Method and device for predicting temperature and moisture of materials at outlet of cut tobacco dryer

Abstract

The invention discloses a method and a device for predicting temperature and moisture of materials at an outlet of a cut-tobacco drier, and belongs to the technical field of temperature and moisture prediction of materials at the outlet of the cut-tobacco drier. Inputting the obtained variables into a trained random forest model based on multi-output regression to obtain preliminary predicted temperature and preliminary predicted moisture; discrete auxiliary data generated in the cut tobacco drying process of the cut tobacco dryer are obtained; inputting the obtained variables, the preliminary predicted moisture and the discrete auxiliary data into a trained first BP neural network model to obtain final predicted moisture; inputting the obtained variables, the preliminary prediction temperature and the discrete auxiliary data into a trained second BP neural network model to obtain a final prediction temperature; according to the method, the moisture and temperature are preliminarily predicted through regression fitting, the influence is corrected through the BP neural network model, various errors of predicted values are reduced, and the prediction precision is improved.

Description

Technical Field

[0001] The invention relates to a method and a device for predicting the temperature and moisture of materials at the outlet of a torrefaction machine, belonging to the technical field of predicting the temperature and moisture of materials at the outlet of a torrefaction machine. Background Art

[0002] The working principle commonly used by cigarette cutting and drying machines is: a certain flow of steam passes through the inner wall of the thin plate of the cylinder, 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, heated, and the filling force is evenly increased, as well as the outlet moisture and moisture. 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] Since the drying process has strong nonlinearity, uncertainty and large hysteresis, coupled with the special properties of tobacco leaves themselves, the material temperature control in the drying process becomes very complicated. In essence, the moisture and temperature control of the materials in the drying machine belong to a nonlinear time-varying large hysteresis system, and there are many interference factors affecting its control, all of which are time-varying nonlinear. The common method of predicting the outlet material of the drying machine is to use big data analysis and machine learning principles to only predict a single value of the outlet moisture or outlet temperature of the drying machine, and predict the outlet moisture and outlet temperature separately. For example, patent (a method for predicting the outlet moisture of the drying machine based on a multivariate linear regression model CN202311493060.7), patent (a prediction model for the wall temperature of the drying machine based on the moisture content of the cut leaf CN201510226105.3), patent (a dynamic prediction method for moisture content of the drying machine based on the dual attention mechanism and XGBoost CN202311506756.9), etc.

[0004] Due to the complex internal mechanism of the tofu drying machine and the overlapping interference factors affecting the moisture and temperature of the materials, and the different degrees of influence of each factor; the moisture and temperature of the materials are also coupled to a certain extent within the process control, and the process of their mutual restriction and influence is also relatively complex. Therefore, the use of different single models to predict the moisture and temperature of the materials lacks their mutual correlation and also ignores the coupling between the two in the prediction stage. Therefore, the prediction accuracy of the single prediction model will be subject to certain limitations. Summary of the invention

[0005] The object of the present invention is to provide a method and device for predicting the temperature and moisture of materials at the outlet of a tofu drying machine, so as to solve the problem of low prediction accuracy in the prior art.

[0006] To achieve the above objectives, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a method for predicting the temperature and moisture content of materials at the outlet of a tofu drying machine, comprising: Acquire the operating variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, the disturbance variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, and the disturbance variables that have an indirect relationship with the predicted result on the mechanism of the tofu drying machine; The obtained variables are input into the trained random forest model based on multi-output regression to obtain the preliminary predicted temperature and preliminary predicted moisture of the material at the outlet of the tofu drying machine; Acquire discrete auxiliary data generated by the drying process of the tofu drying machine; Input the acquired variables, preliminary predicted moisture content and discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture content of the material at the outlet of the tofu drying machine; The obtained variables, preliminary predicted temperature and discrete auxiliary data are input into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

[0007] Furthermore, the operating variables that are directly related to the predicted results in the mechanism of the wire drying machine include the actual value of the drum wall temperature, the actual value of the hot air temperature and the actual value of the hot air volume; the disturbance variables that are directly related to the predicted results in the mechanism of the wire drying machine include: the actual value of the incoming material moisture, the actual value of the incoming material temperature, the actual value of the incoming material flow rate, the steam dryness value and the negative pressure value of the dehumidification fan; the disturbance variables that are indirectly related to the predicted results in the mechanism of the wire drying machine include the ambient temperature value, the ambient humidity value and the steam temperature value; the discrete auxiliary data include the production stage, the production operation time schedule and the number of batches produced on the day.

[0008] Furthermore, the random forest model based on multi-output regression is trained by the following method: Obtain the following data from the historical data: operating variables that are directly related to the prediction results on the mechanism of the wire drying machine, disturbance variables that are directly related to the prediction results on the mechanism of the wire drying machine, and disturbance variables that are indirectly related to the prediction results on the mechanism of the wire drying machine, and obtain the wire drying machine outlet material temperature and wire drying machine outlet material moisture corresponding to the above variables; Based on the measurement time of the material temperature and moisture content at the outlet of the wire drying machine, the obtained variables are aligned in time series; The Pearson correlation coefficients of the disturbance variables with indirect relationship in the mechanism of the torrefaction machine and the outlet material temperature and moisture content of the torrefaction machine are calculated and predicted respectively. If the two Pearson correlation coefficients of a disturbance variable are both less than 0.3, the disturbance variable is discarded. The random forest model based on multi-output regression was trained using the time-series aligned and screened variables, the outlet material temperature of the tofu drying machine, and the outlet material moisture of the tofu drying machine. The training ended when the degree of fit was greater than 0.9.

[0009] Furthermore, the variables acquired based on the measurement time of the temperature of the material at the outlet of the wire drying machine and the moisture of the material at the outlet of the wire drying machine are aligned in time sequence, including: The measurement time of the outlet material temperature and moisture content of the wire drying machine is taken as the timing reference point, and all variables are moved backward in the timing until they are aligned with the timing reference point.

[0010] Furthermore, the measurement time of the temperature of the material at the outlet of the wire drying machine and the moisture of the material at the outlet of the wire drying machine is used as a time series reference point, and all variables are moved backward in the time series until they are aligned with the time series reference point, including: The time from the moisture meter in the drying process to the drying machine inlet is obtained as the first delay time. t 1; The time from the inlet of the wire drying machine to the outlet of the wire drying machine during the wire drying process is obtained as the second delay time t 2; The time from the exit of the silk drying machine to the exit moisture meter during the silk drying process is obtained as the third delay time t 3; Calculating a preliminary time series delay time for each variable to move backward in time series based on the first delay time, the second delay time and the third delay time; Based on the Pearson correlation principle and the preliminary time series delay time, the precise time series alignment delay correction time of each variable moving backward in time series is calculated; The preliminary timing delay time and the precise timing alignment delay correction time corresponding to each variable are added together to obtain the final delay time corresponding to each variable, and each variable is moved backward in timing by the final delay time to complete the timing alignment of each variable.

[0011] Furthermore, the calculation of the preliminary time series delay time for each variable to move backward in time series based on the first delay time, the second delay time and the third delay time is performed by the following method: The initial time delay of the actual value of the cylinder wall temperature, the actual value of the hot air temperature and the actual value of the hot air volume is: t 2 / 2+ t 3. The initial sequence delay time of the actual value of incoming material moisture, actual value of incoming material temperature and actual value of incoming material flow is: t 1+ t 2 +t 3. The initial timing delay time of steam dryness value is t 2 +t 3. The initial timing delay time of the negative pressure value of the dehumidification fan ist 3. The initial timing delay time of the ambient temperature value and the ambient humidity value is t 2 / 2+ t 3. The initial timing delay time of the steam temperature value is t 2+ t 3.

[0012] Furthermore, the calculation of the precise time series alignment delay correction time of each variable moving backward in time series based on the Pearson correlation principle and the preliminary time series delay time is performed by the following method: ; in, i N It indicates the precise timing alignment delay correction time of the Nth variable when the sum of the Pearson correlation coefficients between the Nth variable and the two dependent variables is the largest. N indicates the sequence number of the variable, arg max indicates the arg max function, and corr(a,b) indicates the Pearson correlation coefficient between a and b. T max Indicates the maximum change value of the initial timing delay time, y 1 indicates the moisture content of the material at the outlet of the silk drying machine. y 2 indicates the outlet temperature of the fiber drying machine. Indicates moving backward in time The Nth variable after Indicates the initial timing delay time of the Nth variable.

[0013] Furthermore, the first BP neural network model is trained by the following method: Obtain the following data from the historical data: operating variables that have a direct relationship with the prediction result on the mechanism of the wire drying machine, disturbance variables that have a direct relationship with the prediction result on the mechanism of the wire drying machine, and disturbance variables that have an indirect relationship with the prediction result on the mechanism of the wire drying machine, moisture content of the wire drying machine outlet material corresponding to the above variables, and discrete auxiliary data corresponding to the above variables; The moisture content of the material at the outlet of the tofu drying machine is used as the output of the first BP neural network model, and other data are used as input. The first BP neural network model is reversely trained, and the tanh function is used as the activation function during the reverse training process.

[0014] Furthermore, the second BP neural network model is trained by the following method: Obtain the following data from the historical data: operating variables that have a direct relationship with the predicted result on the mechanism of the torrefaction machine, disturbance variables that have a direct relationship with the predicted result on the mechanism of the torrefaction machine, and disturbance variables that have an indirect relationship with the predicted result on the mechanism of the torrefaction machine, the torrefaction machine outlet material temperature corresponding to the above variables, and discrete auxiliary data corresponding to the above variables; The outlet material temperature of the tofu drying machine is used as the output of the second BP neural network model, and other data are used as input. The second BP neural network model is reversely trained, and the tanh function is used as the activation function during the reverse training process.

[0015] In a second aspect, the present invention provides a device for predicting the temperature and moisture content of materials at the outlet of a tofu drying machine, comprising: The variable acquisition module is configured to: acquire the operating variables that are directly related to the prediction result on the mechanism of the tofu drying machine, the disturbance variables that are directly related to the prediction result on the mechanism of the tofu drying machine, and the disturbance variables that are indirectly related to the prediction result on the mechanism of the tofu drying machine; The preliminary prediction module is configured to: input each acquired variable into a trained random forest model based on multi-output regression to obtain a preliminary predicted temperature and a preliminary predicted moisture content of the material at the outlet of the tofu drying machine; The auxiliary data acquisition module is configured to: acquire discrete auxiliary data generated during the drying process of the tofu drying machine; The moisture prediction module is configured to: input the acquired variables, the preliminary predicted moisture and the discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture of the material at the outlet of the tofu drying machine; The temperature prediction module is configured to: input the acquired variables, preliminary predicted temperature and discrete auxiliary data into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

[0016] Compared with the prior art, the beneficial effects achieved by the present invention are: The present invention provides a method and device for predicting the temperature and moisture of the outlet material of the tofu drying machine. The method and device are used to regress and fit the moisture and temperature of the outlet material of the tofu drying machine through a random forest model based on multi-output regression. At the same time, a BP neural network model based on multiple input factors is used to correct the influence of factors that are unpredictable or have large estimated deviations on the predicted value in dimensions such as coupling and uncertainty, thereby reducing the coupling interference error of the controlled model and the inherent error of the machine learning model, thereby improving the accuracy of the predicted value. The present invention combines a variety of predictable factors and simultaneously fits and predicts the moisture and temperature of the material at the outlet of the tofu drying machine through a random forest model based on multi-output regression. Different from the principle of using different separate machine learning models to fit a single output factor separately in other invention methods, the combination of a complex improved decision tree and dual outputs is used, which better fits the phenomenon that the internal mechanism of the tofu drying machine is complex and the interference factors affecting the moisture and temperature of the material have overlapping effects. It can also weaken the mechanism process of coupling between the moisture and temperature of the material in the process control to a certain extent, so that the process of their mutual restriction and influence is fully fitted by the tree structure model, and at the same time improves the prediction accuracy of the two export indicators; The present invention divides the independent variables into three categories for analysis, and performs Pearson correlation analysis on the indirect disturbance variables of the third category for selection, thereby ensuring the scientific nature of the data and the linkage of the mechanism, so that the overall algorithm starts searching from relatively accurate data, greatly reducing the computational overhead of data search, and also improving the effect of model convergence, preventing the data from falling into the local optimum within other imprecise data ranges; The present invention also performs a time series analysis of the data, uses a data initial alignment scheme to approximate the precise independent variable time series array, and uses the sum of the Pearson correlation coefficients between the independent variable and the two dependent variables to perform precise time series alignment, so that the data of the method of the present invention has the highest correlation in the time series relationship, which is convenient for subsequent model training and ensures the accuracy of the model; After using the random forest model based on multi-output regression for prediction, the present invention also uses the first BP neural network model and the second BP neural network model to correct the errors of the outlet moisture and the outlet temperature respectively. This method utilizes the reverse training model principle of the BP neural network and continuously reduces the error between the predicted value and the actual value by updating the sample size data, preventing the random forest model based on multi-output regression from falling into local optimization, and also can locally improve the accuracy of the two prediction quality indicators respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for predicting the temperature and moisture content of materials at the outlet of a tofu drying machine corresponding to Embodiment 1 of the present invention; Figure 2 It is a flow chart of a method for predicting the temperature and moisture content of materials at the outlet of a tofu drying machine corresponding to Embodiment 2 of the present invention; Figure 3 is a schematic diagram of a model G1 provided in an embodiment of the present invention; Figure 4 is a schematic diagram of a model G2 provided in an embodiment of the present invention; Figure 5 It is a schematic diagram of model G3 provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the protection scope of the present invention.

[0019] Example 1.

[0020] like Figure 1 As shown, the present invention provides a method for predicting the temperature and moisture content of the material at the outlet of a tofu drying machine, comprising: Acquire the operating variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, the disturbance variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, and the disturbance variables that have an indirect relationship with the predicted result on the mechanism of the tofu drying machine; The obtained variables are input into the trained random forest model based on multi-output regression to obtain the preliminary predicted temperature and preliminary predicted moisture of the material at the outlet of the tofu drying machine; Acquire discrete auxiliary data generated by the drying process of the tofu drying machine; Input the acquired variables, preliminary predicted moisture content and discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture content of the material at the outlet of the tofu drying machine; The obtained variables, preliminary predicted temperature and discrete auxiliary data are input into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

[0021] The random forest model based on multi-output regression, the first BP neural network model and the second BP neural network model adopted in the present invention are all common prediction models in the industry in the prior art, and no improvement is made to the model itself.

[0022] The present invention regresses and fits the moisture and temperature of the material at the outlet of the tofu drying machine through a random forest model based on multi-output regression. At the same time, a BP neural network model based on multiple input factors is used to correct the influence of factors that are unpredictable or have large estimated deviations on the predicted value in dimensions such as coupling and uncertainty, thereby reducing the predicted value's coupling interference error caused by the controlled model and the inherent error of the machine learning model, thereby improving the accuracy of the predicted value.

[0023] Example 2.

[0024] like Figure 2 As shown, the present invention provides a method for predicting the temperature and moisture content of the material at the outlet of a tofu drying machine, comprising: Step S101, selection of independent variables and dependent variables based on the mechanism of the tofu drying machine.

[0025] Based on the operating raw materials and material processing technology of the thin plate dryer, the moisture content and temperature of the outlet material of the dryer have a direct mechanistic relationship with the following operating variables: actual value of the drum wall temperature, actual value of the hot air temperature, actual value of the hot air volume; have a direct mechanistic relationship with the following disturbance variables: actual value of incoming material moisture, actual value of incoming material temperature, actual value of incoming material flow rate, steam dryness value; have an indirect mechanistic relationship with the following disturbance variables: ambient temperature value, ambient humidity value, steam temperature value.

[0026] The moisture content and temperature of the material at the outlet of the wire dryer are taken as dependent variables, which are defined as: the moisture content of the material at the outlet of the wire dryer y1 and the temperature of the material at the outlet of the wire dryer y2.

[0027] The operating variables that have a direct relationship in mechanism are defined as Type I independent variables, which are defined as: actual value of cylinder wall temperature X1, actual value of hot air temperature X2, actual value of hot air volume X3.

[0028] The disturbance variables that have a direct relationship in mechanism are defined as Type II independent variables, which are defined as: actual value of incoming moisture X4, actual value of incoming temperature X5, actual value of incoming flow rate X6, steam dryness value X7, and dehumidification fan negative pressure value X8.

[0029] The disturbance variables with indirect relationship in mechanism are defined as type III independent variables, which are defined as: ambient temperature value X9, ambient humidity value X 10 , Steam temperature value X 11 .

[0030] Step S102, analyzing the time domain delay relationship of the independent variable.

[0031] Based on the operation of the thin plate drying machine raw materials and material processing technology, X1 to X 11 There are inconsistencies in the time series of the 11 independent variables. Some independent variables are measured values ​​at the equipment entrance, and some are actual measured values ​​or indirect measured values ​​during the processing. The dependent variable is the actual value of the equipment outlet quality index. There is a time domain delay relationship between the independent variables and the dependent variables. Before establishing the prediction model, the data needs to be aligned in time series.

[0032] Step S102-1, measuring the delay time from the inlet moisture meter to the inlet of the silk drying machine t 1. Place markers with physical properties similar to the material in the production process, record the delay time of the marked material from the entrance moisture meter to the entrance of the silk drying machine multiple times, and calculate the average value, which is recorded as t 1, in seconds.

[0033] Step S102-2, measuring the delay time from the inlet to the outlet of the wire drying machine t2. Place markers with physical properties similar to the material in the production process, record the delay time of the marked material from the entrance to the exit of the dryer for many times, and calculate the average value, which is recorded as t 2, in seconds.

[0034] Step S102-3, measuring the delay time from the outlet of the silk drying machine to the moisture meter at the outlet of the silk drying machine t 3. Place markers with physical properties similar to the material in the production process, record the delay time of the marked material from the exit of the wire drying machine to the moisture meter at the exit of the wire drying machine multiple times, and calculate the average value, which is recorded as t 3. The unit is seconds.

[0035] Step S102-4, preliminary time alignment: The instantaneous time when the moisture meter at the outlet of the tofu drying machine measures the outlet moisture and outlet temperature is used as the time reference point, and all independent variables are moved backward in the time series.

[0036] definition t The moisture data of the material at the outlet of the silk drying machine at this moment is y1( t ), the outlet material temperature data is y2( t ), then there is a correlation with the following delayed data: Including: Actual value of cylinder wall temperature after delay X1( t - t 2 / 2- t 3), the actual value of hot air temperature after delay is X2( t - t 2 / 2- t 3), the actual value of hot air volume after delay is X3( t - t 2 / 2- t 3) The above three factors all assume that the measurement is taken when the material passes through the middle of the cylinder.

[0037] Also includes: the actual value of the incoming moisture after the delay X4 ( t - t 1- t 2- t 3), the actual value of the incoming material temperature after the delay is X5 ( t - t 1- t 2- t 3), the actual value of the incoming flow after delay is X6 ( t - t 1- t 2- t 3) The above three factors all assume that the measurement is taken at the material inlet moisture meter position at the moment of measurement.

[0038] It also includes: the steam dryness value after delay X7 (t - t 2- t 3), the above factors assume that the material is measured at the entrance of the dryer. The negative pressure value of the exhaust fan after the delay is X8 ( t - t 3) The above factors assume that the measurement is taken at the exit of the tofu drying machine.

[0039] Also includes: the ambient temperature value after delay X9 ( t-t 2 / 2- t 3) Ambient humidity value after delay X 10 ( t - t 2 / 2- t 3) If the two values ​​are collected from a real-time system, the principle is similar to the above; if the two values ​​are collected from a weather forecast system, no delay is required and the real-time data X9(t) and X 10 (t)

[0040] Finally, it also includes: the steam temperature value X after delay 11 ( t - t 2- t 3) The above factors assume that the measurement is taken at the entrance of the torrefaction machine.

[0041] So far, the independent variable after preliminary timing alignment is determined to be X N ( t - T dely ), N represents the independent variable number, which is any integer from 1 to 11; T dely is the initial timing delay time, based on t 1. t 2. t 3 Calculate it.

[0042] The initial timing delay time is calculated by the following method: The initial time delay of the actual value of the cylinder wall temperature, the actual value of the hot air temperature and the actual value of the hot air volume is: t 2 / 2+ t 3. The initial sequence delay time of the actual value of incoming material moisture, actual value of incoming material temperature and actual value of incoming material flow is: t 1+ t 2 +t 3. The initial timing delay time of steam dryness value is t 2 +t 3. The initial timing delay time of the negative pressure value of the dehumidification fan is t 3. The initial timing delay time of the ambient temperature value and the ambient humidity value is t 2 / 2+t 3. The initial timing delay time of the steam temperature value is t 2+ t 3.

[0043] Step S102-5: accurate timing alignment. Use the Pearson correlation principle to align the delay times in step S102-4. T dely The time unit is extended and shortened with a step of 1 second, and the Pearson correlation coefficient between the independent variable and the two dependent variables after extension and shortening by integer seconds is studied respectively.

[0044] Define the maximum extension and shortening time units as T max , the unit is seconds, then corresponding to the independent variable number N, study X N ( t - T dely - T max ),X N ( t - T dely - T max +1), X N ( t - T dely - T max +2), …, X N ( t - T dely - T max -2) X N ( t - T dely - T max -1) X N ( t - T dely ),X N ( t - T dely +1), X N ( t - T dely +2), …, X N ( t - T dely + T max -2) XN ( t - T dely + T max -1) X N ( t - T dely + T max ) The Pearson correlation coefficient between the independent variable and the two dependent variables after the above different time delay units. Among them, T dely It is the preliminary timing delay time determined in S102-4, and N represents the independent variable sequence number, which is any integer from 1 to 11.

[0045] Compare the sum of the Pearson correlation coefficients of the independent variable and the two dependent variables after different time delay units, and select the one with the largest sum of the Pearson correlation coefficients as the precise time series delay time. The expression formula is as follows: ; in, i N It indicates the precise timing alignment delay correction time of the Nth variable when the sum of the Pearson correlation coefficients between the Nth variable and the two dependent variables is the largest. N indicates the sequence number of the variable, arg max indicates the arg max function, and corr(a,b) indicates the Pearson correlation coefficient between a and b. T max Indicates the maximum change value of the initial timing delay time, y 1 indicates the moisture content of the material at the outlet of the silk drying machine. y 2 indicates the outlet temperature of the fiber drying machine. Indicates moving backward in time The Nth variable after Indicates the initial timing delay time of the Nth variable.

[0046] i N The value range is - T max To+ T max Any integer between .

[0047] At this point, the precise timing alignment delay time of each independent variable is calculated as the sum of the preliminary timing delay time and the precise timing alignment delay correction time.

[0048] Step S103, establish a random forest model based on multi-output regression (a prediction algorithm for material temperature and material moisture after drying based on random forest of multi-output regression, model G1), such as Figure 3 shown.

[0049] Step S103-1, perform correlation analysis on the disturbance variables of the third type of indirect relationship. Calculate the ambient temperature value X9, the ambient humidity value X 10 , Steam temperature value X 11 The Pearson correlation coefficient with the moisture content y1 of the dried silk machine outlet and the temperature y2 of the dried silk machine outlet is less than 0.3, and the independent variable factor is discarded. The factors of category I and category II are not deleted and are used as independent variables (i.e., inputs of the model) of the random forest model based on multi-output regression.

[0050] Step S103-2, export historical data according to the data format. Take the data collected at fixed time intervals during tobacco silk production as data rows, and export X1-X at fixed time intervals. 11 , y1, y2 values, and align the delay correction time according to the precise timing obtained in step S102, and align all historical data according to different delays.

[0051] Step S103-3, based on X N , y1 and y2, and train a random forest model based on multi-output regression. Divide the data into a data set and a test set, and ensure that the model's fit (R2 value) is greater than 0.9 when doing the test set.

[0052] Step S104, establish prediction error correction algorithms G2 and G3 based on BP neural network, G2 is the first BP neural network model (such as Figure 4 As shown in ), G3 is the second BP neural network model (as shown in Figure 5 as shown).

[0053] Step S104-1, use the multi-output regression based random forest prediction algorithm G1 for post-drying material temperature and material moisture established in step S103, and perform long-term online prediction and data recording, recording the actual average outlet moisture value and the actual average outlet temperature value, so as to compare with the predicted value of the G1 algorithm and further correct it.

[0054] Step S104-2, increase data dimension and import production stage X 12 , production operation time schedule X 13 、Number of batches produced on the day X 14 These three types of auxiliary discrete attribute data increase the data dimension of the auxiliary characteristics that indirectly affect the prediction results and improve the accuracy of the overall prediction algorithm.

[0055] Step S104-3, using the BP neural network to establish error correction models G2 and G3 for the moisture content and temperature of the material at the outlet of the tofu drying machine, respectively.

[0056] Among them, G2 is the outlet moisture error correction model. All independent variables in the prediction algorithm G1 and the prediction result y1 of G1, as well as the discrete auxiliary data added in step S104-2 are input into G2, and the actual outlet moisture average value is taken as the output of G2. The model is reversely trained to reduce the error between the predicted value y1 and the actual outlet moisture average value.

[0057] G3 is the outlet temperature error correction model. All independent variables in the prediction algorithm G1 and the prediction result y2 of G1, as well as the discrete auxiliary data added in step S104-2 are input into G3, and the actual outlet moisture average value is used as the output of G3. The model is trained in reverse order to reduce the error between the predicted value y2 and the actual outlet moisture average value.

[0058] Step S104-4, initializing the weight coefficients of the BP neural network model. To start training the neural network, the weight coefficients W between each layer and the threshold b are set to small non-zero random numbers.

[0059] Step S104-5, forward propagation, passing the input layer by layer to the output layer, calculating the output value of each node. For each node, calculate its input value and output value.

[0060] Step S104-6, select the tanh function as the activation function. Compared with the commonly used sigmoid function, the tanh function can make error training more stable and reliable, can better handle nonlinear relationships and complex time series relationships, can suppress data interference, and can reduce the hardware overhead of data training.

[0061] Step S105, output of predicted values. The output of G2 is used as the moisture value of the material at the outlet of the wire drying machine (the final predicted moisture value of the material at the outlet of the wire drying machine) predicted by the method of the present invention, and the output of G3 is used as the temperature value of the material at the outlet of the wire drying machine (the final predicted temperature of the material at the outlet of the wire drying machine) predicted by the method of the present invention.

[0062] Example 3.

[0063] The present invention provides a device for predicting the temperature and moisture of materials at the outlet of a tofu drying machine, comprising: The variable acquisition module is configured to: acquire the operating variables that are directly related to the prediction result on the mechanism of the tofu drying machine, the disturbance variables that are directly related to the prediction result on the mechanism of the tofu drying machine, and the disturbance variables that are indirectly related to the prediction result on the mechanism of the tofu drying machine; The preliminary prediction module is configured to: input each acquired variable into a trained random forest model based on multi-output regression to obtain a preliminary predicted temperature and a preliminary predicted moisture content of the material at the outlet of the tofu drying machine; The auxiliary data acquisition module is configured to: acquire discrete auxiliary data generated during the drying process of the tofu drying machine; The moisture prediction module is configured to: input the acquired variables, the preliminary predicted moisture and the discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture of the material at the outlet of the tofu drying machine; The temperature prediction module is configured to: input the acquired variables, preliminary predicted temperature and discrete auxiliary data into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

[0064] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0065] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0066] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0068] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting the temperature and moisture content of the material at the outlet of a tofu drying machine, characterized in that: include: Acquire the operating variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, the disturbance variables that have a direct relationship with the predicted result on the mechanism of the tofu drying machine, and the disturbance variables that have an indirect relationship with the predicted result on the mechanism of the tofu drying machine; The obtained variables are input into the trained random forest model based on multi-output regression to obtain the preliminary predicted temperature and preliminary predicted moisture of the material at the outlet of the tofu drying machine; Acquire discrete auxiliary data generated by the drying process of the tofu drying machine; Input the acquired variables, preliminary predicted moisture content and discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture content of the material at the outlet of the tofu drying machine; The obtained variables, preliminary predicted temperature and discrete auxiliary data are input into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

2. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 1, characterized in that: The operating variables that are directly related to the predicted results in the mechanism of the wire drying machine include the actual value of the drum wall temperature, the actual value of the hot air temperature and the actual value of the hot air volume; the disturbance variables that are directly related to the predicted results in the mechanism of the wire drying machine include: the actual value of the incoming material moisture, the actual value of the incoming material temperature, the actual value of the incoming material flow rate, the steam dryness value and the negative pressure value of the dehumidification fan; the disturbance variables that are indirectly related to the predicted results in the mechanism of the wire drying machine include the ambient temperature value, the ambient humidity value and the steam temperature value; the discrete auxiliary data include the production stage, the production operation time schedule and the number of batches produced on the day.

3. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 1, characterized in that: The random forest model based on multi-output regression is trained by the following method: Obtain the following data from the historical data: operating variables that are directly related to the prediction results on the mechanism of the wire drying machine, disturbance variables that are directly related to the prediction results on the mechanism of the wire drying machine, and disturbance variables that are indirectly related to the prediction results on the mechanism of the wire drying machine, and obtain the wire drying machine outlet material temperature and wire drying machine outlet material moisture corresponding to the above variables; Based on the measurement time of the material temperature and moisture content at the outlet of the wire drying machine, the obtained variables are aligned in time series; The Pearson correlation coefficients of the disturbance variables with indirect relationship in the mechanism of the torrefaction machine and the outlet material temperature and moisture content of the torrefaction machine are calculated and predicted respectively. If the two Pearson correlation coefficients of a disturbance variable are both less than 0.3, the disturbance variable is discarded. The random forest model based on multi-output regression was trained using the time-series aligned and screened variables, the outlet material temperature of the tofu drying machine, and the outlet material moisture of the tofu drying machine. The training ended when the degree of fit was greater than 0.

9.

4. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 3, characterized in that: The variables obtained based on the measurement time of the temperature of the material at the outlet of the wire drying machine and the moisture of the material at the outlet of the wire drying machine are aligned in time sequence, including: The measurement time of the outlet material temperature and moisture content of the wire drying machine is taken as the timing reference point, and all variables are moved backward in the timing until they are aligned with the timing reference point.

5. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 4, characterized in that: The measurement time of the temperature of the material at the outlet of the wire drying machine and the moisture of the material at the outlet of the wire drying machine is used as the time series reference point, and all variables are moved backward in the time series until they are aligned with the time series reference point, including: The time from the moisture meter in the drying process to the drying machine inlet is obtained as the first delay time. t 1; The time from the inlet of the wire drying machine to the outlet of the wire drying machine during the wire drying process is obtained as the second delay time t 2; The time from the exit of the silk drying machine to the exit moisture meter during the silk drying process is obtained as the third delay time t 3; Calculating a preliminary time series delay time for each variable to move backward in time series based on the first delay time, the second delay time and the third delay time; Based on the Pearson correlation principle and the preliminary time series delay time, the precise time series alignment delay correction time of each variable moving backward in time series is calculated; The preliminary timing delay time and the precise timing alignment delay correction time corresponding to each variable are added together to obtain the final delay time corresponding to each variable, and each variable is moved backward in timing by the final delay time to complete the timing alignment of each variable.

6. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 5, characterized in that: The calculation of the preliminary time sequence delay time of each variable moving backward in time sequence based on the first delay time, the second delay time and the third delay time is performed by the following method: The initial time delay of the actual value of the cylinder wall temperature, the actual value of the hot air temperature and the actual value of the hot air volume is: t 2 / 2+ t 3. The initial sequence delay time of the actual value of incoming material moisture, actual value of incoming material temperature and actual value of incoming material flow is: t 1+ t 2 +t 3. The initial timing delay time of steam dryness value is t 2 +t 3. The initial timing delay time of the negative pressure value of the dehumidification fan is t 3. The initial timing delay time of the ambient temperature value and the ambient humidity value is t 2 / 2+ t 3. The initial timing delay time of the steam temperature value is t 2+ t 3.

7. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 5, characterized in that: The calculation of the precise time series alignment delay correction time of each variable moving backward in time series based on the Pearson correlation principle and the preliminary time series delay time is performed by the following method: ； in, i N It indicates the precise timing alignment delay correction time of the Nth variable when the sum of the Pearson correlation coefficients between the Nth variable and the two dependent variables is the largest. N indicates the sequence number of the variable, arg max indicates the arg max function, and corr(a,b) indicates the Pearson correlation coefficient between a and b. T max Indicates the maximum change value of the initial timing delay time, y 1 indicates the moisture content of the material at the outlet of the silk drying machine. y 2 indicates the outlet temperature of the fiber drying machine. Indicates moving backward in time The Nth variable after Indicates the initial timing delay time of the Nth variable.

8. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 1, characterized in that: The first BP neural network model is trained by the following method: Obtain the following data from the historical data: operating variables that have a direct relationship with the prediction result on the mechanism of the wire drying machine, disturbance variables that have a direct relationship with the prediction result on the mechanism of the wire drying machine, and disturbance variables that have an indirect relationship with the prediction result on the mechanism of the wire drying machine, moisture content of the wire drying machine outlet material corresponding to the above variables, and discrete auxiliary data corresponding to the above variables; The moisture content of the material at the outlet of the tofu drying machine is used as the output of the first BP neural network model, and other data are used as input. The first BP neural network model is reversely trained, and the tanh function is used as the activation function during the reverse training process.

9. The method for predicting the temperature and moisture content of the outlet material of the tofu drying machine according to claim 1, characterized in that: The second BP neural network model is trained by the following method: Obtain the following data from the historical data: operating variables that have a direct relationship with the predicted result on the mechanism of the torrefaction machine, disturbance variables that have a direct relationship with the predicted result on the mechanism of the torrefaction machine, and disturbance variables that have an indirect relationship with the predicted result on the mechanism of the torrefaction machine, the torrefaction machine outlet material temperature corresponding to the above variables, and discrete auxiliary data corresponding to the above variables; The outlet material temperature of the tofu drying machine is used as the output of the second BP neural network model, and other data are used as input. The second BP neural network model is reversely trained, and the tanh function is used as the activation function during the reverse training process.

10. A device for predicting the temperature and moisture content of materials at the outlet of a tofu drying machine, characterized in that: include: The variable acquisition module is configured to: acquire the operating variables that are directly related to the prediction result on the mechanism of the tofu drying machine, the disturbance variables that are directly related to the prediction result on the mechanism of the tofu drying machine, and the disturbance variables that are indirectly related to the prediction result on the mechanism of the tofu drying machine; The preliminary prediction module is configured to: input each acquired variable into a trained random forest model based on multi-output regression to obtain a preliminary predicted temperature and a preliminary predicted moisture content of the material at the outlet of the tofu drying machine; The auxiliary data acquisition module is configured to: acquire discrete auxiliary data generated during the drying process of the tofu drying machine; The moisture prediction module is configured to: input the acquired variables, the preliminary predicted moisture and the discrete auxiliary data into the trained first BP neural network model to obtain the final predicted moisture of the material at the outlet of the tofu drying machine; The temperature prediction module is configured to: input the acquired variables, preliminary predicted temperature and discrete auxiliary data into the trained second BP neural network model to obtain the final predicted temperature of the outlet material of the tofu drying machine.

Images