CN118098405A - Parameter optimization method for rice paper manufacturing process - Google Patents

Abstract

The invention discloses a parameter optimization method of a rice paper manufacturing process. In order to solve the problem that the prior art can not realize automatic equalization of base material parameters and equipment parameters by too much manpower and then optimize pulp mixing, the method comprises the steps of obtaining basic parameters and equipment parameters temporarily set in the rice paper manufacturing process, presetting ideal quality parameters of paper, relatively providing substitute quality parameters, and establishing a mapping relation of the basic parameters, the equipment parameters and the substitute quality parameters; establishing a first neural network model for training basic parameters and a second neural network for training equipment parameters based on the mapping relation; and adjusting basic parameters and equipment parameters based on the mapping relation and the neural network model until the difference value set balance of the generation quality parameters and the ideal quality parameters is obtained, setting the generation quality parameters at the moment as the optimal quality parameters, outputting the corresponding basic parameters and equipment parameters, and obtaining the optimal parameter set in all parameter constraint conditions to realize slurry distribution optimization.

Description

A parameter optimization method for rice paper making process

Technical Field

The invention relates to the fields of papermaking technology and artificial intelligence, and in particular to a parameter optimization method for a rice paper making process.

Background technique

In the process of preparing Xuan paper or other types of Chinese traditional calligraphy and painting papers, in order to obtain target paper that meets the use standards, historical data of paper production data is usually retrieved based on production experience, and individual equipment adjustments or substrate control are performed to adjust the pulp data. The above method can obtain paper that meets some requirements, but in the process of manually debugging pulp data, too much manpower is consumed, the overall calculation amount of the solution is too large, or it is impossible to comprehensively improve and coordinate the relationship between various coefficients. Therefore, it is urgent to predict the pulping method based on dynamic algorithms to achieve accurate matching of long and short fibers and various types of equipment, and to achieve formula management of different pulps by controlling substrate parameters and machine parameters.

For example, a "method, device and medium for optimizing pressure control of the pulp processing press section" disclosed in Chinese patent literature, with announcement number CN114277602B, synchronously collects paper quality data based on the pressing form and demand; verifies the integrity and synchronization of the paper quality data, and packages and uploads the verified qualified quality data to the database; verifies the correctness and difference of the qualified quality data, and stores the verified qualified quality data in the database; determines the optimal pressing line pressure parameters based on the matching data model, and sends the optimal pressing line pressure parameters to the host computer; the host computer sends the adjusted optimal pressing line pressure parameters to the DCS controller, and the DCS controller dynamically adjusts the pressing section action. Its solution adapts the material selection and pressing form respectively, and obtains the optimal pressing line pressure and other parameters after the system continuously adjusts the parameters, thereby obtaining qualified paper. However, it can only obtain qualified paper and related parameters, and cannot obtain the optimal paper parameters and corresponding basic parameters through continuous calculation. In addition, its text does not contain specific methods and processes for calculation and control, and the relevant description is relatively general.

Summary of the invention

The present invention mainly solves the problem that the prior art is too dependent on manpower and cannot achieve the optimal slurry matching after automatically balancing the substrate parameters and the equipment parameters; a parameter optimization method for the rice paper production process is provided, basic parameters and equipment parameters are set for the substrate and the equipment, a mapping relationship between the two and the target parameters of the rice paper is established, and the basic parameters and the equipment parameters are trained separately through a dual neural network in conjunction with the relevant constraints established by the target parameters until the optimal parameter group within the constraints is obtained, thereby achieving the optimization of slurry matching.

The above technical problems of the present invention are mainly solved by the following technical solutions:

The present invention includes: obtaining the basic parameters and equipment parameters temporarily set in the process of making rice paper, presetting the ideal quality parameters of the paper and relatively proposing the generation quality parameters, establishing the mapping relationship between the basic parameters, the equipment parameters and the generation quality parameters; establishing a first neural network model for training the basic parameters and a second neural network for training the equipment parameters based on the mapping relationship; adjusting the basic parameters and the equipment parameters based on the mapping relationship and the neural network model until the difference group balance between the generation quality parameters and the ideal quality parameters is obtained, setting the generation quality parameters at this time as the optimal quality parameters, and outputting the corresponding basic parameters and equipment parameters. The basic parameters mentioned are the relevant parameters that the substrate may affect the pulping, the equipment parameters are the adjustable parameters that the equipment may affect the pulping, the generation quality parameters are the pulping related parameters that can be achieved in reality, and the ideal quality parameters are the target or preset pulping related parameters. In addition, by establishing a neural network to first train the basic parameters and the equipment parameters separately, their effective ranges can be determined, and by establishing a mapping relationship and continuously adjusting the training parameters, the generation quality parameters that meet the requirements can be obtained. When the contemporary quality parameters are infinitely close to the ideal quality parameters, the required optimal quality parameters are obtained, and the optimal pulping that meets the actual modulation is achieved.

Preferably, the basic parameters include but are not limited to: average length of substrate fiber l, substrate hardness h and/or substrate chromaticity coefficient μ _c , and other basic parameter influencing coefficients k _j ; the equipment parameters include but are not limited to: blade speed v, press force f, beating time T _d and/or pressing time _Ty , and also include beating times Countd, pressing times County and other equipment parameter influencing coefficients k _s . Blade speed and average length of substrate fiber can affect the efficiency of substrate cutting by the equipment, press force and substrate hardness can affect the pressing efficiency of the equipment, and the two together with parameters such as beating time, pressing time, beating times and pressing times can affect the density of the final slurry; and between the pressing and beating processes, related reagents are usually added, and it is temporarily assumed that the number of reagent additions is the same as the number of pressing times, and the reagents and the chromaticity of the substrate jointly affect the chromaticity of the final slurry.

Preferably, the generation quality parameters are further set to be expressed as [ρ ₁ ,ρ ₂ ,μ _X ], wherein ρ ₁ and ρ ₂ represent the first density coefficient and the second density coefficient of the rice paper after beating and pressing, respectively, and μ _X represents the chromaticity of the rice paper at the current density coefficient; the mapping relationship between the basic parameters, the equipment parameters and the generation quality parameters is expressed as: Among them, f(l,v) corresponds to the influence of blade rotation speed and average length of substrate fiber on substrate cutting efficiency, and f(h,f) corresponds to the influence of press force and substrate hardness on substrate pressing efficiency. The material beating time _Td and/or pressing time _Ty , also include the number of beating times Countd and the number of pressing times County as restriction conditions to limit the value of the above functions. /> The above mapping relationship is established based on the basic data, equipment data and parameter types of the quality parameters, which can comprehensively consider the influence of the base material and equipment on the relevant parameters of the slurry, so as to facilitate the subsequent acquisition of the optimal slurry parameters.

Preferably, in the process of establishing the first neural network, the mean of historical equipment parameters is extracted as the equipment parameters for training, and the ideal quality parameters are introduced into [ρ ₁ ,ρ ₂ ,μ _X ], and then the mapping relationship is imported for the parameter values selected within the preset interval of the basic parameters, and the range constraints of the basic parameters are obtained after training with the PID controller: l∈[l _b ,l _t ], h∈[h _b ,h _t ], μ _c ∈[μ _cb ,l _ct ], wherein the parameters with subscripts b and t represent the bottom constraint and top constraint of the corresponding parameters respectively. The basic parameters are trained with fixed and reasonable preset equipment parameters under the first neural network, and balanced training is performed on the three related parameters of the basic parameters under the conditions of the specified generation quality parameters and equipment parameters, so that the reasonable intervals corresponding to each parameter can be obtained, thereby ensuring the pertinence and accuracy in the selection of the substrate.

As a preferred method, in the process of establishing the second neural network, the mean of historical basic parameters is extracted as the basic parameters for training, and the ideal quality parameters are introduced into [ρ ₁ ,ρ ₂ ,μ _X ], and then the parameter values are selected within the preset interval of the device parameters to import the mapping relationship, and the range constraints of the device parameters are obtained after training: v∈[v _b ,v _t ], f∈[f _b ,f _t ], T _d ∈[T _db ,T _dt ], _Ty ∈[T _yb ,T _yt ]; Countd∈[Countd _b ,Countd _t ], County∈[County _b ,County _f ]; the parameters with subscripts b and t represent the bottom constraint and top constraint of the corresponding parameters respectively. Under the condition of preset fixed and reasonable basic parameters, the ideal quality parameters are substituted into the substitute quality parameters, and the six related parameters of the device parameters are trained to obtain their respective range constraint intervals, which can reduce the constraints on the device and debug the device in a targeted manner.

As a preference, an intermediate function is further obtained according to the neural network training result: Substituting the intermediate function into the mapping relationship, we get the quality parameter expression:/> Obtain the range constraint of the generation quality parameter. Express the intermediate function with the relevant basic parameters and device parameters to obtain a specific relationship formula. Express the relationship formula as the intermediate function and bring it into the mapping relationship to obtain the generation quality parameter expression represented by the basic parameters and device parameters, thereby obtaining the range constraint of the generation quality parameter based on the basic parameters and device parameters containing the range constraint. The above expression can improve the accuracy of the subsequent generation quality parameters.

Preferably, an ideal quality parameter is further set and expressed as [ρ ₁₀ ,ρ ₂₀ ,μ _X0 ], wherein ρ ₁₀ and ρ ₂₀ represent the first density coefficient and the second density coefficient of the rice paper under the ideal state, respectively, and μ _X0 is the chromaticity coefficient of the rice paper under the ideal state; according to the range constraint of the generation quality parameter, the estimated difference range between the ideal quality parameter and the generation quality parameter is set as: |ρ ₁₀ -ρ ₁ |∈[ρ _ξb ,ρ _ξt ], |ρ ₂₀ -ρ ₂ |∈[ρ _λb ,ρ _λt ], |μ _X0 -μ _X |∈[μ _cb ,l _ct ]. An array expression form of the ideal quality parameter is added, and the difference between the ideal quality parameter and the generation quality parameter is preset to calculate the gap between the generation quality parameter and the ideal quality parameter. If the difference is within the corresponding difference range, it is considered that the current generation quality parameter is in the candidate range of the optimal quality parameter, which can effectively narrow the data range and improve the judgment range of the optimal quality parameter.

As a preferred method, the basic parameters and equipment parameters within the constraint range are further continuously trained according to the estimated difference range between the ideal quality parameters and the generation quality parameters to obtain several groups of generation quality parameters that meet the range; let η be the preset control value coefficient that narrows the gap between ρ ₁ and ρ ₂ ; select The quality parameter with the smallest difference from the ideal quality parameter is taken as the optimal quality parameter, and the optimal quality parameter and the corresponding basic parameters and equipment parameters are output. /> The establishment of inequalities can meet the requirements of individual pulp coefficients, so that the basic parameters and equipment parameters are in a state of parameter balance, reducing parameter fluctuations. In addition, the alternative quality parameters that are infinitely close to the ideal parameters are selected within the candidate range of the optimal quality parameters to achieve the optimal pulping under actual conditions, and the feasibility of some pulping formulas can be verified.

The beneficial effects of the present invention are:

1. A parameter optimization method for a rice paper production process of the present invention constructs a process quality monitoring model for special paper for calligraphy and painting based on a neural network, and trains the neural network using environmental parameters and observation data of paper quality, thereby establishing a mapping relationship between production process parameters and special paper quality parameters, and obtaining an accurate expression of pulping parameters, thereby achieving the effect of obtaining idealized and optimized pulping parameters through a large range of training parameters;

2. A parameter optimization method for a rice paper making process of the present invention proposes an easily implementable suboptimal control strategy, namely, threshold control; by performing range constraints on pre-trained production process parameters, threshold control is performed on the slurry mixing parameters under the mapping relationship; in addition, ideal slurry mixing parameters are preset, and difference threshold control is performed between them and the actually available slurry mixing parameters, thereby obtaining a variety of optimized parameters that meet the requirements, and further obtaining the optimal slurry mixing parameters under the specified threshold.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a parameter optimization method for a rice paper production process of the present invention;

FIG. 2 is a flow chart of a parameter optimization method for a rice paper production process of the present invention.

Detailed ways

The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

Example:

A parameter optimization method for a rice paper production process of the present embodiment, as shown in FIG1 and FIG2, includes: setting basic parameters as parameters related to the substrate that may affect the slurry preparation, equipment parameters as adjustable parameters of the equipment that may affect the slurry preparation, substitute quality parameters as parameters related to the slurry preparation that can be achieved in reality, and ideal quality parameters as target or preset parameters related to the slurry preparation. By establishing a neural network to first train the basic parameters and equipment parameters separately, their effective ranges can be determined, and by establishing a mapping relationship and continuously adjusting the training parameters, substitute quality parameters that meet the requirements can be obtained. When the contemporary quality parameters are infinitely close to the ideal quality parameters, the required optimal quality parameters are obtained, and the optimal slurry preparation that meets the actual modulation is achieved.

Among them, the basic parameters and equipment parameters are uniformly set as production process parameters. More adjustable and variable production process parameters can be added in the mapping relationship to make the obtained slurry preparation parameters more accurate. In addition, other influencing factors can also be added, such as environmental factors, dust, temperature, humidity, drying conditions, drying conditions and other parameters to enrich the types of parameters, so as to obtain more comprehensive training and obtain the optimal slurry preparation parameters.

The specific steps include: S1. Obtaining the basic parameters and equipment parameters temporarily set in the process of making rice paper, presetting the ideal quality parameters of the paper and relatively proposing alternative quality parameters, and establishing a mapping relationship between the basic parameters, equipment parameters and alternative quality parameters.

The set basic parameters related to the substrate include but are not limited to: average fiber length l of the substrate, hardness h of the substrate and/or chromaticity coefficient μ _c of the substrate, and also include other basic parameter influencing coefficients k _j ; the set equipment parameters include but are not limited to: blade speed v, press force f, beating time T _d and/or pressing time _Ty , and also include beating times Countd, pressing times County and other equipment parameter influencing coefficients k _s .

The blade speed and the length of the substrate fiber can affect the efficiency of the equipment in cutting the substrate, the press force and the substrate hardness can affect the equipment's pressing efficiency of the substrate, and the two together with parameters such as the beating time, pressing time, the number of beatings and the number of pressings can affect the density of the final slurry. Relevant reagents are usually added between the pressing and beating processes. It is temporarily assumed that the number of times the reagents are added is the same as the number of pressings, and the reagents and the color of the substrate together affect the color of the final slurry.

In addition, other basic parameter influence coefficients _kj and other equipment parameter influence coefficients _ks are used as other production process parameters. Other parameters that have no direct effect or little effect on the slurry mixing parameters can be added here, such as related parameters that have indirect effects when there are slight differences between material categories. In addition, parameters related to the type of additives, the amount of additives, or the order of additives can be added to the equipment parameters to more accurately regulate the slurry mixing parameters.

The generation quality parameters are set to be expressed as [ρ ₁ ,ρ ₂ ,μ _X ], where ρ ₁ and ρ ₂ represent the first density coefficient and the second density coefficient of the rice paper after beating and pressing, respectively, and μ _X represents the chromaticity of the rice paper at the current density coefficient; the mapping relationship between the basic parameters, equipment parameters and generation quality parameters is expressed as

Among them, f(l,v) corresponds to the influence of blade rotation speed and average length of substrate fiber on substrate cutting efficiency, and f(h,f) corresponds to the influence of press force and substrate hardness on substrate pressing efficiency. The material beating time _Td and/or pressing time _Ty also include the number of beating times Countd and the number of pressing times County as restriction conditions to limit the value of the above function. The above mapping relationship is established based on the basic data, equipment data and parameter types of the quality parameters, which can comprehensively consider the influence of the base material and equipment on the relevant parameters of the slurry, so as to facilitate the subsequent acquisition of the optimal slurry parameters.

In view of the difficulties such as the many and complex influencing factors in the slurry mixing process and the difficulty in establishing an accurate mathematical model, a dynamic prediction control algorithm is introduced for control. In combination with the above mapping relationship, the development and feasibility verification of different slurry mixing formulas can be realized by dynamically changing parameters. In addition, by combining the uniform transfer technology and the random dynamic programming method, the optimal feedback modulation strategy for allocation is given.

S2. Establishing a first neural network model for training basic parameters and a second neural network model for training device parameters based on the mapping relationship.

During the establishment of the first neural network, the mean value of historical equipment parameters is extracted as the equipment parameters during training, and the ideal quality parameters are substituted into [ρ ₁ ,ρ ₂ ,μ _X ], and then the mapping relationship is imported for the parameter values selected within the preset interval of the basic parameters. After training with the PID controller, the range constraints of the basic parameters are obtained: l∈[l _b ,l _t ], h∈[h _b ,h _t ], μ _c ∈[μ _cb ,l _ct ], where the parameters with subscripts b and t represent the bottom constraint and top constraint of the corresponding parameters, respectively.

In the first neural network, the basic parameters are trained with fixed and reasonable preset equipment parameters. Under the conditions of specified generation quality parameters and equipment parameters, balanced training is carried out on the three related parameters of the basic parameters to obtain the reasonable range corresponding to each parameter, thereby ensuring the pertinence and accuracy in substrate selection.

During the establishment of the second neural network, the mean of historical basic parameters is extracted as the basic parameters for training, and the ideal quality parameters are substituted into [ρ ₁ ,ρ ₂ ,μ _X ], and then the mapping relationship is imported for the parameter values selected within the preset range of the equipment parameters. After training, the range constraints of the equipment parameters are obtained: v∈[v _b ,v _t ], f∈[f _b ,f _t ], T _d ∈[T _db ,T _dt ], _Ty ∈[T _yb ,T _yt ]; Countd∈[Countd _b ,Countd _t ], County∈[County _b ,County _t ]; the parameters with subscripts b and t represent the bottom constraints and top constraints of the corresponding parameters respectively.

Under the preset fixed and reasonable basic parameter conditions, the ideal quality parameters are substituted into the substitute quality parameters, and the six related parameters of the equipment parameters are trained to obtain their respective range constraint intervals, which can reduce the constraints on the equipment and debug the equipment in a targeted manner.

By cooperating with the development of special equipment for precise slurry mixing of complex slurries and additives, based on the key process state parameters such as flow rate and density given by sensor components such as flow meters and online densitometers, the flow of each batching pipeline is adjusted by controlling actuators such as reduction motors and diaphragm pumps to achieve precise matching of long and short fibers and various additives. By changing the parameter structure or parameter data of the equipment, effective threshold control of equipment parameters can be achieved.

Get the intermediate function based on the neural network training results Substituting the intermediate function into the mapping relationship, we get the quality parameter expression:/> Obtain the range constraints of the quality parameters.

The intermediate function is expressed by related basic parameters and device parameters to obtain a specific relationship formula. The relationship formula is expressed as an intermediate function and brought into the mapping relationship to obtain the generation quality parameter expression represented by the basic parameters and device parameters, thereby obtaining the range constraint of the generation quality parameter according to the basic parameters and device parameters containing the range constraint. The above expression can improve the accuracy of the subsequent generation quality parameters.

S3. Adjust the basic parameters and equipment parameters based on the mapping relationship and the neural network model until the difference group balance between the substitute quality parameters and the ideal quality parameters is obtained. Set the substitute quality parameters at this time as the optimal quality parameters, and output the corresponding basic parameters and equipment parameters.

The ideal quality parameters are set and expressed as [ρ ₁₀ ,ρ ₂₀ ,μ _X0 ], where ρ ₁₀ and ρ ₂₀ represent the first density coefficient and the second density coefficient of Xuan paper under the ideal state, respectively, and μ _X0 is the chromaticity coefficient of Xuan paper under the ideal state; according to the range constraint of the generation quality parameter, the estimated difference range between the ideal quality parameter and the generation quality parameter is set as: |ρ ₁₀ -ρ ₁ |∈[ρ _ξb ,ρ _ξt ], |ρ ₂₀ -ρ ₂ |∈[ρ _λb ,ρ _λt ], |μ _X0 -μ _X |∈[μ _cb ,l _ct ].

The preset estimated difference range is convenient for continuous feedback adjustment of basic parameters and equipment parameters when the range is not reached; after reaching the threshold range, a feedback signal is formed, indicating that the current generation quality parameters are qualified, so as to carry out parameter adjustment in the next stage; until the generation quality parameters do not change after several adjustments, it means that the generation quality parameters at this time are infinitely close to the ideal quality parameters, so as to obtain the candidate range of the optimal quality parameters, and the basic parameters and equipment parameters corresponding to the optimal quality parameters.

An array expression of the ideal quality parameter is added, and the difference between the ideal quality parameter and the generation quality parameter is preset to calculate the difference between the generation quality parameter and the ideal quality parameter. If the difference is within the corresponding difference range, it is considered that the current generation quality parameter is in the candidate range of the optimal quality parameter, which can effectively narrow the data range and improve the judgment range of the optimal quality parameter.

According to the estimated difference range between the ideal quality parameter and the generation quality parameter, the basic parameters and equipment parameters within the constraint range are continuously trained to obtain several groups of generation quality parameters that meet the range; let η be the preset control coefficient that narrows the gap between ρ ₁ and ρ ₂ ; select The substitute quality parameter with the smallest difference from the ideal quality parameter is taken as the optimal quality parameter, and the optimal quality parameter and the corresponding basic parameters and equipment parameters are output.

The establishment of inequalities can meet the requirements of individual pulp coefficients, so that the basic parameters and equipment parameters are in a state of parameter balance, reducing parameter fluctuations. In addition, the alternative quality parameters that are infinitely close to the ideal parameters are selected within the candidate range of the optimal quality parameters to achieve the optimal pulping under actual conditions, and the feasibility of some pulping formulas can be verified.

In summary, the method in this embodiment constructs a quality monitoring model for the process flow of special paper for calligraphy and painting based on a neural network, and uses environmental parameters and observation data of paper quality to train the neural network, thereby establishing a mapping relationship between production process parameters and special paper quality parameters, and obtaining an accurate expression of the pulping parameters, thereby achieving the effect of obtaining idealized and optimized pulping parameters by training parameters over a large range. In addition, by subjecting the pre-trained production process parameters to range constraints, the pulping parameters under the mapping relationship are threshold-controlled; in addition, the ideal pulping parameters are preset, and the difference thresholds are controlled between them and the actually obtainable pulping parameters, thereby obtaining a variety of optimized parameters that meet the requirements, and further obtaining the optimal pulping parameters under the specified thresholds.

It should be understood that the embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the appended claims of the application.

Claims (8)

1. The parameter optimization method of the rice paper manufacturing process is characterized by comprising the following steps of:

Acquiring temporary basic parameters and equipment parameters in the rice paper manufacturing process, presetting ideal quality parameters of paper, relatively providing substitute quality parameters, and establishing a mapping relation of the basic parameters, the equipment parameters and the substitute quality parameters;

establishing a first neural network model for training basic parameters and a second neural network for training equipment parameters based on the mapping relation;

And adjusting basic parameters and equipment parameters based on the mapping relation and the neural network model until the difference value group balance between the generation quality parameters and the ideal quality parameters is obtained, setting the generation quality parameters at the moment as the optimal quality parameters, and outputting the corresponding basic parameters and equipment parameters.

2. The method for optimizing parameters of a rice paper making process according to claim 1, wherein the basic parameters include, but are not limited to: the average length of the substrate fiber l, the substrate hardness h and/or the substrate chromaticity coefficient mu _c, and also comprises other basic parameter influence coefficients k _j; the device parameters include, but are not limited to: knife She Zhuaisu v, press force f, length of time of beating T _d and/or length of time of pressing T _y, and also include beating time Countd, pressing time count and other equipment parameter influence coefficient k _s.

3. The parameter optimization method of a rice paper making process according to claim 2, further comprising: setting a generation quality parameter as [ rho ₁,ρ₂,μ_X ], wherein rho ₁ and rho ₂ respectively represent a first density coefficient and a second density coefficient of the rice paper after being subjected to material beating and pressing, and mu _X represents chromaticity of the rice paper under the current density coefficient; the mapping relation of the basic parameters, the equipment parameters and the generation quality parameters is expressed as

4. A method for optimizing parameters of a rice paper making process according to claim 2 or 3, further comprising: in the first neural network building process, extracting a historical equipment parameter mean value as an equipment parameter during training, taking an ideal quality parameter into [ rho ₁,ρ₂,μ_X ], then importing a mapping relation to a selected parameter value in a preset interval of a basic parameter, and obtaining a range constraint of the basic parameter after training by using a PID controller: l e [ l _b,l_t],h∈[h_b,h_t],μ_c∈[μ_cb,l_ct ], wherein the parameters of the subscripts comprising b and t are the bottom constraint and the top constraint representing the corresponding parameters, respectively.

5. The method for optimizing parameters of a rice paper making process according to claim 4, further comprising: in the second neural network building process, the historical basic parameter mean value is extracted to serve as a basic parameter in training, ideal quality parameters are brought into [ rho ₁,ρ₂,μ_X ], then parameter values selected in a preset interval of equipment parameters are imported into a mapping relation, and parameters of which the range constraint ：v∈[v_b,v_t],f∈[f_b,f_t],T_d∈[T_db,T_dt],T_y∈[T_yb,T_yt];Countd∈[Countd_b,Countd_t],County∈[County_b,County_t]; subscripts of the acquired equipment parameters comprise b and t are respectively representing bottom constraint and top constraint of corresponding parameters after training.

6. The parameter optimization method of a rice paper making process according to claim 2 or 5, further comprising: obtaining an intermediate function according to the training result of the neural networkAnd carrying out the intermediate function into the mapping relation to obtain a generation quality parameter expression: /(I)And obtaining the range constraint of the substitution quality parameter.

7. The method for optimizing parameters of a rice paper making process according to claim 6, further comprising: setting ideal quality parameters and expressing as [ rho ₁₀,ρ₂₀,μ_X0 ], wherein rho ₁₀ and rho ₂₀ respectively represent a first density coefficient and a second density coefficient of the rice paper in an ideal state, and mu _X0 is a chromaticity coefficient of the rice paper in the ideal state; setting the estimated difference range of the ideal quality parameter and the generation quality parameter as follows according to the generation quality parameter range constraint ：|ρ₁₀-ρ₁|∈[ρ_ξb,ρ_ξt],|ρ₂₀-ρ₂|∈[ρ_λb,ρ_λt],|μ_X0-μ_X|∈[μ_cb,l_ct].

8. The method for optimizing parameters of a rice paper making process according to claim 7, further comprising: continuously training basic parameters and equipment parameters in a constraint range according to the estimated difference value range of the ideal quality parameters and the generation quality parameters to obtain a plurality of groups of generation quality parameters conforming to the range; let eta be the preset control value coefficient for narrowing the gap between rho ₁ and rho ₂; selectingAnd the generation quality parameter with the smallest difference value with the ideal quality parameter is used as the optimal quality parameter, and the optimal quality parameter, the corresponding basic parameter and the corresponding equipment parameter are output.