CN108142976A - A kind of cut tobacco Drying Technology Parameter optimization method - Google Patents

Abstract

The invention discloses a method for optimizing parameters of shredded leaf drying process. The invention provides a method for optimizing the setting of shredded leaf drying process parameters, and solves the problem that the traditional method cannot establish the process function model of shredded leaf and it is difficult to optimize the setting of technological parameters. problem, while improving the operation efficiency and prediction accuracy of the prediction model, by establishing a lightweight data-driven prediction model to construct the mapping relationship between each process parameter and the moisture content of the shredded shreds during the drying process, and based on the mapping relationship to find the The combination of the optimal value of the water content and the corresponding optimal process parameters of shredded shreds drying realizes the precise optimization of the shredded shreds drying process parameters and the moisture content of the shredded leaves even when there are many parameters of the shredded shreds drying process.

Description

A Method for Optimizing Process Parameters of Shredded Leaf Drying

technical field

The invention relates to a process parameter optimization method for shredded leaves, belonging to the field of drying agricultural and sideline products.

Background technique

Leaf shred drying is an important process in the drying process of agricultural and sideline products. By optimizing the process parameters of the drying process, the moisture content of leaf shreds can be kept stable, so as to improve and control the quality of leaf shreds. At present, the optimal setting of shredded drying process parameters mainly depends on the experience of technicians, and it is difficult to adopt the optimal setting method. The main reason is that the shredded shredded drying process is a complex process involving multi-field and multidisciplinary coupling such as physics and chemistry. The relationship between the process parameters and the moisture content of shredded leaves is very complicated, and it is difficult to determine the functional relationship by traditional methods.

Contents of the invention

In order to solve the problems that the traditional method cannot establish a shredded leaf drying process function model, and it is difficult to optimize the setting of technological parameters, etc., the invention provides a shredded leaf drying process parameter optimization method.

The technical scheme of the present invention is: a kind of shredded leaf drying process parameter optimization method, described method step is as follows:

Step 1, removing abnormal data and error data in the shredded leaf drying process parameter data to obtain the process parameter data to be optimized;

Step 2. Perform dimensionality reduction processing on the process parameter data to be optimized to obtain lightweight parameter data;

Step 3. Create an initial BP neural network, substitute lightweight parameter data and silk moisture content training data into it for training, and obtain a lightweight data-driven prediction model;

Step 4, performing random screening on the process parameter data to be optimized to obtain the silk drying process parameter population;

Step 5, performing dimensionality reduction operations on individuals in the shredded leaf drying process parameter population;

Step 6. Substituting the dimensionally reduced population individual data into the lightweight data-driven prediction model to obtain the predicted value y _i of the moisture content of shredded leaves;

Step 7, Convergence Judgment: Make a difference between the predicted value y _i of the water content of the shredded leaf and the set value y ₀ of the water content of the shredded leaf, if the difference is less than or equal to the convergence accuracy e , then output the predicted value of the water content of the shredded leaf Individuals in the population corresponding to the silk drying process parameters, otherwise not output.

The dimensionality reduction process adopts principal component analysis method.

In the step 7, if it is not output, update the shredded leaf drying process parameter population, and repeat steps 5 to 7 until |y _i - y ₀ | ≤ e .

The beneficial effect of the present invention is that it provides a method for the optimal setting of the shredded drying process parameters, solves the problem that the traditional method cannot establish the shredded shredded drying process function model, and it is difficult to optimize the setting of the technological parameters, and at the same time improves the operation of the prediction model Efficiency and prediction accuracy, by establishing a lightweight data-driven prediction model to construct the mapping relationship between each process parameter and the moisture content of the shredded leaf during the drying process, and based on the mapping relationship to find the optimal value of the shredded moisture content and its corresponding The combination of the optimal process parameters for shredded shreds drying realizes the precise optimization of shredded shreds drying process parameters and moisture content of shreds even when there are many shredded shreds drying process parameters.

Description of drawings

Fig. 1 is a flowchart of the present invention.

Detailed ways

Embodiment 1: as shown in Figure 1, a kind of shredded leaf drying process parameter optimization method, described method step is as follows:

Step 1, removing abnormal data and error data in the shredded leaf drying process parameter data to obtain the process parameter data to be optimized;

Step 2. Perform dimensionality reduction processing on the process parameter data to be optimized to obtain lightweight parameter data;

Step 3. Create an initial BP neural network, substitute lightweight parameter data and silk moisture content training data into it for training, and obtain a lightweight data-driven prediction model;

Step 4, performing random screening on the process parameter data to be optimized to obtain the silk drying process parameter population;

Step 5, performing dimensionality reduction operations on individuals in the shredded leaf drying process parameter population;

Step 6. Substituting the dimensionally reduced population individual data into the lightweight data-driven prediction model to obtain the predicted value y _i of the moisture content of shredded leaves;

Step 7, Convergence Judgment: Make a difference between the predicted value y _i of the water content of the shredded leaf and the set value y ₀ of the water content of the shredded leaf, if the difference is less than or equal to the convergence accuracy e , then output the predicted value of the water content of the shredded leaf Individuals in the population corresponding to the silk drying process parameters, otherwise not output.

Further, it may be set that the dimensionality reduction process adopts a principal component analysis method.

Embodiment 2: as shown in Figure 1, a kind of shredded leaf drying process parameter optimization method, described method step is as follows:

Step 1, removing abnormal data and error data in the shredded leaf drying process parameter data to obtain the process parameter data to be optimized;

Step 2. Perform dimensionality reduction processing on the process parameter data to be optimized to obtain lightweight parameter data;

Step 3. Create an initial BP neural network, substitute lightweight parameter data and silk moisture content training data into it for training, and obtain a lightweight data-driven prediction model;

Step 4, performing random screening on the process parameter data to be optimized to obtain the silk drying process parameter population;

Step 5, performing dimensionality reduction operations on individuals in the shredded leaf drying process parameter population;

Step 6. Substituting the dimensionally reduced population individual data into the lightweight data-driven prediction model to obtain the predicted value y _i of the moisture content of shredded leaves;

Step 7, Convergence Judgment: Make a difference between the predicted value y _i of the water content of the shredded leaf and the set value y ₀ of the water content of the shredded leaf, if the difference is less than or equal to the convergence accuracy e , then output the predicted value of the water content of the shredded leaf Individuals in the corresponding leaf shred drying process parameter population, otherwise update the leaf shred drying process parameter population, and perform steps 5 to 7 again until |y _i - y ₀ | ≤ e .

Further, it may be set that the dimensionality reduction process adopts a principal component analysis method.

Embodiment 3: as shown in Figure 1, a kind of shredded leaf drying process parameter optimization method, described method step is as follows:

Step 1, remove the abnormal data and error data in the shredded leaf drying process parameter data, and obtain the process parameter data to be optimized, as shown in Table 1:

Table 1:

The data of variable x ₄ in Table 1 will no longer change from the previous data from the 8th row, and the data of variable x ₈ will be all 0s from the 9th row, so the data after the 8th row will be removed; the data of variable x ₉ All are 0, so all the data of the variable x ₉ are eliminated, and the process parameter data to be optimized shown in Table 2 are obtained:

Table 2:

Step 2, perform dimension reduction processing on the process parameter data to be optimized to obtain lightweight parameter data, such as: use the principal component analysis method to establish the following dimension reduction formula:

Principal component 1= 0.050027*x ₁ +0.020110*x ₂ -0.257952*x ₃ -0.035438*x ₄ +0.386489*x ₅

+0.443235*x ₆ -0.000825*x ₇ +0.144985*x ₈ +0.014484*x ₁₀ -0.004486*x ₁₁ -0.027720*x ₁₂

-0.051757*x ₁₃ +0.080297*x ₁₄ +0.013425*x ₁₅ -0.018750*x ₁₆ -0.016826*x ₁₇

Principal Component 2= -0.055614*x ₁ -0.030874*x ₂ +0.506334*x ₃ -0.012715*x ₄ -0.103676*x ₅

-0.210429*x ₆ -0.052316*x ₇ +0.222817*x ₈ +0.015285*x ₁₀ -0.015888*x ₁₁ +0.029754*x ₁₂

+0.049319*x ₁₃ +0.310010*x ₁₄ +0.283559*x ₁₅ -0.037438*x ₁₆ -0.005331*x ₁₇

Principal component 3= -0.186842*x ₁ +0.033591*x ₂ -0.069666*x ₃ -0.150656*x ₄ -0.002547*x ₅

+0.013832*x ₆ +0.018056*x ₇ -0.102104*x ₈ +0.391237*x ₁₀ -0.316293*x ₁₁ -0.102645*x ₁₂

+0.029974*x ₁₃ +0.073417*x ₁₄ +0.137466*x ₁₅ +0.094621*x ₁₆ +0.404656*x ₁₇

Substitute the data of process parameters x1~x17 to be optimized obtained in step 1 (except _x9 ) into the dimensionality reduction formulas of principal components 1~principal components 3, and obtain the light weight parameter data and leaf silk moisture content training data shown in Table 3:

table 3:

Step 3: Create an initial BP neural network, and substitute the lightweight parameter data and leaf silk moisture content training data into it for training to obtain a lightweight data-driven prediction model, such as: use principal components 1 to 3 as input variables, y is substituted into the neural network as an output variable for training, and the mapping relationship between principal components 1~principal components 3 and y constructed by it is a lightweight data-driven prediction model;

Step 4: Randomly screen the data of process parameters to be optimized to obtain a population of shredded drying process parameters. For example, randomly screen one data from the data of process parameters to be optimized x1~x17 (except x ₉ ) obtained in step 1 to obtain a population Repeat the screening process 5 times to obtain 5 individuals. These 5 individuals form the shredded leaf drying process parameter population, as shown in Table 4:

Table 4:

Step 5, perform dimensionality reduction processing on the individuals in the cut leaf drying process parameter population, such as: Substitute five individuals in the cut leaf drying process parameter population in step 4 into the dimensionality reduction formula in step 2, and the results are shown in Table 5:

table 5:

Step 6. Substitute the reduced population individual data into the lightweight data-driven prediction model to obtain the predicted value y _i of the moisture content of leaf shreds; for example, for each set of data input from principal component 1 to principal component 3, a The predicted value y _i of moisture content of shredded leaves is shown in Table 6:

Table 6:

Step 7, convergence judgment. Specifically, the difference between the predicted value y _i of moisture content of shredded leaves and the set value y ₀ of moisture content of shredded leaves is made. If the difference is less than or equal to the convergence accuracy e , the optimal value of shredded shredded moisture content is obtained, and the optimal value of shredded shredded moisture content is output. The combination of the optimal shred drying process parameters corresponding to the excellent value, the step ends; if the difference is greater than the convergence accuracy e , then update the shredded shred drying process parameter population, and repeat steps 5 to 7 until |y _i - y ₀ | ≤ e . For example: take e = 0.001 for the convergence accuracy, and take y ₀ = 12.7 for the moisture content of the shredded leaf. When y _i = 12.699915, |y _i - y ₀ | =0.000585≤0.001, then y _i = 12.699915 is the shredded leaf The optimal value of water content, and the corresponding data of individual 3 leaf shred drying process parameters x1~x17 before dimensionality reduction (except x ₉ ) is the combination of optimal process parameters, and the step ends; when |y _i - y When ₀ | ≥ 0.001, if the combination of the optimal value of leaf silk moisture content and optimal process parameters is not obtained, the genetic algorithm is used to perform selection, crossover, and mutation operations on individuals 1 to 5 to generate new individuals and populations , repeat steps 5 to 7 using new individuals and new populations until the convergence conditions are met.

The specific implementation of the present invention has been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned implementation, within the knowledge of those of ordinary skill in the art, it can also be made without departing from the gist of the present invention. Variations.

Claims (3)

1. a leaf shred drying process parameter optimization method is characterized in that: the method steps are as follows: Step 1, removing abnormal data and error data in the shredded leaf drying process parameter data to obtain the process parameter data to be optimized; Step 2. Perform dimensionality reduction processing on the process parameter data to be optimized to obtain lightweight parameter data; Step 3. Create an initial BP neural network, substitute lightweight parameter data and silk moisture content training data into it for training, and obtain a lightweight data-driven prediction model; Step 4, performing random screening on the process parameter data to be optimized to obtain the silk drying process parameter population; Step 5, performing dimensionality reduction operations on individuals in the shredded leaf drying process parameter population; Step 6. Substituting the dimensionally reduced population individual data into the lightweight data-driven prediction model to obtain the predicted value y _i of the moisture content of shredded leaves; Step 7, Convergence Judgment: Make a difference between the predicted value y _i of the water content of the shredded leaf and the set value y ₀ of the water content of the shredded leaf, if the difference is less than or equal to the convergence accuracy e , then output the predicted value of the water content of the shredded leaf Individuals in the population corresponding to the silk drying process parameters, otherwise not output. 2. The process parameter optimization method for leaf shreds drying according to claim 1, characterized in that: the dimensionality reduction process adopts principal component analysis method. 3. The process parameter optimization method for shredded leaf drying according to claim 1, characterized in that: in step 7, if not output, update the shredded leaf drying process parameter population, and repeat steps 5 to 7 until |y _i - y ₀ | ≤ e .