Cigarette total ventilation rate intelligent prediction method and device based on laser drilling morphological parameters, electronic equipment and storage medium

Abstract

The invention relates to the technical field of cigarette laser drilling ventilation rate prediction, in particular to a cigarette total ventilation rate intelligent prediction method based on laser drilling morphological parameters. Cigarette punching form parameter data and cigarette punching quality index data are determined; the cigarette punching shape parameter data and the cigarette punching quality index data are preprocessed, the preprocessed cigarette punching shape parameter data and the preprocessed cigarette punching quality index data are input into an ECA-DSVR prediction model for training, and the cigarette punching shape parameter data and the cigarette punching quality index data are obtained based on a punching detection system arranged in laser punching equipment on a cigarette production line. And determining real-time data of the cigarette punching form parameters, inputting the real-time data of the cigarette punching form parameters into the trained ECA-DSVR prediction model, and outputting a cigarette total ventilation rate prediction result, so as to solve the problem that the cigarette total ventilation rate cannot be predicted due to complex hole shape parameters and non-uniform characteristic contribution in the cigarette laser punching process. And the existing method for predicting the ventilation rate of the cigarette laser drilling is low in prediction precision.

Description

Technical Field

[0001] This invention relates to the field of cigarette laser perforation ventilation rate prediction technology, and particularly to an intelligent prediction method, device, electronic device, and storage medium for the total ventilation rate of cigarettes based on laser perforation morphology parameters. Background Technology

[0002] Laser drilling allows for online adjustment of multiple indicators of cigarette products by setting drilling parameters. The laser drilling process involves the combined effects of various drilling process parameters, particularly the geometric characteristics of these parameters, which directly influence the smoke flow path and ventilation performance. Core morphological parameters such as the number of holes, hole length, hole width, and hole depth have a significant impact on the overall ventilation rate. However, in actual production, the laser drilling process is affected by factors such as equipment precision, paper characteristics, and fluctuations in processing conditions. This leads to complex nonlinear coupling relationships among these drilling parameters, indirectly making it difficult to accurately predict the overall ventilation rate using traditional linear models or empirical formulas.

[0003] Existing techniques for predicting the ventilation rate of cigarette laser perforation generally use single-factor regression or simple neural network structures to describe the mapping relationship between perforation parameters and ventilation rate. However, these techniques generally suffer from insufficient fitting ability, inadequate feature learning, and weak generalization performance, making them unable to adapt to the dynamic changes of multiple parameters in complex production environments. Furthermore, traditional statistical modeling methods such as multiple linear regression and BP neural networks are prone to getting trapped in local optima when dealing with high-dimensional, nonlinear, and multi-feature interactive data, resulting in room for improvement in prediction accuracy and model stability.

[0004] Therefore, it is necessary to improve and optimize the existing methods for predicting the ventilation rate of cigarette laser perforation in order to solve the problem of low prediction accuracy caused by the complexity of hole shape parameters and uneven feature contributions during the cigarette laser perforation process. Summary of the Invention

[0005] The purpose of this invention is to propose an intelligent prediction method, device, electronic device, and storage medium for the total ventilation rate of cigarettes based on laser perforation morphology parameters, in order to solve the problem that existing methods for predicting the ventilation rate of cigarette laser perforation have low prediction accuracy due to the complexity of hole shape parameters and uneven feature contributions during the laser perforation process of cigarettes.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] This application proposes a method for intelligent prediction of the total ventilation rate of cigarettes based on laser-drilled morphological parameters, comprising the following steps:

[0008] Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the perforation morphological parameter data and perforation quality index data of cigarettes are determined.

[0009] The perforation morphology parameters and perforation quality index data of cigarettes are preprocessed to obtain the preprocessed perforation morphology parameters and perforation quality index data of cigarettes.

[0010] The preprocessed cigarette punching morphology parameters and preprocessed cigarette punching quality index data are input into the ECA-DSVR prediction model for training to obtain a trained ECA-DSVR prediction model. The ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network and SVR support vector regression network.

[0011] Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the real-time data of the perforation morphology parameters of the cigarette is determined. The real-time data of the perforation morphology parameters of the cigarette is input into the trained ECA-DSVR prediction model, and the prediction result of the total ventilation rate of the cigarette is output.

[0012] Prior to this, the determination of cigarette perforation morphology parameters and perforation quality indicators based on the perforation detection system in the laser perforation equipment configured on the cigarette production line includes the following steps:

[0013] Data on the perforation morphology of cigarettes is collected from the perforation detection system in the laser perforation equipment installed on the cigarette production line. The perforation morphology data includes the number of perforations, the length of the perforations, the width of the perforations, and the depth of the perforations.

[0014] Based on previous data on cigarette perforation morphology parameters, previous cigarette perforation quality index data were determined; among them, the cigarette perforation quality index data is the total ventilation rate of the cigarette.

[0015] Prior to this, the preprocessing of the cigarette perforation morphology parameter data and the cigarette perforation quality index data to obtain preprocessed cigarette perforation morphology parameter data and preprocessed cigarette perforation quality index data includes the following steps:

[0016] Outlier removal was performed on the pre-processed cigarette perforation morphology parameter data and the pre-processed cigarette perforation quality index data to obtain the outlier-removed cigarette perforation morphology parameter data and the outlier-removed cigarette perforation quality index data.

[0017] The missing value filling process was performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing outliers, respectively, to obtain the missing value filled cigarette perforation morphology parameter data and the missing value filled cigarette perforation quality index data.

[0018] Based on the 3σ principle, abnormal data identification and removal are performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after missing value filling, so as to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing the outliers.

[0019] The perforation morphology parameter data and perforation quality index data of cigarettes after removing outliers were normalized to obtain normalized perforation morphology parameter data and normalized perforation quality index data.

[0020] Feature importance analysis was conducted based on the random forest model, normalized cigarette perforation morphology parameters, and normalized cigarette perforation quality index data to determine the cigarette perforation morphology parameters and cigarette perforation quality index data with lower feature importance rankings.

[0021] Based on the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking, the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation morphology parameter data and the cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation quality index data to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removal.

[0022] The perforation morphology parameters of the cigarettes after rejection were used as the perforation morphology parameters of the cigarettes after preprocessing, and the perforation quality index data of the cigarettes after rejection were used as the perforation quality index data of the cigarettes after preprocessing.

[0023] Preferably, the ECA–DSVR prediction model includes an ECA attention mechanism, a DNN neural network, and an SVR support vector regression network;

[0024] Input the real-time parameter data of the cigarette perforation pattern into the trained ECA-DSVR prediction model, and output the prediction result of the total ventilation rate of the cigarette, including the following steps:

[0025] The real-time data of cigarette punching morphology parameters are enhanced by using the ECA attention mechanism to obtain an enhanced feature vector that corresponds to the real-time data of cigarette punching morphology parameters.

[0026] High-dimensional feature extraction and fusion processing of feature vectors are performed using DNN neural networks to obtain fused deep abstract feature vectors.

[0027] The fused deep abstract feature vectors are fitted using a nonlinear regression method via an SVR support vector regression network to obtain the predicted total ventilation rate of cigarettes.

[0028] Preferably, the ECA attention mechanism includes a one-dimensional convolutional layer, an efficient channel attention mechanism layer, and a feature fusion layer;

[0029] The real-time data of cigarette perforation morphology parameters are enhanced using the ECA attention mechanism to obtain an enhanced feature vector that corresponds to the real-time data of cigarette perforation morphology parameters. The process includes the following steps:

[0030] One-dimensional convolutional layers are used to process the features in the real-time data of cigarette perforation morphology parameters in order to capture the correlation between local channels; where local channels are the feature channels corresponding to each feature in the real-time data of cigarette perforation morphology parameters.

[0031] The importance weights of each feature channel are calculated through an efficient channel attention mechanism layer to obtain weighted morphological features; among them, the different morphological features are the number of holes, hole length and hole width in the real-time data of cigarette perforation morphological parameters.

[0032] The different morphological features after weighting are fused by a feature fusion layer to obtain an enhanced feature vector that corresponds to the real-time data of the perforation morphological parameters of the cigarette.

[0033] Preferably, the DNN neural network includes a multi-layer fully connected network and hidden layers;

[0034] High-dimensional feature extraction and fusion of feature vectors using a DNN neural network to obtain a fused deep abstract feature vector includes the following steps:

[0035] The feature vectors are processed by feature mapping and nonlinear transformation through a multi-layer fully connected network to obtain the feature vectors after feature mapping and nonlinear transformation.

[0036] The feature vectors after feature mapping and nonlinear transformation are processed by activation functions in the hidden layer to perform complex nonlinear relationship processing, so as to obtain the fused deep abstract feature vector. In the process of processing the feature vectors after feature mapping and nonlinear transformation by activation functions in the hidden layer, batch normalization and Dropout techniques are used to suppress overfitting of activation functions in the hidden layer.

[0037] Preferably, the SVR support vector regression network includes a kernel function;

[0038] The fused deep abstract feature vectors are fitted using a support vector regression network (SVR) to obtain the predicted total ventilation rate of cigarettes. The steps include:

[0039] The fused deep abstract feature vectors are nonlinearly mapped using a kernel function, and an optimal hyperplane is constructed in the feature space to obtain the predicted total ventilation rate of cigarettes. In the process of nonlinearly mapping the fused deep abstract feature vectors using a kernel function and constructing the optimal hyperplane in the feature space, the penalty coefficient and relaxation variable in the kernel function are also optimized.

[0040] Prior to this, after determining the real-time data of cigarette perforation morphology parameters based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, inputting the real-time data of cigarette perforation morphology parameters into the trained ECA-DSVR prediction model, and outputting the predicted result of the total ventilation rate of the cigarette, the method further includes the following steps:

[0041] Based on the prediction results of the total ventilation rate of cigarettes, the online accurate assessment of the quality of laser real-time perforation and the dynamic prediction of the ventilation rate of cigarettes are carried out during the cigarette production process.

[0042] In its second aspect, this application proposes an intelligent prediction device for the total ventilation rate of cigarettes based on laser-drilled morphological parameters, comprising:

[0043] The cigarette punching data acquisition module is used to obtain cigarette punching morphological parameter data and cigarette punching quality index data based on the punching detection system in the laser punching equipment configured on the cigarette production line.

[0044] The cigarette punching data preprocessing module is used to preprocess the cigarette punching morphological parameter data and the cigarette punching quality index data to obtain the preprocessed cigarette punching morphological parameter data and the preprocessed cigarette punching quality index data.

[0045] The cigarette total throughput model training module is used to input the preprocessed cigarette perforation morphology parameter data and the preprocessed cigarette perforation quality index data into the ECA-DSVR prediction model for training, so as to obtain the trained ECA-DSVR prediction model; wherein, the ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network and SVR support vector regression network.

[0046] The real-time prediction module for total cigarette ventilation rate is used to determine the real-time data of cigarette perforation morphology parameters based on the perforation detection system in the laser perforation equipment on the cigarette production line. The real-time data of cigarette perforation morphology parameters is input into the trained ECA-DSVR prediction model, and the prediction result of total cigarette ventilation rate is output.

[0047] This application provides, in a third aspect, an electronic device comprising:

[0048] One or more processors;

[0049] A memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a method for intelligent prediction of total ventilation rate of cigarettes based on laser perforation morphology parameters.

[0050] In its fourth aspect, this application proposes a computer-readable storage medium storing computer instructions for causing a processor to execute an intelligent prediction method for the total ventilation rate of cigarettes based on laser-drilled morphological parameters.

[0051] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0052] 1. The present invention provides an intelligent prediction method for the total ventilation rate of cigarettes based on laser perforation morphology parameters. First, based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the perforation morphology parameter data and perforation quality index data of the cigarettes are determined. Then, the perforation morphology parameter data and perforation quality index data are preprocessed to obtain preprocessed perforation morphology parameter data and preprocessed perforation quality index data. Finally, the preprocessed perforation morphology parameter data and preprocessed perforation quality index data are input into the ECA-DSVR prediction model for training to obtain a trained model. The ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network, and SVR support vector regression network. Finally, based on the punching detection system in the laser punching equipment configured on the cigarette production line, the real-time data of the cigarette punching morphology parameters are determined. The real-time data of the cigarette punching morphology parameters are input into the trained ECA-DSVR prediction model, and the total ventilation rate of the cigarette is predicted. This solves the problem of low prediction accuracy of existing methods for predicting the ventilation rate of cigarette laser punching due to the complexity of the hole shape parameters and uneven feature contributions during the cigarette laser punching process.

[0053] 2. The present invention provides an intelligent prediction method that integrates ECA, DNN, and SVR. By introducing the ECA mechanism to weight the importance of multidimensional punching geometric parameters, the model's ability to identify and express key influencing features is significantly improved.

[0054] 3. The present invention provides an intelligent prediction method that integrates ECA, DNN, and SVR. It utilizes a DNN structure to perform deep feature extraction and nonlinear mapping on the features enhanced by ECA. In a multi-layer neuron structure, it automatically learns the complex coupling relationship between parameters such as the number of holes, hole diameter, hole width, and hole depth and the total ventilation rate of cigarettes. Compared with traditional shallow regression methods, it has greater generalization and expressive power, laying the foundation for subsequent high-precision prediction.

[0055] 4. The intelligent prediction method of this invention, which integrates ECA, DNN and SVR, achieves nonlinear regression and global generalization in a high-dimensional feature space through support vector machine. In scenarios with limited samples and high-dimensional input features, it can not only effectively suppress overfitting, but also ensure the stability and adaptability of the model in actual production, and improve the prediction accuracy of the total ventilation rate of cigarettes.

[0056] 5. The intelligent prediction method integrating ECA, DNN and SVR in this invention fully integrates the advantages of feature importance enhancement, deep nonlinear feature extraction and high-precision regression prediction. Compared with traditional SVR or single neural network structure, it has better fitting effect and stronger generalization ability in the task of predicting cigarette laser perforation with multiple parameters, strong nonlinearity and high feature noise. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating an intelligent prediction method for the total ventilation rate of cigarettes based on laser perforation morphology parameters, as described in this invention.

[0058] Figure 2 This is a schematic diagram of the ECA–DSVR prediction model structure constructed in this invention.

[0059] Figure 3 A visualization of the importance of features in a random forest.

[0060] Figure 4 A schematic diagram of a sliding window

[0061] Figure 5 This is a schematic diagram of the fitting curve between the predicted and actual values ​​of the total ventilation rate prediction performed by ECA-DSVR in an embodiment of the present invention.

[0062] Figure 6 This is a connection diagram of an intelligent prediction device for the total ventilation rate of cigarettes based on laser perforation morphology parameters, as described in this invention.

[0063] Figure 7 This is a schematic diagram of an electronic device structure according to the present invention.

[0064] Table 1 shows the historical data on the perforation morphology parameters and perforation quality indicators of cigarettes.

[0065] Table 2 shows the experimental prediction results of an intelligent prediction method for the total ventilation rate of cigarettes based on laser perforation morphological parameters. Detailed Implementation

[0066] Example 1

[0067] Figure 1-7 This is a flowchart illustrating an intelligent prediction method for the total ventilation rate of cigarettes based on laser perforation morphology parameters, as provided in Embodiment 1 of the present invention. This embodiment is applicable to the prediction of the ventilation rate of cigarettes after laser perforation. The method can be executed by an intelligent prediction device for the total ventilation rate of cigarettes based on laser perforation morphology parameters. This intelligent prediction device can be implemented in hardware and / or software, and can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:

[0068] S110. Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, determine the perforation morphological parameter data and perforation quality index data of the cigarette, as follows:

[0069] (1) Collect past cigarette punching morphology parameter data from the punching detection system in the laser punching equipment configured on the cigarette production line; wherein, the cigarette punching morphology parameter data includes the number of holes punched in the cigarette, the length of the holes punched in the cigarette, the width of the holes punched in the cigarette, and the depth of the holes punched in the cigarette.

[0070] (2) Based on the previous data on the perforation morphology of cigarettes, determine the previous data on the perforation quality index of cigarettes; among which, the perforation quality index of cigarettes is the total ventilation rate of cigarettes.

[0071] S120. The perforation morphology parameter data and perforation quality index data of cigarettes are preprocessed to obtain the preprocessed perforation morphology parameter data and preprocessed perforation quality index data of cigarettes, as detailed below:

[0072] (1) The pre-processed cigarette punching morphology parameter data and the pre-processed cigarette punching quality index data are subjected to outlier removal processing to obtain the cigarette punching morphology parameter data and the cigarette punching quality index data after outlier removal. The cigarette punching morphology parameter data and the cigarette punching quality index data after outlier removal are subjected to missing value filling processing to obtain the cigarette punching morphology parameter data and the cigarette punching quality index data after missing value filling.

[0073] In this example, the data was collected from the online detection system of the laser perforation production process, focusing on the laser perforation characteristic parameters of cigarettes in the cigarette rolling machine section. The system records the geometric features of each cigarette on the production line in real time, including the number of holes, hole length, hole width, and hole depth, and simultaneously collects the corresponding total ventilation rate. After cleaning and imputing outliers and missing values, 500 sets of sample data were selected to construct the training and testing datasets. Some of the original sample data are shown in Table 1.

[0074] Table 1 shows the previous data on cigarette perforation morphological parameters and cigarette perforation quality indicators.

[0075]

[0076] (3) Based on the 3σ principle, abnormal data identification and removal are performed on the cigarette punching morphology parameter data and the cigarette punching quality index data after missing value filling, so as to obtain the cigarette punching morphology parameter data and the cigarette punching quality index data after removing the abnormal values.

[0077] The formula for the 3σ method is shown below:

[0078]

[0079] In the formula, x represents the data after removing outliers. Let σ represent the mean and σ represent the variance.

[0080] (4) Normalize the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing outliers to obtain normalized cigarette perforation morphology parameter data and normalized cigarette perforation quality index data.

[0081] The calculation formula for the normalization method is as follows:

[0082]

[0083] In the formula, X is the normalized variable, and x is the original data; x max and x min These are the maximum and minimum values ​​of x, respectively.

[0084] (5) Based on the random forest model, the normalized cigarette punching morphology parameter data and the normalized cigarette punching quality index data, feature importance analysis was performed to determine the cigarette punching morphology parameter and the cigarette punching quality index data with lower feature importance ranking.

[0085] Among them, the Random Forest (RF) model (such as Figure 3Perform feature importance analysis on the sample data. Let the input feature vector be x = [x1, x2, ..., xn]. n The output target is the total ventilation rate y. The random forest model consists of T regression trees, and its predicted output is:

[0086]

[0087] The importance of each feature is calculated based on the decrease in the average variance:

[0088]

[0089] After normalization, the feature weights W are obtained. j This reflects the relative impact of different parameters on the total ventilation rate. The analysis results show that the number of holes, hole depth, hole width, and hole length are of high importance. The visualization results of the importance of each feature are as follows: Figure 6 As shown, based on the ranking of feature importance, feature parameters with low contribution to prediction, such as opening and suction, pore spacing, and upper surface area, are removed to reduce redundant input variables and further improve the model's prediction accuracy.

[0090] Furthermore, a sliding window method is used to segment the dataset, resulting in an observation sequence with a fixed window size as the model's input, and a fixed number of subsequent observations as the model's output, which serves as the subsequent observation sequence. The model makes a prediction on the data every 5 observations. Figure 4 The process for a sliding window used for time-step prediction has an input sequence size of 5 and an output sequence size of 1. The preprocessed dataset is divided into training and test sets in an 8:2 ratio. The temporal order is maintained during the partitioning to simulate prediction scenarios in real-world production.

[0091] (6) Based on the cigarette perforation morphology parameter data and the cigarette perforation quality index data with low feature importance ranking, remove the cigarette perforation morphology parameter data and the cigarette perforation quality index data with low feature importance ranking from the normalized cigarette perforation morphology parameter data to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removal.

[0092] (7) The perforation morphology parameter data of the cigarette after rejection processing is used as the perforation morphology parameter data of the cigarette after preprocessing, and the perforation quality index data of the cigarette after rejection processing is used as the perforation quality index data of the cigarette after preprocessing.

[0093] S130. Input the pre-processed cigarette punching morphology parameter data and the pre-processed cigarette punching quality index data into the ECA-DSVR prediction model for training, so as to obtain the trained ECA-DSVR prediction model.

[0094] The ECA–DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network, and SVR support vector regression network. The ECA attention mechanism, DNN neural network, and SVR support vector regression network are all existing technologies.

[0095] Specifically, such as Figure 2 As shown, the ECA–DSVR prediction model is constructed, including: using the perforation morphology features in the preprocessed sample dataset as input. The input data includes 10 geometric feature parameters of laser perforation, namely, the number of holes, hole length, hole width, and hole depth. First, the preprocessed sample data is input into the ECA module. This module achieves feature weighting through a local cross-channel interaction mechanism to identify and enhance key perforation features that significantly affect the total ventilation rate of cigarettes. The core idea of ​​the ECA module is to achieve efficient information interaction between channels through one-dimensional convolution without introducing additional dimensional compression, thereby avoiding the information loss problem that may occur in the traditional SE module.

[0096] Specifically, the ECA module improves computational efficiency by introducing lightweight one-dimensional convolution operations and proposes an adaptive selection strategy for cross-channel size. Assume X∈R C×H×W Let X be the input feature map, and C, H, and W be the number of channels, height, and width of the feature map, respectively. First, X undergoes global average pooling (GAP) to obtain the aggregated feature g(X) for each channel. Then, C is used to adaptively compute the number of channels k. Finally, the weights W represent the weights of each channel, calculated using a one-dimensional convolution with a kernel size of k and a sigmoid function, and assigned to the original feature map. The specific equations are as follows:

[0097]

[0098] ω=σ(C1D k (g(X)))

[0099] In the formula, g(·) represents the GAP function. ψ(·) is a function that adaptively calculates the number of cross-channels k; C1D k (·) represents one-dimensional convolution calculation; σ(·) represents the Sigmoid function; b and γ are constants.

[0100] The features output from the ECA module are fed into a deep neural network (DNN) layer for feature abstraction and nonlinear mapping. A DNN is a multi-layered unsupervised neural network consisting of an input layer, multiple hidden layers, and an output layer. These layers are fully connected, meaning that any neuron in one layer is connected to every neuron in the next layer. The entire DNN model consists of linear functions and activation functions, as shown below:

[0101] a=∑w i x i +b i

[0102] In the formula, x i It is the input value of each neuron; w i It is the coefficient of the linear relationship, b i It's a deviation.

[0103] Assuming there are L hidden layers in a DNN, the calculation of the output value can be represented as follows:

[0104] f(x)=f[a L+1 (h L (…(h 2 (a 2 (h 1 (a 1 (x))))))))] a L (x)=W L x+b

[0105] In the formula, L represents the Lth layer; x is the matrix of input variables; W and b are high-dimensional matrices; and f(x) is the introduced activation function used to increase the nonlinearity of the neural network to approximate any nonlinear function of many nonlinear models.

[0106] Finally, the high-dimensional features extracted by the DNN layer are input into the Support Vector Regression (SVR) module for prediction. The SVR module aims to minimize structural risk by mapping features to a high-dimensional space using a kernel function, constructing a nonlinear regression model to achieve accurate prediction of the total ventilation rate. The basic form of SVR is:

[0107]

[0108] In the formula, K(x) i (x) is the kernel function, a i , Let be the Lagrange multiplier and b be the bias term. Through a cascaded structure of DNN and SVR, this invention achieves a combination of deep feature extraction and high-precision regression prediction, possessing both nonlinear learning capabilities and robust generalization performance.

[0109] In summary, the overall structure of the ECA-DSVR prediction model proposed in this invention is as follows: Figure 3 As shown, this model uses laser-drilled perforation morphology parameters as input features. ECA (Electronic Dynamic Alignment) is introduced into the model structure to achieve weighted modeling of key perforation features, improving the model's attention to and discrimination ability regarding important geometric parameters affecting the overall ventilation rate of cigarettes. In the feature abstraction stage, a DNN network structure is used to perform multi-layer nonlinear mapping on the input features, capturing the complex nonlinear coupling relationships and high-dimensional interaction features between perforation parameters. In the prediction output stage, an SVR (Simultaneous Dynamic Regression) module is introduced, using kernel function mapping to achieve high-precision fitting and regression prediction of ventilation rate changes. Through the synergistic integration of the above structures, the fitting performance and generalization ability in cigarette ventilation rate prediction tasks can be improved, providing effective technical support for perforation parameter optimization and intelligent ventilation quality control in cigarette production.

[0110] Furthermore, the preprocessed sample set was divided into a training set and a test set in an 8:2 ratio. The ECA-DSVR prediction model was trained using the training set. The trained ECA-DSVR prediction model was then validated using the test set, and a comparative experiment was designed to verify the model's effectiveness. The percentage absolute error (MAPE), mean absolute error (MAE), and coefficient of determination (R²) were used as evaluation metrics. The formulas for calculating these metrics are as follows:

[0111]

[0112] in, This is a predicted value; Average value; y i n represents the actual value; n represents the total number of samples; the smaller the MAPE and MAE values, the closer the predicted value is to the actual value; the closer R2 is to 1, the higher the model fit.

[0113] Specifically, the constructed ECA-DSVR model is trained using the Adam optimizer to iteratively update the network parameters, with mean squared error (MSE) as the objective function to minimize the difference between predicted and actual values. During training, a ReduceLROnPlateau learning rate adjustment mechanism is introduced, automatically reducing the learning rate when the validation set error no longer decreases after several consecutive iterations to avoid the model getting trapped in local optima. Simultaneously, an early stopping mechanism is employed, terminating the training process prematurely when the validation set error shows no significant improvement after 20 consecutive training iterations to prevent overfitting.

[0114] Furthermore, the main parameters of the model are set as follows: the activation function is ReLU; the optimizer is Adam, and the initial learning rate is set to 0.001; the training batch size is 32, and the number of training epochs is set to 100; the learning rate adjustment epochs are 25, and the decrease factor is 0.1. After the DNN feature extraction is completed, the extracted high-dimensional feature vector is input into the SVR module for ventilation rate regression prediction. The SVR kernel function is the radial basis function (RBF), the penalty coefficient is set to 100, and the kernel width parameter is set to 0.1 to achieve accurate mapping and prediction of the nonlinear feature space.

[0115] S140. Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, determine the real-time data of the cigarette perforation morphology parameters, input the real-time data of the cigarette perforation morphology parameters into the trained ECA-DSVR prediction model, and output the predicted result of the total ventilation rate of the cigarette, as follows:

[0116] The process involves inputting real-time parameter data of cigarette perforation patterns into a trained ECA-DSVR prediction model to output the predicted total ventilation rate of the cigarette, including the following steps:

[0117] (1) The real-time data of cigarette punching morphology parameters are enhanced by ECA attention mechanism to obtain the enhanced feature vector corresponding to the real-time data of cigarette punching morphology parameters.

[0118] The ECA attention mechanism comprises a one-dimensional convolutional layer, an efficient channel attention mechanism layer, and a feature fusion layer. It enhances the features of real-time cigarette perforation morphology parameters using the ECA attention mechanism to obtain an enhanced feature vector corresponding to the real-time data. This process includes the following steps: First, a one-dimensional convolutional layer is used to perform one-dimensional convolution on the features in the real-time cigarette perforation morphology parameters to capture the correlation between local channels; where local channels refer to the individual feature channels corresponding to each feature in the real-time data. Second, the efficient channel attention mechanism layer calculates the importance weights of each feature channel to obtain weighted morphological features; where different morphological features refer to the number, length, and width of holes in the real-time data. Finally, the feature fusion layer fuses the weighted morphological features to obtain an enhanced feature vector corresponding to the real-time cigarette perforation morphology parameters.

[0119] (2) High-dimensional feature extraction and fusion processing of feature vectors are performed by DNN neural network to obtain fused deep abstract feature vectors;

[0120] The DNN neural network includes a multi-layer fully connected network and hidden layers. High-dimensional feature extraction and fusion processing of feature vectors using the DNN neural network yields a fused deep abstract feature vector. This process includes the following steps: feature mapping and nonlinear transformation of the feature vectors using a multi-layer fully connected network to obtain a feature vector after feature mapping and nonlinear transformation; and complex nonlinear relationship processing of the feature vector after feature mapping and nonlinear transformation using an activation function in the hidden layer to obtain the fused deep abstract feature vector. During the complex nonlinear relationship processing of the feature vector after feature mapping and nonlinear transformation using an activation function in the hidden layer, batch normalization and Dropout techniques are used to suppress overfitting of the activation function in the hidden layer.

[0121] (3) The fused deep abstract feature vectors are subjected to nonlinear regression fitting through the SVR support vector regression network to obtain the prediction results of the total ventilation rate of cigarettes.

[0122] The SVR (Support Vector Regression) network includes a kernel function. The SVR network is used to perform nonlinear regression fitting on the fused deep abstract feature vectors to obtain the predicted total ventilation rate of cigarettes. This includes the following steps: performing nonlinear mapping on the fused deep abstract feature vectors using the kernel function and constructing an optimal hyperplane in the feature space to obtain the predicted total ventilation rate of cigarettes; wherein, during the process of performing nonlinear mapping on the fused deep abstract feature vectors using the kernel function and constructing the optimal hyperplane in the feature space, the penalty coefficient and slack variables in the kernel function are also optimized.

[0123] It should be noted that, based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the real-time data of the cigarette perforation morphology parameters are determined. After inputting the real-time data of the cigarette perforation morphology parameters into the trained ECA-DSVR prediction model and outputting the prediction result of the total ventilation rate of the cigarette, the system also includes online accurate evaluation of the laser real-time perforation quality and dynamic prediction of the cigarette ventilation rate during the cigarette production process based on the prediction result of the total ventilation rate of the cigarette.

[0124] The technical solution of this invention first determines the perforation morphological parameters and quality indicators of cigarettes based on the perforation detection system in the laser perforation equipment configured on the cigarette production line. Then, the perforation morphological parameters and quality indicators are preprocessed to obtain preprocessed perforation morphological parameters and quality indicators. Finally, the preprocessed perforation morphological parameters and quality indicators are input into the ECA-DSVR prediction model for training to obtain a trained ECA-DSVR prediction model. The ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network, and SVR support vector regression network. Finally, based on the punching detection system in the laser punching equipment configured on the cigarette production line, the real-time data of the cigarette punching morphology parameters are determined. The real-time data of the cigarette punching morphology parameters are input into the trained ECA-DSVR prediction model, and the total ventilation rate prediction result of the cigarette is output. This solves the problem of low prediction accuracy of existing methods for predicting the ventilation rate of cigarette laser punching due to the complexity of hole shape parameters and uneven feature contribution during the cigarette laser punching process.

[0125] In addition, to verify the predictive performance of the model, the proposed ECA-DSVR prediction model was compared with four deep learning models—CNN, RF, CNN-SVR, and ECA-DNN—on the same dataset. The results of the comparison experiments are shown in Table 2. Absolute percentage error (MAPE), mean absolute error (MAE), and coefficient of determination (R²) were used as comparative analysis indicators. The results show that the R² value of the proposed ECA-DSVR prediction model is 0.9593, which is closest to 1, indicating that the model has the best predictive performance. The fitting curves of the predicted and measured values ​​of the ECA-DSVR model on the test set are shown in Table 2. Figure 5 As shown.

[0126] Table 2 shows the experimental prediction results of an intelligent prediction method for the total ventilation rate of cigarettes based on laser perforation morphology parameters.

[0127]

[0128] In summary, through experiments comparing the proposed ECA-DSVR prediction model with four typical models—CNN, RF, CNN-SVR, and ECA-DNN—on the same dataset, it is evident that the traditional RF model possesses strong nonlinear fitting capabilities. However, due to its tree structure's difficulty in fully capturing the coupling relationships between variables under high-dimensional continuous features, its prediction accuracy is low, and its coefficient of determination R0 is low. 2The R² value is only 0.8069. CNN models extract features through local convolutions, enabling them to identify local correlations among morphological variables such as aperture length and width to some extent. However, limited by the fixed kernel size and insufficient global perception capability, their prediction results still exhibit bias. Furthermore, the CNN-SVR model improves prediction accuracy by inputting convolutional features into a support vector regressor for nonlinear fitting, achieving an R² value of only 0.8069. 2 The performance was improved to 0.9016, but it still has shortcomings in modeling the correlation between multiple features. The ECA-DNN model introduces ECA, which strengthens the model's response weights to key laser drilling parameters, further improving prediction performance. R 2 The accuracy reached 0.9148. In comparison, the ECA-DVR model proposed in this invention introduces a support vector regression layer on the basis of the ECA module to achieve synergistic optimization of deep features and nonlinear regression, significantly improving prediction accuracy and generalization ability. Therefore, the ECA-DSVR model exhibits higher accuracy and stability in the nonlinear mapping modeling between laser perforation morphology parameters and total ventilation rate, providing an efficient and reliable modeling method for intelligent prediction of cigarette ventilation rate and optimization of process parameters.

[0129] Example 2

[0130] Figure 6 This invention provides a structural block diagram of an intelligent prediction device for the total ventilation rate of cigarettes based on laser perforation morphology parameters. This embodiment is applicable to predicting the ventilation rate of cigarettes affected by laser perforation. This intelligent prediction device for the total ventilation rate of cigarettes based on laser perforation morphology parameters can be implemented in hardware and / or software, and can be configured in an electronic device with data processing capabilities. Figure 3 As shown, the intelligent prediction device for total ventilation rate of cigarettes based on laser perforation morphology parameters in this embodiment may include: a cigarette perforation data acquisition module 210, a cigarette perforation data preprocessing module 220, a cigarette total ventilation rate model training module 230, and a cigarette total ventilation rate real-time prediction module 240.

[0131] in:

[0132] The cigarette punching data acquisition module 210 is used to obtain cigarette punching morphological parameter data and cigarette punching quality index data based on the punching detection system in the laser punching equipment configured on the cigarette production line.

[0133] The cigarette punching data preprocessing module 220 is used to preprocess the cigarette punching morphological parameter data and the cigarette punching quality index data to obtain the preprocessed cigarette punching morphological parameter data and the preprocessed cigarette punching quality index data.

[0134] The cigarette total throughput model training module 230 is used to input the preprocessed cigarette perforation morphology parameter data and the preprocessed cigarette perforation quality index data into the ECA-DSVR prediction model for training, so as to obtain the trained ECA-DSVR prediction model; wherein, the ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network and SVR support vector regression network.

[0135] The cigarette total ventilation rate real-time prediction module 240 is used to determine the real-time data of cigarette perforation morphology parameters based on the perforation detection system in the laser perforation equipment on the cigarette production line, input the real-time data of cigarette perforation morphology parameters into the trained ECA-DSVR prediction model, and output the prediction result of cigarette total ventilation rate.

[0136] The technical solution of this invention first uses a cigarette perforation data acquisition module 210 based on a perforation detection system in a laser perforation device configured on a cigarette production line to obtain cigarette perforation morphological parameter data and cigarette perforation quality index data. Then, a cigarette perforation data preprocessing module preprocesses the cigarette perforation morphological parameter data and cigarette perforation quality index data to obtain preprocessed cigarette perforation morphological parameter data and preprocessed cigarette perforation quality index data. Finally, a cigarette total throughput model training module 230 inputs the preprocessed cigarette perforation morphological parameter data and preprocessed cigarette perforation quality index data into an ECA-DSVR prediction model for training to obtain the trained model. The trained ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network, and SVR support vector regression network. Finally, the real-time prediction module 240 for the total ventilation rate of cigarettes determines the real-time data of the cigarette perforation morphology parameters based on the perforation detection system in the laser perforation equipment on the cigarette production line. The real-time data of the cigarette perforation morphology parameters is input into the trained ECA-DSVR prediction model, and the prediction result of the total ventilation rate of cigarettes is output. This solves the problem of low prediction accuracy of existing methods for predicting the ventilation rate of cigarette laser perforation due to the complexity of the hole shape parameters and the uneven contribution of features during the laser perforation process of cigarettes.

[0137] Based on the above embodiments, optionally, the cigarette punching data acquisition module 210 is specifically used for:

[0138] Data on the perforation morphology of cigarettes is collected from the perforation detection system in the laser perforation equipment installed on the cigarette production line. The perforation morphology data includes the number of perforations, the length of the perforations, the width of the perforations, and the depth of the perforations.

[0139] Based on previous data on cigarette perforation morphology parameters, previous cigarette perforation quality index data were determined; among them, the cigarette perforation quality index data is the total ventilation rate of the cigarette.

[0140] Based on the above embodiments, optionally, the cigarette punching data preprocessing module 220 is specifically used for:

[0141] Outlier removal was performed on the pre-processed cigarette perforation morphology parameter data and the pre-processed cigarette perforation quality index data to obtain the outlier-removed cigarette perforation morphology parameter data and the outlier-removed cigarette perforation quality index data.

[0142] The missing value filling process was performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing outliers, respectively, to obtain the missing value filled cigarette perforation morphology parameter data and the missing value filled cigarette perforation quality index data.

[0143] Based on the 3σ principle, abnormal data identification and removal are performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after missing value filling, so as to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing the outliers.

[0144] The perforation morphology parameter data and perforation quality index data of cigarettes after removing outliers were normalized to obtain normalized perforation morphology parameter data and normalized perforation quality index data.

[0145] Feature importance analysis was conducted based on the random forest model, normalized cigarette perforation morphology parameters, and normalized cigarette perforation quality index data to determine the cigarette perforation morphology parameters and cigarette perforation quality index data with lower feature importance rankings.

[0146] Based on the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking, the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation morphology parameter data and the cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation quality index data to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removal.

[0147] The perforation morphology parameters of the cigarettes after rejection were used as the perforation morphology parameters of the cigarettes after preprocessing, and the perforation quality index data of the cigarettes after rejection were used as the perforation quality index data of the cigarettes after preprocessing.

[0148] Based on the above embodiments, optionally, the real-time prediction module 240 for total cigarette ventilation rate is specifically used for:

[0149] Input the real-time parameter data of the cigarette perforation pattern into the trained ECA-DSVR prediction model, and output the prediction result of the total ventilation rate of the cigarette, including the following steps:

[0150] The real-time data of cigarette punching morphology parameters are enhanced by using the ECA attention mechanism to obtain an enhanced feature vector that corresponds to the real-time data of cigarette punching morphology parameters.

[0151] High-dimensional feature extraction and fusion processing of feature vectors are performed using DNN neural networks to obtain fused deep abstract feature vectors.

[0152] The fused deep abstract feature vectors are fitted using a nonlinear regression method via an SVR support vector regression network to obtain the predicted total ventilation rate of cigarettes.

[0153] Example 3

[0154] Figure 7 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0155] like Figure 7 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0156] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0157] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a method for intelligent prediction of total ventilation rate of cigarettes based on laser perforation morphology parameters.

[0158] In some embodiments, a method for intelligently predicting the total ventilation rate of cigarettes based on laser-drilled perforation morphology parameters can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for intelligently predicting the total ventilation rate of cigarettes based on laser-drilled perforation morphology parameters described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform a method for intelligently predicting the total ventilation rate of cigarettes based on laser-drilled perforation morphology parameters by any other suitable means (e.g., by means of firmware).

[0159] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0160] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0161] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0162] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0163] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0164] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0165] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and no limitation is imposed herein.

[0166] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0167] Although the invention has been described herein with reference to several illustrative embodiments, it should be understood that many other modifications and implementations can be devised by those skilled in the art, which will fall within the scope and spirit of the principles disclosed herein. More specifically, various variations and modifications can be made to the components and / or layout of the subject matter arrangement within the scope of the disclosure, drawings, and claims. Besides variations and modifications to the components and / or layout, other uses will be apparent to those skilled in the art.

Claims

1. A method for intelligent prediction of total ventilation rate of cigarettes based on laser perforation morphology parameters, characterized in that: Includes the following steps: Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the perforation morphological parameter data and perforation quality index data of cigarettes are determined. The perforation morphology parameters and perforation quality index data of cigarettes are preprocessed to obtain the preprocessed perforation morphology parameters and perforation quality index data of cigarettes. The preprocessed cigarette punching morphology parameters and preprocessed cigarette punching quality index data are input into the ECA-DSVR prediction model for training to obtain a trained ECA-DSVR prediction model. The ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network and SVR support vector regression network. Based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, the real-time data of the perforation morphology parameters of the cigarette is determined. The real-time data of the perforation morphology parameters of the cigarette is input into the trained ECA-DSVR prediction model, and the prediction result of the total ventilation rate of the cigarette is output.

2. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The perforation detection system, based on the laser perforation equipment configured on the cigarette production line, determines the perforation morphological parameters and perforation quality indicators of the cigarettes, including the following steps: Data on the perforation morphology of cigarettes is collected from the perforation detection system in the laser perforation equipment installed on the cigarette production line. The perforation morphology data includes the number of perforations, the length of the perforations, the width of the perforations, and the depth of the perforations. Based on previous data on cigarette perforation morphology parameters, previous cigarette perforation quality index data were determined; among them, the cigarette perforation quality index data is the total ventilation rate of the cigarette.

3. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The preprocessing of the cigarette perforation morphology parameter data and the cigarette perforation quality index data to obtain preprocessed cigarette perforation morphology parameter data and preprocessed cigarette perforation quality index data includes the following steps: Outlier removal was performed on the pre-processed cigarette perforation morphology parameter data and the pre-processed cigarette perforation quality index data to obtain the outlier-removed cigarette perforation morphology parameter data and the outlier-removed cigarette perforation quality index data. The missing value filling process was performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing outliers, respectively, to obtain the missing value filled cigarette perforation morphology parameter data and the missing value filled cigarette perforation quality index data. Based on the 3σ principle, abnormal data identification and removal are performed on the cigarette perforation morphology parameter data and the cigarette perforation quality index data after missing value filling, so as to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removing the outliers. The perforation morphology parameter data and perforation quality index data of cigarettes after removing outliers were normalized to obtain normalized perforation morphology parameter data and normalized perforation quality index data. Feature importance analysis was conducted based on the random forest model, normalized cigarette perforation morphology parameters, and normalized cigarette perforation quality index data to determine the cigarette perforation morphology parameters and cigarette perforation quality index data with lower feature importance rankings. Based on the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking, the cigarette perforation morphology parameter data and cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation morphology parameter data and the cigarette perforation quality index data with low feature importance ranking are removed from the normalized cigarette perforation quality index data to obtain the cigarette perforation morphology parameter data and the cigarette perforation quality index data after removal. The perforation morphology parameters of the cigarettes after rejection were used as the perforation morphology parameters of the cigarettes after preprocessing, and the perforation quality index data of the cigarettes after rejection were used as the perforation quality index data of the cigarettes after preprocessing.

4. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The ECA–DSVR prediction model includes an ECA attention mechanism, a DNN neural network, and an SVR support vector regression network. Input the real-time parameter data of the cigarette perforation pattern into the trained ECA-DSVR prediction model, and output the prediction result of the total ventilation rate of the cigarette, including the following steps: The real-time data of cigarette punching morphology parameters are enhanced by using the ECA attention mechanism to obtain an enhanced feature vector that corresponds to the real-time data of cigarette punching morphology parameters. High-dimensional feature extraction and fusion processing of feature vectors are performed using DNN neural networks to obtain fused deep abstract feature vectors. The fused deep abstract feature vectors are fitted using a nonlinear regression method via an SVR support vector regression network to obtain the predicted total ventilation rate of cigarettes.

5. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The ECA attention mechanism includes a one-dimensional convolutional layer, an efficient channel attention mechanism layer, and a feature fusion layer. The real-time data of cigarette perforation morphology parameters are enhanced using the ECA attention mechanism to obtain an enhanced feature vector that corresponds to the real-time data of cigarette perforation morphology parameters. The process includes the following steps: One-dimensional convolutional layers are used to process the features in the real-time data of cigarette perforation morphology parameters in order to capture the correlation between local channels; where local channels are the feature channels corresponding to each feature in the real-time data of cigarette perforation morphology parameters. The importance weights of each feature channel are calculated through an efficient channel attention mechanism layer to obtain weighted morphological features; among them, the different morphological features are the number of holes, hole length and hole width in the real-time data of cigarette perforation morphological parameters. The different morphological features after weighting are fused by a feature fusion layer to obtain an enhanced feature vector that corresponds to the real-time data of the perforation morphological parameters of the cigarette.

6. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The DNN neural network includes a multi-layer fully connected network and hidden layers; High-dimensional feature extraction and fusion of feature vectors using a DNN neural network to obtain a fused deep abstract feature vector includes the following steps: The feature vectors are processed by feature mapping and nonlinear transformation through a multi-layer fully connected network to obtain the feature vectors after feature mapping and nonlinear transformation. The feature vectors after feature mapping and nonlinear transformation are processed by activation functions in the hidden layer to perform complex nonlinear relationship processing, so as to obtain the fused deep abstract feature vector. In the process of processing the feature vectors after feature mapping and nonlinear transformation by activation functions in the hidden layer, batch normalization and Dropout techniques are used to suppress overfitting of activation functions in the hidden layer.

7. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The SVR support vector regression network includes a kernel function; The fused deep abstract feature vectors are fitted using a support vector regression network (SVR) to obtain the predicted total ventilation rate of cigarettes. The steps include: The fused deep abstract feature vectors are nonlinearly mapped using a kernel function, and an optimal hyperplane is constructed in the feature space to obtain the predicted total ventilation rate of cigarettes. In the process of nonlinearly mapping the fused deep abstract feature vectors using a kernel function and constructing the optimal hyperplane in the feature space, the penalty coefficient and relaxation variable in the kernel function are also optimized.

8. The intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters according to claim 1, characterized in that: The process, based on the perforation detection system in the laser perforation equipment configured on the cigarette production line, determines real-time data of the cigarette perforation morphology parameters. After inputting this real-time data into the trained ECA-DSVR prediction model and outputting the predicted total ventilation rate of the cigarette, the process further includes the following steps: Based on the prediction results of the total ventilation rate of cigarettes, the online accurate assessment of the quality of laser real-time perforation and the dynamic prediction of the ventilation rate of cigarettes are carried out during the cigarette production process.

9. A smart prediction device for the total ventilation rate of cigarettes based on laser perforation morphology parameters, characterized in that: include: The cigarette punching data acquisition module is used to obtain cigarette punching morphological parameter data and cigarette punching quality index data based on the punching detection system in the laser punching equipment configured on the cigarette production line. The cigarette punching data preprocessing module is used to preprocess the cigarette punching morphological parameter data and the cigarette punching quality index data to obtain the preprocessed cigarette punching morphological parameter data and the preprocessed cigarette punching quality index data. The cigarette total throughput model training module is used to input the preprocessed cigarette perforation morphology parameter data and the preprocessed cigarette perforation quality index data into the ECA-DSVR prediction model for training, so as to obtain the trained ECA-DSVR prediction model; wherein, the ECA-DSVR prediction model is constructed based on the ECA attention mechanism, DNN neural network and SVR support vector regression network. The real-time prediction module for total cigarette ventilation rate is used to determine the real-time data of cigarette perforation morphology parameters based on the perforation detection system in the laser perforation equipment on the cigarette production line. The real-time data of cigarette perforation morphology parameters is input into the trained ECA-DSVR prediction model, and the prediction result of total cigarette ventilation rate is output.

10. An electronic device, characterized in that, The electronic device includes: One or more processors; A memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a method for intelligent prediction of total ventilation rate of cigarettes based on laser perforation morphology parameters as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the intelligent prediction method for total ventilation rate of cigarettes based on laser perforation morphology parameters as described in any one of claims 1-8.

Images