Method and device for predicting outlet material state in cut tobacco drying process
Through the support vector regression model, the problem that the existing technology cannot finely control the wire making and drying process is solved, and higher prediction accuracy and control fineness are achieved.
Patent Information
- Application Number
- CN202411561181.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-02-14
AI Technical Summary
Existing wire dryers cannot directly reflect and predict changes in the export material state, resulting in the inability to finely control the wire making and drying process.
The support vector regression model is used to predict the state of the export material. By obtaining real-time data and performing hyperplane fitting of the feature space, the division criteria and independent variables of the material state are determined, and iterative training and hyperparameter optimization are carried out.
The prediction accuracy of the export material state is improved, and the changes in the material state during the drying process can be more accurately reflected, thereby achieving more refined control.
Smart Images
Figure CN119442183A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of silk making and drying, and more specifically, to a method and device for predicting the state of outlet materials in a silk drying process. Background Art
[0002] The working principle commonly used by tobacco cutting and drying machines is: a certain flow of steam passes through the inner wall of the cylinder thin plate, and the steam condenses in the steam channel to effectively transfer the heat to the thin plate and the tobacco. At the same time, the fan sends the ambient air into the hot air generated by the steam-heated heat exchanger. During the flow through the drum, the heat is transferred to the tobacco and the moisture is taken away by convection, so that the tobacco can be evenly dried, evenly heated, and the filling force is evenly increased, and the outlet moisture is constant. At the same time, a dehumidification system is set inside the cylinder to extract part of the water-containing and hot exhaust gas for discharge to keep the moisture content of the material constant and indirectly affect the moisture content of the material.
[0003] In order to better control the silk making and drying process, the internal program of the silk drying machine defines the host operating states such as preheating, standby, production, and residual material output, so that the silk drying machine can carry out refined processing in stages. The host operating state is a mandatory definition of the equipment, which is to drive the silk drying machine to match the material to carry out different control strategies. The factor that directly affects the control strategy is the state of the outlet material of the silk drying process. However, the switching logic of the host operating state is defined by the material signal state on the production line, which cannot directly reflect the change of the outlet material state of the silk drying machine, nor can it be used as a direct determining variable for predicting the outlet material state of the silk drying process. Therefore, the outlet material state cannot be predicted based solely on the host operating state. Summary of the invention
[0004] The present application provides a method and device for predicting the state of outlet materials in a torrefied silk drying process. The output data of the model is determined according to the changes in the outlet materials in the torrefied silk drying process, and prediction is performed using a support vector regression model. By finding a hyperplane in the feature space, the sample points are located as close to the hyperplane as possible, and a certain error is allowed within a tolerance range. This method has high adaptability to the classification of the outlet material state, and the prediction accuracy of the outlet material state is guaranteed to a great extent.
[0005] The present application provides a method for predicting the state of outlet materials in a tofu drying process, comprising:
[0006] Obtain real-time data related to the status of exported materials generated during the drying process;
[0007] Input the real-time data into the support vector regression model to obtain the export material state corresponding to the real-time data, where the export material state is one of a dry head state, a dry head transition state, a stable production state, a production fluctuation state, a dry tail transition state and a dry tail state;
[0008] Among them, the training of the support vector regression model includes:
[0009] Identify multiple independent variables related to the status of the exported material;
[0010] Determine the classification standard of the state of the outlet materials according to the changes of the outlet materials in the drying process;
[0011] Preprocessing the historical data set to obtain a training data set; wherein the preprocessing includes classifying the states of the export materials in the historical data set according to the classification standard;
[0012] The support vector regression model is iteratively trained using the training dataset.
[0013] Preferably, during the iterative training of the support vector regression model, the hyper parameters of the model are also optimized.
[0014] Preferably, the independent variables related to the outlet material state include independent variables having a direct mechanism relationship with the outlet material state and independent variables having an indirect mechanism relationship with the outlet material state.
[0015] Preferably, preprocessing the historical data set also includes time series alignment of independent variables.
[0016] Preferably, the independent variables are time-series aligned, specifically including:
[0017] Determine a first delay time from the entrance of the wire drying machine to the middle position of the cylinder of the wire drying machine;
[0018] Determine a second delay time from the middle position of the cylinder of the wire drying machine to the outlet of the wire drying machine;
[0019] A first timing alignment is performed on at least one first independent variable that needs to be delayed according to the first delay time, and a second timing alignment is performed on at least one second independent variable that needs to be delayed according to the second delay time.
[0020] Preferably, the hyper parameters of the model are optimized, specifically including:
[0021] Determine multiple hyper parameters to be optimized;
[0022] Searching at least one optimal first node set with a first step length within a first value range, wherein the first node set includes at least one first node composed of first values of all super parameters to be optimized;
[0023] Repeat the following search process until the cross-validation accuracy of the first node set meets the requirement: first, the area between two adjacent first nodes in the first node set is used as the search area, and at least one optimal second node set is searched with a second step length; wherein the second step length is half of the first step length; then, the first node set is updated with the second node set, and the first step length is updated with the second step length;
[0024] The optimal second node is searched in the finally obtained first node set, and the value of each super parameter to be optimized in the second node is used as the super parameter of the support vector regression model.
[0025] Preferably, searching at least one optimal first node set with a first step length within the first value range specifically includes:
[0026] Search all first nodes within each step length in the first value range with the first step length;
[0027] The 5-fold cross validation method is used to find the subset with the smallest mean square error value to form the first node set.
[0028] The present application also provides an outlet material state prediction device for a tofu drying process, comprising a real-time data acquisition module, a prediction module and a model training module;
[0029] The real-time data acquisition module is used to acquire the real-time data related to the state of the export materials generated during the drying process;
[0030] The prediction module is used to input the real-time data into the support vector regression model to obtain the export material state corresponding to the real-time data, and the export material state is one of the dry head state, the dry head transition state, the stable production state, the production fluctuation state, the dry tail transition state and the dry tail state;
[0031] The model training module includes an independent variable determination module, a standard determination module, a preprocessing module and an iterative training module;
[0032] The independent variable determination module is used to determine multiple independent variables related to the export material status;
[0033] The standard determination module is used to determine the classification standard of the state of the outlet material according to the change of the outlet material in the drying process;
[0034] The preprocessing module is used to preprocess the historical data set to obtain a training data set; wherein the preprocessing includes classifying the states of the export materials in the historical data set according to the classification criteria;
[0035] The iterative training module is used to iteratively train the support vector regression model using the training data set.
[0036] Preferably, the iterative training module is also used to optimize the super parameters of the model during the iterative training process.
[0037] Preferably, the independent variables related to the outlet material state include independent variables having a direct mechanism relationship with the outlet material state and independent variables having an indirect mechanism relationship with the outlet material state.
[0038] Preferably, the preprocessing module is also used to perform time alignment on the independent variables.
[0039] Preferably, when the preprocessing module performs time alignment on the independent variables, it first determines the first delay time from the entrance of the wire drying machine to the middle position of the cylinder of the wire drying machine and the second delay time from the middle position of the cylinder of the wire drying machine to the exit of the wire drying machine; then, according to the first delay time, a first time alignment is performed on at least one first independent variable that needs to be delayed, and according to the second delay time, a second time alignment is performed on at least one second independent variable that needs to be delayed and advanced.
[0040] Preferably, the iterative training module includes a parameter determination module, a first search module, a second search module and a third search module;
[0041] The parameter determination module is used to determine multiple super parameters to be optimized;
[0042] The first search module is used to search at least one optimal first node set with a first step length within a first value range, wherein the first node set includes at least one first node composed of first values of all super parameters to be optimized;
[0043] The second search module is used to repeat the following search process until the cross-validation accuracy of the first node set meets the requirement: first, the area between two adjacent first nodes in the first node set is used as the search area, and at least one optimal second node set is searched with a second step length; wherein the second step length is half of the first step length; then, the first node set is updated with the second node set, and the first step length is updated with the second step length;
[0044] The third search module is used to search for the best second node in the first node set finally obtained, and use the value of each super parameter to be optimized in the second node as the super parameter of the support vector regression model.
[0045] Preferably, the first search module is used to first search all first nodes in each step within the first value range with the first step length; then use the 5-fold cross validation method to find the subset with the smallest mean square error value to form a first node set.
[0046] Other features and advantages of the present application will become apparent from the following detailed description of exemplary embodiments of the present application with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the application and, together with the description, serve to explain the principles of the application.
[0048] Figure 1 A flow chart of a method for predicting the state of outlet materials in a silk drying process provided in this application;
[0049] Figure 2 Training flow chart of the support vector regression model provided for this application;
[0050] Figure 3 This is a structural diagram of the outlet material status prediction device for the silk drying process provided in this application. DETAILED DESCRIPTION
[0051] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present application.
[0052] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.
[0053] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.
[0054] In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.
[0055] The present application provides a method and device for predicting the state of outlet materials in a drying process. The output data of the model is determined according to the change of the outlet materials in the drying process, and the support vector regression model is used for prediction. By finding a hyperplane in the feature space, the sample points are located as close to the hyperplane as possible, and a certain error is allowed within the tolerance range. It has high adaptability to the classification of the outlet material state, and the prediction accuracy of the outlet material state is guaranteed to a great extent. In addition, the present application selects a specific area search, searches for parameter combinations in sequence according to the step size within the specified parameter value range, and achieves higher search accuracy and effectively improves search efficiency by continuously shrinking the search area and step size, thereby improving the optimization search efficiency and avoiding the calculation collapse problem of high-dimensional data.
[0056] like Figure 1 As shown, the method for predicting the state of the outlet material of the silk drying process provided by the present application includes:
[0057] S110: Acquire real-time data related to the state of the exported materials generated during the fiber drying process.
[0058] S120: Input the real-time data into the support vector regression model to obtain the export material state corresponding to the real-time data, where the export material state is one of the dry head state, dry head transition state, stable production state, production fluctuation state, dry tail transition state and dry tail state.
[0059] Among them, Figure 2 As shown, the training of the support vector regression model includes:
[0060] P1: Identify multiple independent variables related to the status of export materials.
[0061] As an embodiment, the independent variables related to the export material status include independent variables having a direct mechanism relationship with the export material status (Type I independent variables) and independent variables having an indirect mechanism relationship with the export material status (Type II independent variables).
[0062] As an example, based on the operating raw materials and material processing technology of the thin plate type tofu drying machine, the independent variables that have a direct mechanism relationship with the outlet material state of the tofu drying machine include the host operating state X 1 , weight of material passed / total weight of material X 2 , production time X 3 , Material outlet temperature X 4 、Material outlet moisture X 5 , Actual value of cylinder wall temperature X 6 , Actual value of hot air temperature X 7 , Actual value of hot air volume X 8 , Actual value of incoming material moisture X 9 , Actual value of incoming material temperature X 10 , Actual value of incoming material flow X 11 , Steam dryness value X 12 The independent variables that have an indirect mechanism relationship with the outlet material state of the tofu drying machine include the ambient temperature value X 13 , Ambient humidity value X 14 , Steam temperature value X 15 .
[0063] P2: Determine the classification standard of the export material status according to the changes of the export material during the silk drying process.
[0064] Based on the two dimensional data of the outlet material moisture and outlet material temperature of the silk drying machine, there are six changing states of the material during the silk drying process: dry head state, dry head transition state, stable production state, production fluctuation state, dry tail transition state and dry tail state.
[0065] The characteristics of the dry head state are: as soon as the material to be dried enters the wire drying machine, the material flow rate gradually increases to a constant flow rate, resulting in the inevitable "dry head" material state in the wire drying process. In this state, the flow rate of the wire drying outlet material increases nonlinearly and slowly, and the moisture and temperature of the material rise in a curve.
[0066] The characteristics of the dry head transition state are as follows: after the material flow rate becomes constant, the heat exchange inside the wire drying machine tends to be stable, the wire drying feedback control loop is in a process of regulating oscillation and tending to converge, the flow rate of the wire drying outlet material in this state fluctuates slightly and tends to be stable, and the moisture and temperature of the material fluctuate and oscillate slowly and tend to be stable.
[0067] The stable production state is characterized by constant material flow, balanced heat exchange, and multiple control loops in a steady state, and the material moisture and temperature tend to be stable and convergent process characteristics.
[0068] The characteristics of the production fluctuation state are: after a period of stable production, the moisture and temperature of the material fluctuate due to the disturbance of the incoming material flow, the incoming material state, and the internal process quantity of the wire drying.
[0069] The characteristics of the dry tail transition state are: the material flow rate gradually decreases, the silk drying has not yet entered the final state, and it is still controlled by the production steady-state process control method. The material moisture and temperature fluctuate slowly and tend to be dry, but the drying process cannot be predicted.
[0070] The characteristics of the dry tail state are: the material flow rate is significantly reduced, the drying machine recognizes that the material state has changed, and it completely enters the final state, the material moisture and temperature fluctuate greatly, and the process characteristics continue to be dry.
[0071] P3: Preprocess the historical data set to obtain the training data set.
[0072] Preprocessing the historical data set includes dividing the state of the export materials in the historical data set according to the division criteria. First, the time domain is segmented according to the historical trend chart of the material outlet temperature and the material outlet moisture in the historical data set, and then the export material state corresponding to each time domain (i.e., the value of the dependent variable Y) is calibrated according to the segmentation result. For example, the dry head state Y=1; the dry head transition state Y=2; the stable production state Y=3; the production fluctuation state Y=4; the dry tail transition state Y=5; the dry tail state Y=6.
[0073] Based on the raw materials and material processing technology of the thin-plate tofu drying machine, some of the I-type independent variables have inconsistencies in time series. Some independent variables are measured values at the equipment entrance, and some are actual measured values or indirect measured values during the processing. And there is a time domain delay relationship between each independent variable and the dependent variable. Based on such considerations, preferably, preprocessing the historical data set also includes time series alignment of the independent variables.
[0074] Perform time alignment on the independent variables, including:
[0075] Q1: Determine the first delay time from the entrance of the wire drying machine to the middle position of the cylinder of the wire drying machine.
[0076] Specifically, a marker is placed in the material during the production process, and the physical properties of the marker are similar to those of the material. The delay time from the material inlet to the middle position of the drying machine cylinder is marked by recording the marker multiple times, and the average value is calculated and recorded as t 1 , in seconds.
[0077] Q2: Determine the second delay time from the middle position of the cylinder of the silk drying machine to the outlet of the silk drying machine.
[0078] During the production process, a marker is placed in the material, and the physical properties of the marker are similar to those of the material. The delay time of the material from the middle position of the barrel of the silk drying machine to the outlet of the silk drying machine is marked by recording the marker multiple times, and the average value is calculated and recorded as t 2 , in seconds.
[0079] Q3: Perform a first timing alignment on at least one first independent variable that needs to be delayed according to the first delay time, and perform a second timing alignment on at least one second independent variable that needs to be delayed according to the second delay time.
[0080] Among them, the material outlet temperature X 4 、Material outlet moisture X 5 is the independent variable that needs to be delayed and advanced, the actual value of the incoming moisture content X 9 , Actual value of incoming material temperature X 10 , Actual value of incoming material flow X 11 is the independent variable that needs to be delayed.
[0081] Assume that the outlet temperature of the material at time t is X 4 (t), the material outlet moisture is X 5 (t), the actual moisture value of the incoming material is X 9 (t), the actual value of the incoming material temperature is X 10 (t), the actual value of incoming material flow is X 11 (t), then the material outlet temperature after the time delay is X 4 (t+t 2), the material outlet moisture after the delay is X 5 (t+t 2 ), the actual moisture value of the incoming material after the time delay is X 9 (tt 1 ), the actual value of the incoming material temperature after the time delay is X 10 (tt 1 ), the actual value of the incoming flow after the time delay is X 11 (tt 1 ).
[0082] As an embodiment, preprocessing the historical data set further includes removing outliers in the historical data set.
[0083] Specifically, based on past production records, the data of batches with faulty shutdowns and test production batches are removed, and then a historical trend chart is drawn for the independent variables and dependent variables in the historical data set, and the data of the batches where the outliers are located are removed.
[0084] Preferably, preprocessing the historical data set also includes reducing the data dimension.
[0085] As an embodiment, the pre-processed feature data is processed to reduce redundancy and correlation using the Pearson correlation coefficient method, data with a Pearson correlation coefficient greater than 0.3 are selected for model fitting, and dimensions with a correlation coefficient less than 0.3 are actively discarded.
[0086] The data set obtained after all preprocessing of the historical data set is used as the training data set.
[0087] P4: Iteratively train the support vector regression (SVR) model using the training dataset.
[0088] SVR maps regression data to high-dimensional space through nonlinear mapping, constructs regression function in high-dimensional space, and solves it through Lagrangian dual problem. Due to the nonlinear relationship between input features and target prediction values, radial basis kernel function (RBF) is specially selected for high-dimensional mapping of regression data. It is assumed that the radial kernel function is expressed as:
[0089]
[0090] In the formula, is the kernel function that satisfies the Mercer condition; g is the kernel function parameter, X′ is the origin value, and X is the value of the sample point.
[0091] Preferably, during the iterative training of the support vector regression model, the hyper parameters of the model are also optimized.
[0092] As an embodiment, a parameter combination optimization algorithm is used to solve the optimization problem of SVR. Parameter combination optimization optimizes model performance by traversing multiple possible parameter combinations through a specific efficient method. It systematically tries different parameter combinations and selects the best parameters through cross-validation results.
[0093] Optimize the model's hyper parameters, including:
[0094] P41: Identify multiple hyper parameters to be optimized.
[0095] As an example, the regularization parameter C i and kernel function parameter g i The optimization is performed by combining the two methods, where the tolerance deviation ε is obtained by artificial experience and multiple tests.
[0096] P42: In the first value range [2 -N , 2 N ] searches for at least one optimal first node set with a first step length M, the first node set including at least one first node (i.e., parameter combination) composed of first values of all super parameters to be optimized.
[0097] Searching at least one optimal first node set with a first step length within a first value range specifically includes:
[0098] P421: Search all first nodes within each step length with the first step length within the first value range. During the search process, all first nodes within each first step length (given range) are all possible parameter combinations within the given range.
[0099] P422: Use the 5-fold cross validation method to find the subset with the smallest mean square error value to form the first node set.
[0100] Specifically, all first nodes obtained by P421 are equally divided into 5 subsets (including at least one first node). Any subset is selected as a test set, and the other subsets are used as training sets for cross-validation, and the mean square error is calculated. After traversing all subsets (as test sets), the subset with the smallest mean square error is found as the first node set.
[0101] P43: Repeat the following search process until the cross-validation accuracy of the first node set meets the requirement: First, the area between two adjacent first nodes in the first node set is the search area, and at least one optimal second node set is searched with a second step length; wherein the second step length is half of the first step length (i.e. where i is the current search number); then the first node set is updated with the second node set, and the first step length is updated with the second step length.
[0102] It can be understood that the process of obtaining the second node set is the same as that of obtaining the first node set.
[0103] If the cross-validation accuracy of the second node set meets the requirement after a certain search, the first node set is updated with the second node set, and the first node set is used as the final first node set, thereby obtaining the best search range for optimization.
[0104] P44: Search for the optimal second node in the first node set finally obtained, and use the value of each super parameter to be optimized in the second node as the super parameter of the support vector regression model.
[0105] As an embodiment, the GridSearchCV function is used to automatically perform parameter search within the search range determined by the first node set finally obtained, and the performance of each parameter combination is evaluated through 5-fold cross validation, and the optimal parameter combination and its corresponding model performance score are output.
[0106] This application uses specific area search to search for parameter combinations in the specified parameter value range in steps, and achieves higher search accuracy and effectively improves search efficiency by continuously shrinking the search area and step size, thereby improving optimization search efficiency and avoiding computational collapse problems with high-dimensional data.
[0107] After obtaining the optimal hyper parameters, the optimal hyper parameters are input into the SVR model, and the model is trained using the training data set. The training results are then used to adjust the hyper parameters, thereby achieving model optimization and training.
[0108] As an embodiment, a mean square error (MSE) less than 0.5 is used as an indicator to mark the end of optimization and training.
[0109] Based on the above, the present application also provides a device for predicting the state of outlet materials in a drying process. Figure 3 As shown, the export material status prediction device includes a real-time data acquisition module 310, a prediction module 320 and a model training module 330.
[0110] The real-time data acquisition module 310 is used to acquire real-time data related to the state of the outlet material generated during the drying process.
[0111] The prediction module 320 is used to input real-time data into the support vector regression model to obtain the export material state corresponding to the real-time data. The export material state is one of the dry head state, dry head transition state, stable production state, production fluctuation state, dry tail transition state and dry tail state.
[0112] The model training module 330 includes an independent variable determination module 3301 , a standard determination module 3302 , a preprocessing module 3303 and an iterative training module 3304 .
[0113] The independent variable determination module 3301 is used to determine multiple independent variables related to the export material status.
[0114] The standard determination module 3302 is used to determine the classification standard of the outlet material status according to the change of the outlet material in the silk drying process.
[0115] The preprocessing module 3303 is used to preprocess the historical data set to obtain a training data set, wherein the preprocessing includes classifying the states of the export materials in the historical data set according to the classification standard.
[0116] The iterative training module 3304 is used to perform iterative training on the support vector regression model using the training data set.
[0117] Preferably, the iterative training module 3304 is also used to optimize the super parameters of the model during the iterative training process.
[0118] Preferably, the independent variables related to the outlet material state include independent variables having a direct mechanism relationship with the outlet material state and independent variables having an indirect mechanism relationship with the outlet material state.
[0119] Preferably, the preprocessing module 3303 is also used to perform time alignment on the independent variables.
[0120] Preferably, when the preprocessing module 3303 performs time sequence alignment on the independent variables, it first determines a first delay time from the inlet of the wire drying machine to the middle position of the cylinder of the wire drying machine and a second delay time from the middle position of the cylinder of the wire drying machine to the outlet of the wire drying machine. Then, according to the first delay time, a first time sequence alignment is performed on at least one first independent variable that needs to be delayed, and according to the second delay time, a second time sequence alignment is performed on at least one second independent variable that needs to be delayed.
[0121] Preferably, the iterative training module 3304 includes a parameter determination module 33041 , a first search module 33042 , a second search module 33043 and a third search module 33044 .
[0122] The parameter determination module 33041 is used to determine multiple super parameters to be optimized.
[0123] The first search module 33042 is used to search at least one optimal first node set with a first step length within a first value range, and the first node set includes at least one first node composed of first values of all super parameters to be optimized.
[0124] The second search module 33043 is used to repeat the following search process until the cross-validation accuracy of the first node set meets the requirement: first, the area between two adjacent first nodes in the first node set is used as the search area, and at least one optimal second node set is searched with a second step length. The second step length is half of the first step length. Then, the first node set is updated with the second node set, and the first step length is updated with the second step length.
[0125] The third search module 33044 is used to search for the optimal second node in the first node set finally obtained, and use the value of each super parameter to be optimized in the second node as the super parameter of the support vector regression model.
[0126] Preferably, the first search module 33042 is used to first search all first nodes in each step length within the first value range, and then use a 5-fold cross validation method to find a subset with the minimum mean square error value to form a first node set.
[0127] Although some specific embodiments of the present application have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are only for illustration, not for limiting the scope of the present application. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.
Claims
1. A method for predicting the state of outlet materials in a drying process, characterized in that: include: Obtain real-time data related to the status of exported materials generated during the drying process; Input the real-time data into a support vector regression model to obtain an export material state corresponding to the real-time data, wherein the export material state is one of a dry head state, a dry head transition state, a stable production state, a production fluctuation state, a dry tail transition state, and a dry tail state; The training of the support vector regression model includes: Identify multiple independent variables related to the status of the exported material; Determine the classification standard of the state of the outlet materials according to the changes of the outlet materials in the drying process; Preprocessing the historical data set to obtain a training data set; wherein the preprocessing includes classifying the states of the export materials in the historical data set according to the classification standard; The support vector regression model is iteratively trained using the training data set.
2. The method for predicting the outlet material state of the silk drying process according to claim 1, characterized in that: During the iterative training of the support vector regression model, the super parameters of the model are also optimized.
3. The method for predicting the outlet material state of the silk drying process according to claim 1, characterized in that: The independent variables related to the export material status include independent variables having a direct mechanism relationship with the export material status and independent variables having an indirect mechanism relationship with the export material status.
4. The method for predicting the outlet material state of the tofu drying process according to claim 1, characterized in that: Preprocessing the historical data set also includes time series alignment of the independent variables.
5. The method for predicting the outlet material state of the tofu drying process according to claim 4, characterized in that: Perform time alignment on the independent variables, including: Determine a first delay time from the entrance of the wire drying machine to the middle position of the cylinder of the wire drying machine; Determine a second delay time from the middle position of the cylinder of the wire drying machine to the outlet of the wire drying machine; A first timing alignment is performed on at least one first independent variable that needs to be delayed according to the first delay time, and a second timing alignment is performed on at least one second independent variable that needs to be delayed according to the second delay time.
6. A device for predicting the state of outlet materials in a drying process, characterized in that: Includes real-time data acquisition module, prediction module and model training module; The real-time data acquisition module is used to acquire real-time data related to the state of the export material generated during the silk drying process; The prediction module is used to input the real-time data into a support vector regression model to obtain the export material state corresponding to the real-time data, wherein the export material state is one of a dry head state, a dry head transition state, a stable production state, a production fluctuation state, a dry tail transition state and a dry tail state; The model training module includes an independent variable determination module, a standard determination module, a preprocessing module and an iterative training module; The independent variable determination module is used to determine a plurality of independent variables related to the state of the export material; The standard determination module is used to determine the classification standard of the outlet material state according to the change of the outlet material in the silk drying process; The preprocessing module is used to preprocess the historical data set to obtain a training data set; wherein the preprocessing includes classifying the states of the export materials in the historical data set according to the classification standard; The iterative training module is used to perform iterative training on the support vector regression model using the training data set.
7. The device for predicting the state of outlet materials in the process of drying silk according to claim 6, characterized in that: The iterative training module is also used to optimize the super parameters of the model during the iterative training process.
8. The device for predicting the state of outlet materials in the fiber drying process according to claim 6, characterized in that: The independent variables related to the export material status include independent variables having a direct mechanism relationship with the export material status and independent variables having an indirect mechanism relationship with the export material status.
9. The device for predicting the state of outlet materials in a fiber drying process according to claim 6, characterized in that: The preprocessing module is also used to perform time sequence alignment on the independent variables.
10. The device for predicting the outlet material state of the fiber drying process according to claim 9, characterized in that: When the preprocessing module performs time alignment on the independent variables, it first determines a first delay time from the entrance of the wire drying machine to the middle position of the cylinder of the wire drying machine and a second delay time from the middle position of the cylinder of the wire drying machine to the exit of the wire drying machine; then, based on the first delay time, a first time alignment is performed on at least one first independent variable that needs to be delayed, and based on the second delay time, a second time alignment is performed on at least one second independent variable that needs to be delayed.