A method and system for intelligent reverse optimization control of moisture at a cut tobacco outlet

Abstract

The application discloses a kind of intelligent reverse optimization control method and system of moisture of cut tobacco outlet, belong to industrial intelligent control technical field, wherein method includes: obtaining the historical processing parameters of cut tobacco outlet moisture;The historical processing parameters are preprocessed;Feature extraction is carried out to the historical processing parameters after preprocessing, and parameter matrix is constructed according to the extracted features;Discrete processing is carried out to cut tobacco outlet moisture using K-means discrete method, and multiple discrete center values are obtained, each discrete center value corresponds to a label;Parameter matrix is input into the trained outlet moisture content prediction model, and the probability value conforming to each label is output;According to the probability value, the set parameters of cut tobacco outlet moisture in actual operation are backstepped;The method realizes the reverse optimization of outlet moisture influence parameters in cut tobacco drying process, provides a new optimization tool for actual industrial production, and also provides a reference for the application of deep learning technology in solving actual industrial production problems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial intelligent control, and more particularly to a method and system for intelligent reverse optimization control of moisture at the outlet of a cut tobacco dryer. BACKGROUND

[0002] In the process of cigarette manufacturing, the process of making tobacco is particularly important, directly affecting the quality and sensory experience of the product. In particular, the control of the moisture content at the outlet of the tobacco drying process has an important influence on the quality control of the subsequent cooling, blending and flavoring processes, as well as the efficiency of the entire tobacco making process. However, the tobacco drying process involves numerous processing procedures and has strong correlations, so it is difficult to establish an accurate numerical model. How to accurately predict and control the outlet moisture content has become a challenging task. Existing research mainly relies on equipment adjustment or uses linear methods for optimization. These methods often fail to achieve ideal results when dealing with complex, nonlinear processing parameters. At the same time, due to the difficulty in uncovering the complex relationship between the outlet moisture content and various processing parameters, the optimization results are often limited. Therefore, simple planning and empirical adjustment methods cannot meet the needs of production. More challenging is how to start from the desired outlet moisture content (e.g., 13.0000%) and reverse constrain and optimize each influencing factor to provide efficient and accurate data support for management personnel to adjust equipment parameters and process parameters, which is an important concern in the tobacco making process.

[0003] In addition, the widely used PID control method has challenges when dealing with systems with hysteresis, as it relies on immediate feedback of errors. If the system has a large hysteresis, such as the cylinder wall and outlet moisture temperature in the tobacco making process, the adjustment of the PID controller may cause overshoot or oscillation, affecting the stability of the system. Moreover, traditional PID control methods are difficult to effectively control complex systems with multiple influencing factors. With the application of digital technology in production, a large amount of on-site process data has been obtained. From the perspective of data, new optimization methods are sought to better handle the outlet moisture control problem and the constraint problem of influencing parameters in the tobacco drying process.

[0004] Currently, in the process of cigarette tobacco making, the control of outlet moisture content mainly relies on the experience of workers to adjust various parameters to regulate the moisture content of tobacco leaves or tobacco. However, due to the complexity of the production process and the presence of multi-source data disturbances, nonlinearity and high coupling, the control effect of moisture content is not ideal. The following are some problems faced by intelligent control of outlet moisture content in the current cigarette tobacco making process:

[0005] ①Data quality is not high, at present in the actual production process collected original data has not been any processing, resulting in internal noise of data is larger, for subsequent modeling and analyzing data and internal information of equipment parameters exist obstacles and influence.

[0006] ②The drying process of cut tobacco involves many processing procedures and has strong correlation. Due to the internal process of the procedure, it is difficult to establish an accurate numerical model, and how to accurately predict and control the outlet moisture becomes a challenging task. The current research mainly based on equipment adjustment or linear method for optimization, but the internal process of the procedure is often nonlinear, and the nonlinear processing parameters often cannot get ideal results.

[0007] ③It is impossible to get the corresponding key equipment parameters from the target outlet moisture. At present, the production personnel set the equipment parameters in the drying link according to the production experience. However, there is no research on the relationship between parameters and parameters, and if the new production personnel is insufficient, the set value may be deviated.

[0008] Therefore, how to predict the outlet moisture of the cut tobacco dryer, effectively adjust the input process parameters, and provide a more intuitive and more accurate parameter setting method for the management personnel of the cut tobacco process are problems that need to be solved by the technical personnel in the field. SUMMARY

[0009] In view of the above problems, the present application provides a cut tobacco dryer outlet moisture intelligent reverse optimization control method and system to at least solve part of the technical problems mentioned in the background art.

[0010] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0011] On the one hand, the present application provides a cut tobacco dryer outlet moisture intelligent reverse optimization control method, comprising the following steps:

[0012] S1, obtaining the historical processing parameters of the cut tobacco dryer outlet moisture;

[0013] S2, preprocessing the historical processing parameters;

[0014] S3, performing feature extraction on the preprocessed historical processing parameters, and constructing a parameter matrix according to the extracted features;

[0015] S4, performing discretization processing on the cut tobacco dryer outlet moisture by using K-means discrete method, obtaining a plurality of discretization center values, each of which corresponds to a label;

[0016] S5, inputting the parameter matrix into the trained outlet moisture prediction model, and outputting the probability value conforming to each label;

[0017] S6, according to the probability value, backstepping the setting parameter of the moisture at the outlet of the cut tobacco machine in actual operation.

[0018] Further, the historical processing parameters include: cylinder wall temperature, hot air temperature and moisture exhaust air volume.

[0019] Further, the step S2 specifically includes:

[0020] S21, removing the abnormal values and missing values in the historical processing parameters to obtain valid data;

[0021] S22, screening the steady-state data from the valid data;

[0022] S23, moving average processing is performed on the steady-state data.

[0023] Further, the step S22 specifically includes:

[0024] According to the change point detection theory and the Schwarz information criterion, the valid data is dynamically divided into sub-sequences;

[0025] The mean and variance of the sub-sequences are calculated;

[0026] The mean and variance are compared with the preset threshold value respectively to determine whether the corresponding historical processing data is steady-state data.

[0027] Further, the step S3 specifically includes: feature extraction is performed on the steady-state data after moving average processing by random forest method to obtain a factor combination which has an influence on the moisture at the outlet of the cut tobacco machine and an influence value exceeding a preset value; and a parameter matrix is constructed according to the factor combination.

[0028] Further, in the above step S4, the moisture at the outlet of the cut tobacco machine is divided into 4 parts, wherein:

[0029] The discretization center value corresponding to label 0 is 13.0202, and the data range is [12.9995, 13.0466];

[0030] The discretization center value corresponding to label 1 is 12.9263, and the data range is [12.8090, 12.9522];

[0031] The discretization center value corresponding to label 2 is 13.0729, and the data range is [13.0466, 13.2922];

[0032] The discretization center value corresponding to label 3 is 12.9787, and the data range is [12.9525, 12.9994].

[0033] Further, in the step S5, the outlet moisture content prediction model is a double-layer LSTM model based on an attention mechanism.

[0034] Further, the double-layer LSTM model comprises an upper-layer LSTM network and a lower-layer LSTM network.

[0035] The upper-layer LSTM network utilizes the hidden information of the lower-layer LSTM network to capture the context-dependent relationship of data.

[0036] The hidden layer of the upper-layer LSTM network is connected to the hidden layer of the lower-layer LSTM network in a fully connected manner.

[0037] Further, the attention mechanism is a soft attention mechanism.

[0038] In another aspect, the embodiments of the present application also provide a tobacco cutter outlet moisture intelligent reverse optimization control system, which applies the above method, and the system comprises an acquisition module, a preprocessing module, a feature extraction module, a discretization processing module, a calculation module and an analysis module.

[0039] The acquisition module is configured to acquire historical processing parameters of the outlet moisture of the tobacco cutter.

[0040] The preprocessing module is configured to preprocess the historical processing parameters.

[0041] The feature extraction module is configured to perform feature extraction on the preprocessed historical processing parameters, and construct a parameter matrix according to the extracted features.

[0042] The discretization processing module is configured to perform discretization processing on the outlet moisture of the tobacco cutter by using a K-means method, and obtain a plurality of discretization center values, each of which corresponds to a label.

[0043] The calculation module is configured to input the parameter matrix into a trained outlet moisture content prediction model, and output a probability value corresponding to each label.

[0044] The analysis module is configured to inversely deduce the set parameters of the outlet moisture of the tobacco cutter in actual operation according to the probability value.

[0045] According to the above technical solutions, compared with the prior art, the present application provides a tobacco cutter outlet moisture intelligent reverse optimization control method and system, which has the following technical effects:

[0046] 1. Currently, when production personnel set the key equipment parameters for leaf drying, they usually rely on past experience and lack full utilization of data. However, this invention fully mines the information within the data and uses information from past production data to determine the setting of key parameters, providing managers of the leaf-making process with a more intuitive and accurate parameter setting method.

[0047] 2. Deep learning has broad application potential in various fields. Its advantage lies in its ability to fully utilize the information in the data, while capturing the impact of time on the target and the nonlinear relationships between data. Based on deep learning, this invention can use past export moisture data to reverse control the setting of key parameters, reducing the risk of control errors caused by insufficient experience of production workers when setting parameters. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the intelligent reverse optimization control method for moisture content at the outlet of a wire drying machine provided in an embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of the steady-state data filtering process provided in an embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of a two-layer LSTM model structure provided in an embodiment of the present invention.

[0052] Figure 4 A schematic diagram of the soft attention mechanism structure provided in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] See Figure 1 As shown in the figure, an embodiment of the present invention discloses an intelligent reverse optimization control method for moisture content at the outlet of a yarn drying machine, comprising the following steps:

[0055] S1, acquire historical processing parameters of the moisture content at the outlet of the cut tobacco dryer;

[0056] S2, preprocess the historical processing parameters;

[0057] S3, extract features from the preprocessed historical processing parameters, and construct a parameter matrix according to the extracted features;

[0058] S4, discretize the moisture content at the outlet of the cut tobacco dryer by using the K-means method, and obtain a plurality of discretization center values, each of which corresponds to a label;

[0059] S5, input the parameter matrix into the trained outlet moisture content prediction model, and output probability values corresponding to each label;

[0060] S6, according to the probability values, back-propagate the set parameters of the moisture content at the outlet of the cut tobacco dryer in actual operation.

[0061] The method systematically studies the reverse optimization problem of the influence parameters of the moisture content at the outlet of the cut tobacco dryer in the tobacco manufacturing process from the data perspective. Based on this, the method provides a new optimization tool for actual industrial production and provides a reference for the application of deep learning technology in solving actual industrial production problems.

[0062] Next, the above steps will be described in detail.

[0063] In the above step S1, the historical processing parameters of the moisture content at the outlet of the cut tobacco dryer, such as the cylinder wall temperature, hot air temperature, and moisture removal air volume, are directly acquired from the cut tobacco dryer device.

[0064] In the above step S2, the effective data and steady-state data are mainly acquired, which specifically includes:

[0065] S21, remove the abnormal values and missing values in the historical processing parameters, and acquire the effective data in each batch of historical processing data;

[0066] S22, filter the steady-state data from the effective data; this step can be seen from Figure 2 According to the change point detection theory and the Schwarz information criterion, the effective data is dynamically divided into subsequences; the mean and variance of the subsequences are calculated; the mean and variance are compared with the preset threshold, respectively, to determine whether the corresponding historical processing data is steady-state data; wherein the judgment condition corresponding to the change point detection theory is represented as:

[0067]

[0068] wherein j represents the change point position; m represents the number of change points and m∈{1,...,T-1}) ; T represents the time of each change point; C represents the cost function; τj represents the jth turning point; and βf(m) represents a penalty term for preventing overfitting.

[0069] Since the online collection of production process inevitably mixes a large amount of non-steady-state data, if the non-steady-state data cannot be effectively identified, it will inevitably affect the objective judgment of the actual production process and the accuracy of process control diagnosis; therefore, the identification of the steady-state data in the step S22 is crucial to guarantee the objectivity of quality evaluation and the accuracy of quality control.

[0070] S23, moving average processing is performed on the steady-state data; taking the moisture value as an example, when the cut tobacco enters the tobacco cutter, the moisture value Y is measured, and after passing through the tobacco cutter, the moisture value Y' is obtained at the outlet; however, in the whole process, the conveying belt continuously conveys the cut tobacco to the tobacco cutter, which will result in Y'≠Y, and the point-to-point problem of the moisture value cannot be solved; therefore, the moving average method is needed to replace the original point value with the average value to solve the point-to-point problem in the actual production data; in the embodiment of the present application, the moving average processing can be represented as:

[0071]

[0072] represents the outlet moisture value after the moving average calculation; Y' u represents the original outlet moisture value at the u time; and k represents the size of the set backward moving average.

[0073] In the above step S3, a large number of parameters are involved in the collected data, however, only part of the equipment parameters have an impact on the outlet moisture in actual production, therefore, feature screening becomes crucial, in the embodiment of the present application, the moving average processed steady-state data is subjected to feature extraction by the random forest method, to obtain a factor combination that has an impact on the outlet moisture of the tobacco cutter, and the impact value exceeds a preset value; a parameter matrix is constructed according to the factor combination.

[0074] In the above step S4, since the quality of the cut tobacco after the drying process is not a single outlet moisture value, but a range or interval in actual production, the outlet moisture of different intervals is adjusted to realize accurate control of the process parameters of the tobacco cutting process; in order to better process this characteristic, the K-means method is adopted in the embodiment of the present application to discretize the outlet moisture, so as to better reflect the characteristics of the tobacco cutting process; the following formula is used to update the cluster center in the K-means discretization process, so as to optimize the discretization effect:

[0075]

[0076] wherein, m b represents the average value of the data points contained therein; and Cb Represents the b-th cluster center; |C b | Represents the cluster center C b The number of data points included; x l This represents the l-th data point;

[0077] Based on the K-means discretization method described above, in this embodiment of the invention, the moisture content at the outlet of the drying machine is discretized into four parts. Considering actual production needs, label "0" is used as the standard value for the moisture content at the outlet of the drying machine. Labels "1" and "3" are positioned as outlet moisture values ​​below the standard value. Label "2" is defined as an outlet moisture value above the standard value. The discretization center value of label "0" is the center value of the standard value for the moisture content at the outlet of the drying machine. See Table 1 below for details.

[0078] Table 1 Discretization Criteria and Scope

[0079] Tag value Range Discretization center value 0 [12.9995,13.0466] 13.0202 1 [12.8090,12.9522] 12.9263 2 [13.0466,13.2922] 13.0729 3 [12.9525,12.9994] 12.9787

[0080] In step S5 above, the present invention constructs an intelligent control model for the outlet moisture content of key processes based on a two-layer LSTM model with an attention mechanism, namely the outlet moisture content prediction model (A-DLSTM model); this model can also be applied to the outlet moisture content control of each key process segment in the silk making process.

[0081] The two-layer LSTM model used in this embodiment of the invention can better learn information and hidden contextual relationships in the data, reduce the risk of overfitting the model to the training data, and improve the model's generalization ability on unseen data. Its structure diagram can be found in [reference needed]. Figure 3 As shown; in Figure 3 In the text, x represents the input information. t+1 y represents the input information at time t+1; y represents the output result. t+1 This represents the output result at time t+1. This represents the state of the hidden layer of the lower LSTM at time t+1. This represents the state of the hidden layer of the upper LSTM at time t+1;

[0082] The parameter matrix constructed in step S3 above is used as the input to the outlet moisture content prediction model, and the formula is used. The importance of input information compared to previous information is determined by c. t f represents the cell state at time t. t Represents the forgetting gate at time t, i t This represents the input gate at time t. Represents the candidate cell state at time t; and uses h t =o t ⊙tanh(c t) to determine whether to update the information of the hidden layer, wherein o t represents the output gate of the t time, represents dot multiplication, and tanh(·) represents an activation function.

[0083] The double-layer LSTM model comprises an upper-layer LSTM network and a lower-layer LSTM network; the output of the upper-layer LSTM network is used as the input of the lower-layer LSTM network, forming a hierarchical structure; the structure allows the model to learn and process the features of data independently; in addition, the upper-layer LSTM network uses the hidden information of the lower-layer LSTM network to capture the up-down dependency of data; the hidden layer of the upper-layer LSTM network is connected with the hidden layer of the lower-layer LSTM network through a full connection layer; the full connection layer is used for outputting the prediction interval of the exit moisture; wherein the output of the hidden layer state of the upper layer is represented as:

[0084]

[0085] wherein W2 represents a weight matrix between the lower-layer LSTM network and the upper-layer LSTM network; b2 represents a bias vector between the lower-layer LSTM network and the upper-layer LSTM network. represents the hidden layer of the upper-layer LSTM network. represents the hidden layer of the lower-layer LSTM.

[0086] In the embodiment of the application, an attention model is introduced on the basis of the double-layer LSTM model, which is used for learning the dependency in the sequence to enhance the generalization ability of the model, so that the exit moisture content prediction model can pay more attention to the important part; in the embodiment of the application, a soft attention mechanism is used, and a structural diagram thereof can be seen from Figure 4 . Figure 4 In the formula, parameter Z represents a scoring function, and parameter q represents a query vector; the principle is that assuming that a certain system comprises n input information, the attention weight of the input information at each time step is represented as:

[0087]

[0088] wherein n represents the number of input information of data; β p is the attention distribution when the number of input information is p; w p is a splicing vector when the number of input information is p; p<=n.

[0089] In the step S6, the reverse control of the device parameters is realized by using the forward outlet moisture prediction model; by establishing the forward outlet moisture content prediction model, the setting value of the key device parameters can be deduced reversely; in order to test the effect of the step, the outlet moisture content prediction model trained on the training set can be applied to the test set, so that the label number output, the maximum probability value of the target outlet moisture on the test set is obtained, and the corresponding sample value is the reference value required by the embodiment of the application;

[0090] In actual production, the specific point value cannot fully reflect the actual production situation, therefore, the mean value of the device sample value is calculated to optimize the estimated value of the influence factor in the present application; the method can more accurately reflect the influence of the device parameters on the outlet moisture, and provides strong support for the control and optimization of the production process.

[0091] In another aspect, the embodiment of the application also provides an intelligent reverse optimization control system for the outlet moisture of a cut tobacco dryer, which applies the method described above, and the system comprises an acquisition module, a preprocessing module, a feature extraction module, a discretization processing module, a calculation module and an analysis module; wherein:

[0092] The acquisition module is used to acquire the historical processing parameters of the outlet moisture of the cut tobacco dryer;

[0093] The preprocessing module is used to preprocess the historical processing parameters;

[0094] The feature extraction module is used to extract features from the preprocessed historical processing parameters, and construct a parameter matrix according to the extracted features;

[0095] The discretization processing module is used to discretize the outlet moisture of the cut tobacco dryer by using the K-means method, and obtain a plurality of discretization center values, each of which corresponds to a label;

[0096] The calculation module is used to input the parameter matrix into the trained outlet moisture content prediction model, and output the probability value corresponding to each label;

[0097] The analysis module is used to back-calculate the setting parameters of the outlet moisture of the cut tobacco dryer in actual operation according to the probability value.

[0098] In summary, the tobacco dryer outlet moisture intelligent reverse optimization control method and system provided by the embodiment of the application, in the outlet moisture content prediction model composed of the attention mechanism and the double-layer LSTM model, the feature parameter matrix is taken as the input, the outlet moisture of the tobacco dryer is predicted, and the set values of each influencing factor are backstepped. The method proposed in the paper can not only directly consider the target value of the outlet moisture, but also can handle multiple complex parameters to adjust the input process parameters. This strategy can provide more intuitive and accurate parameter setting method for the managers of the tobacco drying process, thereby improving the efficiency and quality of the tobacco drying process.

[0099] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0100] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent reverse optimization control of moisture at the outlet of a cut tobacco dryer, characterized in that, The method comprises the following steps: S1, obtaining historical processing parameters of the moisture content at the outlet of the cut tobacco dryer; S2, preprocessing the historical processing parameters; S3, extracting features from the preprocessed historical processing parameters, and constructing a parameter matrix according to the extracted features; S4, discretizing the moisture content at the outlet of the cut tobacco dryer by using the K-means discrete method to obtain a plurality of discrete center values, each of which corresponds to a label; S5, inputting the parameter matrix into a trained outlet moisture content prediction model to output a probability value corresponding to each label; S6, according to the probability value, backstepping the setting parameters of the moisture content at the outlet of the cut tobacco dryer during actual operation; The historical processing parameters include: cylinder wall temperature, hot air temperature and moisture removal air volume; The step S2 specifically comprises: S21, removing outliers and missing values in the historical processing parameters to obtain valid data; S22, screening steady-state data from the valid data; S23, moving average processing of the steady-state data; The step S22 specifically comprises: According to the change point detection theory and the Schwarz information criterion, the valid data is dynamically divided into sub-sequences; Calculate the mean and variance of the sub-sequences; Compare the mean and variance with the preset threshold value respectively to determine whether the corresponding historical processing data is steady-state data; The step S3 specifically comprises: extracting features from the moving average processed steady-state data by using the random forest method to obtain a factor combination that has an impact on the moisture content at the outlet of the cut tobacco dryer, and the impact value exceeds a preset value; and constructing a parameter matrix according to the factor combination; In the above step S4, the moisture content at the outlet of the cut tobacco dryer is discretized into 4 parts, wherein: The discrete center value corresponding to label 0 is 13.0202, and the data range is [12.9995, 13.0466]; The discrete center value corresponding to label 1 is 12.9263, and the data range is [12.8090, 12.9522]; The discrete center value corresponding to label 2 is 13.0729, and the data range is [13.0466, 13.2922]; The discrete center value corresponding to label 3 is 12.9787, and the data range is [12.9525, 12.9994]; In the above step S5, the outlet moisture content prediction model is a double-layer LSTM model based on attention mechanism.

2. The intelligent moisture reverse optimization control method for a tobacco cutter outlet according to claim 1, characterized in that, The double-layer LSTM model comprises an upper-layer LSTM network and a lower-layer LSTM network; The upper-layer LSTM network uses the hidden information of the lower-layer LSTM network to capture the up-down dependency of the data; The hidden layer of the upper-layer LSTM network is connected to the hidden layer of the lower-layer LSTM network in a fully connected manner.

3. The intelligent moisture reverse optimization control method for a tobacco cutter outlet according to claim 1, characterized in that, The attention mechanism is a soft attention mechanism.

4. A smart reverse optimization control system for moisture content at the outlet of a yarn drying machine, characterized in that, The system comprises an obtaining module, a preprocessing module, a feature extraction module, a discretization processing module, a calculation module and an analysis module; The obtaining module is configured to obtain historical processing parameters of the moisture content at the outlet of the cut tobacco dryer; The preprocessing module is configured to preprocess the historical processing parameters; The feature extraction module is configured to perform feature extraction on the preprocessed historical processing parameters, and construct a parameter matrix according to the extracted features; The discretization processing module is configured to perform discretization processing on the moisture at the outlet of the cut tobacco dryer by using a K-means method, and obtain a plurality of discretization center values, each of which corresponds to a label; The calculation module is configured to input the parameter matrix into a trained outlet moisture content prediction model, and output a probability value corresponding to each label; The analysis module is configured to back-calculate a set parameter of the moisture at the outlet of the cut tobacco dryer during actual operation according to the probability value.

Images