccurate prediction results or the model cannot work properly. In this embodiment, by ensuring that the feature order is consistent with the input requirements of the neural network model, dislocation in the feature integration link can be avoided, and the model can be guaranteed to correctly understand the meaning of the features.

[0021] In some preferred embodiments, the continuous characteristic data includes the inlet VOC concentration detection data, the outlet VOC concentration detection data, the air volume, temperature and adsorption material usage time in the operating parameters, and the humidity and air pressure in the environmental parameters; the classified characteristic data includes the adsorption material state in the operating parameters; and the time series characteristic data includes the historical VOC recovery rate. It should be noted that different types of characteristic data usually require different preprocessing or encoding methods before model input. Among them, continuous characteristic data (such as VOC concentration, temperature, air volume, etc.) are usually directly normalized or standardized before being used as model input; classified characteristic data (such as adsorption material state) requires One-Hot encoding or numerical mapping; time series characteristic data (such as historical VOC recovery rate) needs to retain the time order or extract statistical features. In this embodiment, by classifying these features, it is helpful to carry out subsequent targeted processing, reduce confusion, and ensure the accuracy and consistency of data integration. In addition, in the VOC recovery process of membrane printing, the inlet / outlet VOC concentration, temperature, humidity, air pressure, etc. are all continuously distributed, the adsorption material state is often a discrete category, and the historical recovery rate accumulates and changes over time. The essential properties of the three are not the same. Clearly classifying these features can be closer to the logic of actual production conditions, making it easier for managers and technicians to understand and use them. In addition, by distinguishing different types of features and implementing differentiated processing, the model can more fully learn the characteristic distribution of various features and their impact on VOC recovery rates, thereby improving prediction accuracy.

[0022] In some preferred embodiments, when training the neural network model, it includes: collecting data samples including inlet VOC concentration detection data, outlet VOC concentration detection data, operating parameters of the VOC recovery device, environmental parameters during the operation of the VOC recovery device, and historical VOC recovery rates; standardizing the collected data samples to obtain standardized multidimensional feature data with a unified scale, reduced relative weight of outliers, and consistent numerical range; using the standardized multidimensional feature data after standardization as training set input features to train the neural network model, and using the corresponding actual VOC recovery rate as a label value to supervise the training of the neural network model. It should be noted that in this embodiment, by collecting real operating data, performing standardization processing, and using supervised training to construct and optimize the neural network model, the accuracy and reliability of subsequent prediction results are ensured, thereby providing solid technical support for the intelligent evaluation of VOC recovery effects in the membrane printing process. Before predicting and evaluating the VOC recovery rate in actual industrial scenarios, a large amount of real and diverse operating data is required as learning material for the model. In this embodiment, by collecting inlet / outlet VOC concentrations, operating parameters (air volume, temperature, adsorption material status / duration of use, etc.), environmental parameters (humidity, air pressure, etc.) and historical VOC recovery rates, sufficient multi-dimensional information can be provided to the model, enabling it to better learn and capture the key factors affecting VOC recovery rates and their internal relationships. Using the corresponding actual VOC recovery rates as label values, the neural network model can continuously adjust its own parameters (weights, biases) during the training process to minimize the error between the predicted value and the true value. This supervised learning process ensures that the model makes full use of the large amount of data samples collected, gradually improves the prediction accuracy, and has good generalization capabilities when facing new data.

[0023] In some preferred embodiments, the neural network model includes an input layer, a hidden layer and an output layer; the number of nodes in the input layer is consistent with the dimension of the multidimensional input feature vector; the hidden layer is provided with a number of neurons for learning the complex relationship between the input features; the output layer is provided with a node for outputting the predicted VOC recovery rate. It should be noted that the number of nodes in the input layer is set to be consistent with the dimension of the multidimensional input feature vector to ensure that each feature has a corresponding neuron receiving data at the network input end, thereby avoiding feature loss or repeated input, and ensuring that the model can fully obtain all standardized feature information (such as inlet VOC concentration, outlet VOC concentration, operating parameters, environmental parameters and historical VOC recovery rate, etc.). In the process of film printing, the factors affecting the VOC recovery rate are often not simple linear relationships (such as the interactive effects of temperature and humidity, air volume and VOC concentration). By setting a number of hidden layer neurons, high-order features or implicit patterns can be automatically extracted in the process of nonlinear mapping, these complex relationships can be learned and the accuracy of prediction can be effectively improved. The number of neurons in the hidden layer can be adjusted according to the data scale, model complexity requirements, etc., to seek a balance between fitting ability and overfitting risk. In addition, the output layer is set as a node, which directly corresponds to a continuous numerical prediction (i.e., VOC recovery rate), which is consistent with the property that the recovery rate target variable is a real value. The output form that matches the regression task can simplify the interpretation of the model results. When training the model, it is only necessary to minimize the error between the single output and the true recovery rate to complete the learning process, which is convenient for connecting with subsequent processes in practical applications, such as comparing the predicted recovery rate with the target recovery rate, determining the operating status of the device and generating optimization suggestions.

[0024] In some preferred embodiments, the standardized multidimensional feature data after standardization is used as the input feature of the training set to train the neural network model, including: using the mean square error as the loss function to calculate the error between the predicted VOC recovery rate and the actual VOC recovery rate; using the gradient descent method to minimize the loss function and adjust the parameters of the neural network model; optimizing the model weights through multiple iterations until the loss function converges; using an independent validation data set to evaluate the prediction accuracy of the model to ensure that the model has good generalization ability on unseen data; testing the performance of the model to obtain the trained neural network model. It should be noted that the mean square error (MSE) is a commonly used regression task loss function that can quantify the error between the predicted value and the true value (i.e., the predicted VOC recovery rate and the actual VOC recovery rate). MSE is more sensitive to large deviations and can urge the model to minimize large errors, thereby improving the prediction accuracy of VOC recovery rates. The gradient descent method is an efficient and general optimization algorithm that is suitable for the needs of large-scale parameter updates in neural network training. By back-propagating the gradient and then iteratively updating the model weights using the gradient information, it can quickly converge to a better solution in a high-dimensional feature space. In addition, neural networks often have many weights and bias parameters, and multiple rounds of iterations are required to gradually approach the global or local optimal point. Through repeated training and error feedback, it is ensured that the neural network learns enough information, so that it has stronger generalization ability and stability when predicting VOC recovery rate. In the training process of the neural network model, only good performance on the training set does not guarantee the model's ability to predict new data. The validation data set (the part that is not involved in the training) can help to detect overfitting or underfitting in time and evaluate the performance of the model on unseen data. After completing the training and verification, it is also necessary to use the test set (or new data in the real production environment) to perform a final performance evaluation on the model to obtain objective evaluation indicators and finally obtain the trained neural network model. Among them, the neural network model can use a multilayer perceptron.

[0025] Embodiment 2 See also Figure 1The present embodiment provides an electronic device film. When printing the electronic device film, the intelligent evaluation method for VOC recovery effect in film printing described in any of the above embodiments is used to evaluate the VOC recovery effect of the VOC recovery device. By collecting the inlet VOC concentration detection data, outlet VOC concentration detection data, operating parameters of the VOC recovery device, environmental parameters during the operation of the VOC recovery device, and historical VOC recovery rates in the operation records of the VOC recovery device during film printing, the VOC recovery rate of the VOC recovery device is predicted using a trained neural network model based on the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, environmental parameters during the operation of the VOC recovery device, and the historical VOC recovery rates. Based on the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device using the neural network model, an intelligent evaluation and optimization suggestion for the recovery effect of the VOC recovery device are provided, thereby realizing effective evaluation of the VOC recovery effect during the printing process of the electronic device film, fine management of VOC recovery, and ensuring printing safety.

[0026] It should be pointed out that the above embodiments are only preferred specific implementation modes of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An intelligent evaluation method for VOC recovery effect in film printing, characterized in that: include: Collect the inlet VOC concentration detection data of the VOC recovery device during film printing, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters when the VOC recovery device is running, and the historical VOC recovery rate in the VOC recovery device operation record; According to the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters when the VOC recovery device is running, and the historical VOC recovery rate, the VOC recovery rate of the VOC recovery device is predicted using the trained neural network model; According to the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device using the neural network model, an intelligent evaluation and optimization suggestion for the recovery effect of the VOC recovery device are provided.

2. The intelligent evaluation method for VOC recovery effect in film printing according to claim 1, characterized in that: When the VOC recovery rate of the VOC recovery device is predicted by using the trained neural network model, the method includes: standardizing the inlet VOC concentration detection data, the outlet VOC concentration detection data, the operating parameters of the VOC recovery device, the environmental parameters during the operation of the VOC recovery device, and the historical VOC recovery rate to obtain standardized multidimensional feature data with a unified scale, reduced relative weights of abnormal values, and consistent numerical ranges; integrating the standardized multidimensional feature data into a multidimensional input feature vector for inputting a model; and inputting the multidimensional input feature vector into the trained neural network model to predict the VOC recovery rate of the VOC recovery device.

3. The intelligent evaluation method for VOC recovery effect in film printing according to claim 2, characterized in that: When the standardized multidimensional feature data is integrated into a multidimensional input feature vector for inputting a model, it includes: classifying the standardized multidimensional feature data to obtain continuous feature data, classified feature data and time series feature data; sorting the continuous feature data to obtain continuous feature data arranged in a fixed order; sorting the classified feature data to obtain numerically encoded classified feature data; sorting the time series feature data to obtain complete time series feature data; combining all the sorted continuous feature data, classified feature data and time series feature data in a fixed order to form a feature set with a unified scale and no redundancy between features; converting the combined multidimensional feature data into a vector form to form the multidimensional input feature vector for inputting a neural network model, wherein each feature corresponds to a component of the feature vector.

4. The intelligent evaluation method for VOC recovery effect in film printing according to claim 3, characterized in that: After forming the multidimensional input feature vector for input into the neural network model, check whether the formed multidimensional input feature vector contains all the standardized feature data, and ensure that the feature order is consistent with the input requirements of the neural network model to ensure the integrity of the multidimensional input feature vector.

5. The intelligent evaluation method for VOC recovery effect in film printing according to claim 3, characterized in that: The continuous characteristic data includes the inlet VOC concentration detection data, the outlet VOC concentration detection data, the air volume, temperature and adsorption material usage time in the operating parameters, and the humidity and air pressure in the environmental parameters; the classified characteristic data includes the adsorption material state in the operating parameters; the time series characteristic data includes the historical VOC recovery rate.

6. The intelligent evaluation method for VOC recovery effect in film printing according to claim 2, characterized in that: When training the neural network model, it includes: collecting data samples including inlet VOC concentration detection data, outlet VOC concentration detection data, operating parameters of the VOC recovery device, environmental parameters during the operation of the VOC recovery device, and historical VOC recovery rates; standardizing the collected data samples to obtain standardized multidimensional feature data with a unified scale, reduced relative weights of abnormal values, and consistent numerical ranges; using the standardized multidimensional feature data after standardization as training set input features to train the neural network model, and using the corresponding actual VOC recovery rate as a label value to supervise the training of the neural network model.

7. The intelligent evaluation method for VOC recovery effect in film printing according to claim 6, characterized in that: The neural network model includes an input layer, a hidden layer and an output layer; the number of nodes in the input layer is consistent with the dimension of the multi-dimensional input feature vector; the hidden layer is provided with a number of neurons for learning the complex relationship between the input features; the output layer is provided with a node for outputting the predicted VOC recovery rate.

8. The intelligent evaluation method for VOC recovery effect in film printing according to claim 6, characterized in that: The standardized multidimensional feature data after standardization is used as the input feature of the training set to train the neural network model, including: using the mean square error as the loss function to calculate the error between the predicted VOC recovery rate and the actual VOC recovery rate; using the gradient descent method to minimize the loss function and adjust the parameters of the neural network model; optimizing the model weights through multiple iterations until the loss function converges; using an independent validation data set to evaluate the prediction accuracy of the model to ensure that the model has good generalization ability on unseen data; and testing the performance of the model to obtain the trained neural network model.

9. The intelligent evaluation method for VOC recovery effect in film printing according to any one of claims 1 to 8, characterized in that: According to the prediction result obtained by predicting the VOC recovery rate of the VOC recovery device by the neural network model, an intelligent evaluation and optimization suggestions for the recovery effect of the VOC recovery device are provided, including: comparing the VOC recovery rate predicted by the neural network model with the target recovery rate, judging whether the current working state of the VOC recovery device meets expectations, so as to provide an intelligent evaluation and optimization suggestions for the recovery effect of the VOC recovery device.

10. An electronic device film, characterized in that: When printing the electronic device film, the VOC recovery effect of the VOC recovery device is evaluated using the intelligent evaluation method for VOC recovery effect in film printing as described in any one of claims 1 to 9.

Images