CN111476428A - Big data analysis-based brewing process optimization method - Google Patents

Big data analysis-based brewing process optimization method Download PDF

Info

Publication number
CN111476428A
CN111476428A CN202010301141.2A CN202010301141A CN111476428A CN 111476428 A CN111476428 A CN 111476428A CN 202010301141 A CN202010301141 A CN 202010301141A CN 111476428 A CN111476428 A CN 111476428A
Authority
CN
China
Prior art keywords
model
link
brewing process
output
grain
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010301141.2A
Other languages
Chinese (zh)
Other versions
CN111476428B (en
Inventor
林锋
刘淼
张宿义
张睿
宋建勋
张程
李德林
王海
蔡小波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Luzhou Laojiao Co Ltd
Luzhou Laojiao Brewing Co Ltd
Original Assignee
Luzhou Laojiao Co Ltd
Luzhou Laojiao Brewing Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Luzhou Laojiao Co Ltd, Luzhou Laojiao Brewing Co Ltd filed Critical Luzhou Laojiao Co Ltd
Priority to CN202010301141.2A priority Critical patent/CN111476428B/en
Publication of CN111476428A publication Critical patent/CN111476428A/en
Application granted granted Critical
Publication of CN111476428B publication Critical patent/CN111476428B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Development Economics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Operations Research (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Game Theory and Decision Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Educational Administration (AREA)
  • Primary Health Care (AREA)
  • Manufacturing & Machinery (AREA)
  • Apparatus Associated With Microorganisms And Enzymes (AREA)

Abstract

The invention relates to the technical field of wine brewing, aims to solve the problem of poor optimization effect of the existing wine brewing process optimization method, and provides a wine brewing process optimization method based on big data analysis.

Description

Big data analysis-based brewing process optimization method
Technical Field
The invention relates to the technical field of wine brewing, in particular to a wine brewing process optimization method.
Background
The solid brewing process of the strong aromatic Chinese spirits mainly comprises a saccharification link, a vinasse preparation link, a fermentation link and a distillation link, and the main process comprises the following steps: pretreatment (such as warming saccharification and the like) of grains (such as sorghum); mixing the saccharified grain with Daqu, bran coat and distilled fermented grains in a certain proportion, and adding into a fermentation tank; fermenting for 30-45 days, taking out alcohol-containing fermented grains, mixing with bran shell, and loading into wine retort; introducing steam into a bottom pot below the wine retort, wherein the steam penetrates through the fermented grains from bottom to top to extract alcohol and fragrant substances in the fermented grains, and according to the characteristic that the quality of the wine varies along with the distillation time, a worker segments distilled wine by a method of watching flowers and picking the wine and stores the segmented distilled wine into raw wine storage tanks (generally divided into head wine, second-stage wine, third-stage wine and tail wine) of different grades; spreading and cooling the distilled fermented grains, and then using the fermented grains for the next round of fermentation (generally discarding after 6-7 cycles); the head wine and the tail wine have poor quality and can be returned to the bottom pot for reuse; the second section and the third section are raw wine obtained by production; the raw wine leaves the factory after cellaring, wine blending and packaging.
The production of the white spirit is an intermittent production process which takes biological fermentation as a core, has many interference factors influencing the quality of the white spirit and has a long period. On the basis of mechanization of the existing production process, an automatic control system is added to control the production process in an optimal operation state (optimal control) so as to achieve the expected aims (optimal targets) of improving the utilization rate of grains and the quality of white spirit, saving energy and consumption, reducing production cost and the like, and the method is the development requirement of numerous white spirit factories.
At present, a research method for optimizing the white wine production process mainly aims at a certain production link (such as a fermentation link) and obtains optimal material proportion, operation conditions (such as fermentation temperature and fermentation time) and other optimization results through mechanism modeling or numerical simulation or simulation test and analysis under laboratory conditions. For example, Liu Deng of Jiangnan university, aiming at the fermentation process of Agrobacterium ATCC31749, proposes a fermentation kinetic model with dissolved oxygen concentration as a key control variable, and solves the model by a bird swarm algorithm so as to maximize the concentration of the fermentation product (CN 110334373A). The guanya of China university of Petroleum proposes a phellinus igniarius fermentation experiment data optimization and experimental product yield prediction method (CN109859804A) based on a neural network. The research is a local optimization method, local optimization can only optimize a local target, and usually cannot reach the optimal state of the whole process production flow, because the local optimization of a single link is not equal to global optimization; the global optimization effect requires that all links are operated under the optimization operation state at the same time, and the optimization degrees of all links need to be considered mutually.
In the doctor's paper of jialin of Shanghai university of transportation, the modeling and optimization of the Pichia pastoris are researched, and a dynamics model of the Pichia pastoris is established on the basis of the hypothesis that the average cell volume does not change in the fermentation process and the like by analyzing the metabolic pathway of the Pichia pastoris matrix. The model is combined with a bioreactor model, the relationship between a manipulated variable (feeding rate) and a state variable (cell concentration, substrate concentration, protein concentration and fermentation liquor volume) is obtained, and the optimization of a glycerol feeding strategy is obtained to improve the optimization aim of heterologous protein yield (Jialin. Pichia pastoris fermentation process modeling and optimization [ D ] Shanghai transportation university, 2007). The method for solving the optimized data by using the mechanism model and the empirical model is carried out under certain assumed conditions, the actual solid-state brewing process of the white spirit is long in existing time, complex in object characteristics, multiple in interference factors and easy to change, the production link characteristics described by the mechanism model and the empirical model are deviated from the actual characteristics, and the obtained optimized result is deviated from the actual optimal operation state in the actual production process.
Yangbai and the like research the optimal operating conditions of yellow water biological fermentation under laboratory conditions to obtain the conclusion that the fermentation time is 8 days and the fermentation temperature is 35 ℃ (Yangbai, Zhang dori and the like, research of yellow water biological fermentation technology, brewing technology, 2018No. 1). The static optimization means that optimized operation parameters are obtained under certain experimental conditions and then used for production, which is equivalent to artificial expansion of specific experimental conditions (grain and working environment); the optimization state in the actual production is related to various factors such as the types and the quality of raw materials (grains, yeast, and the like), the environmental temperature, the running state of production equipment, the fermentation environment and the like, and the change of the factors can cause the actual optimization operation point to deviate; the static optimization method does not consider the above factors and cannot realize the optimization operation in the production process.
In conclusion, the existing research method for brewing process optimization has the characteristics of local objects, staticizing modes, dependence on mechanism models and the like, and has the problem of poor optimization effect.
Disclosure of Invention
The invention aims to solve the problem of poor optimization effect of the existing brewing process optimization method, and provides a brewing process optimization method based on big data analysis.
The technical scheme adopted by the invention for solving the technical problems is as follows: the brewing process optimization method based on big data analysis comprises the following steps:
step 1, obtaining historical parameter information of a wine brewing process, wherein the wine brewing process at least comprises a saccharification link, a vinasse preparation link, a fermentation link and a distillation link;
step 2, respectively establishing an RBF model for a saccharification link, an ANFIS model for a vinasse preparation link, an L S-SVM model for a fermentation link and a BP neural network model for a distillation link according to the historical parameter information;
and 3, acquiring real-time parameter information of the brewing process, calculating parameter optimization results of all links of the brewing process based on all models according to the real-time parameter information, and performing variable control on corresponding links of the brewing process according to the parameter optimization results.
Further, to implement the establishment of the ANFIS model, the historical parameter information includes: the method comprises the following steps of establishing an ANFIS model for a grain blending link, wherein the ANFIS model comprises the following steps of grain saccharification picture information, mother grain water content, grain temperature, mother grain starch content and grain blending proportion:
acquiring gray level co-occurrence matrix contrast characteristics and HSV information in the grain saccharified picture information;
and taking the gray level co-occurrence matrix contrast characteristic, HSV information, the water content of the mother grains, the grain temperature and the starch content of the mother grains as the input of an ANFIS model, and taking the grain preparation proportion as the output of the ANFIS model for training to obtain the ANFIS model for the grain preparation link.
Further, in order to improve the optimization effect of the lees matching link, the ANFIS model is divided into five layers, including: a fuzzy layer, a rule reasoning layer, a normalization layer, a de-fuzzy layer and an output layer, wherein a membership function u is customized for each fuzzy subsetA(x，θ1) Wherein, theta1＝(a1i，a2i，b1i，b2i，c1i，c2i) When the membership function is a bell-shaped function, the output formula of the fuzzy layer is as follows:
Figure BDA0002454034530000031
wherein i is 1, 2;
each node of the rule inference layer corresponds to a fuzzy rule and is used for calculating the excitation strength of each fuzzy rule, and the calculation formula is as follows:
Figure BDA0002454034530000032
wherein i is 1, 2;
the normalization layer is used for normalizing the excitation intensity, calculating to obtain a weighting coefficient output by each fuzzy rule, and calculating the formula as follows:
Figure BDA0002454034530000033
the de-blurring layer is used for blurring the image according to a function f (x, R, theta) in a blurring rule2) Converting the fuzzy output to a precise output, wherein θ2(p, q, r) are the parameters of the back piece, and the conversion formula is as follows:
Figure BDA0002454034530000034
the output layer is used for summing output results of the de-fuzzy layer, and the summation formula is as follows:
Figure BDA0002454034530000035
further, the learning algorithm of the ANFIS model is a BP and L SM hybrid algorithm, and the learning steps are as follows:
A. fixing the front piece coefficient to propagate forward to obtain an output for adjusting the back piece parameter, wherein the output is a linear combination of the back piece parameter, and an equation is as follows:
Figure BDA0002454034530000036
the equation y is a × θ1The least squares solution of (c) is:
θ1＝A+y+(I-A+A)z；
in the formula, A+Moore-Penrose generalized inverse of A, z being any real number, whichThe minimum two-times solution calculation formula is as follows:
x＝A+b；
B. and adjusting the parameters of the back piece according to the output, calculating errors according to the output after the parameters of the back piece are adjusted, reversely propagating the errors to the fuzzy layer, and updating the parameters of the front piece by using a gradient descent method.
Further, to implement the establishment of the RBF model, the historical parameter information includes: the method comprises the following steps of saccharifying time, water adding temperature in a saccharifying link, grain preparation proportion, mother grain starch content and pit entering grain starch content, wherein the method for establishing the RBF model for the saccharifying link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link, the ratio of the prepared grains and the starch content of the mother grains as the input of an RBF model, and taking the starch content of the fermented grains entering the cellar as the output of the RBF model for training to obtain the RBF model for the saccharification link.
Further, in order to improve the optimization effect of the saccharification link, the method further comprises the following steps:
and calculating the model deviation of the PBF model based on historical parameter information, establishing a correction model according to the model deviation, and correcting the RBF model according to the correction model.
Further, in order to realize the establishment of L S-SVM model, the historical parameter information comprises saccharification time, water adding temperature in the saccharification link, fermentation time, cellaring alcohol concentration, butyl butyrate concentration and ethanol concentration, and the method for establishing the L S-SVM model for the fermentation link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link and the fermentation time as the input of an L S-SVM model, and taking the cellaring alcohol concentration, the butyl butyrate concentration and the ethanol concentration as the output of a L S-SVM model for training to obtain a L S-SVM model for the fermentation link.
Further, optimizing the L S-SVM model by a genetic algorithm to obtain the saccharifying time, the water adding temperature in the saccharifying link and the fermenting time corresponding to the maximum alcohol concentration, butyl butyrate concentration and ethanol concentration in the cellar, and performing variable control on the fermenting link by taking the saccharifying time, the water adding temperature in the saccharifying link and the fermenting time as parameter optimization results.
Further, to implement the establishment of the BP neural network model, the historical parameter information includes: the method for establishing the BP neural network model for the distillation link comprises the following steps of:
and taking the steam flow and the distillation time as the input of a BP neural network model, and taking the two-stage liquor yield and the three-stage liquor yield as the output of the BP neural network model for training to obtain the BP neural network model for the distillation link.
Further, in order to optimize parameters of the distillation link, the method further comprises the following steps: and optimizing the BP neural network model by a particle swarm algorithm to obtain the steam flow and the distillation time corresponding to the maximum two-stage liquor yield and the maximum three-stage liquor yield, and performing variable control on the distillation link by taking the steam flow and the distillation time as parameter optimization results.
The invention has the beneficial effects that: the wine brewing process optimization method based on big data analysis can enable the whole production process to run in an optimized state through the whole process optimization method of the white wine brewing process, improve the utilization rate of raw grains and the proportion of high-quality wine, can identify characteristic drift and various random interference factors of each main link in the white wine production process through a method of combining a mechanism model of the main production link and a dynamic numerical model based on big data, can obtain the actual dynamic characteristics of each main joint in the white wine production process, and improves the parameter optimization effect of the wine brewing process.
Drawings
FIG. 1 is a schematic flow chart of a brewing process optimization method based on big data analysis according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an optimization process of the brewing process optimization method based on big data analysis according to the embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The wine brewing process optimization method based on big data analysis comprises the following steps of 1, obtaining historical parameter information of a wine brewing process, wherein the wine brewing process at least comprises a saccharification link, a vinasse preparation link, a fermentation link and a distillation link, 2, respectively establishing an RBF model for the saccharification link, an ANFIS model for the vinasse preparation link, an L S-SVM model for the fermentation link and a BP neural network model for the distillation link according to the historical parameter information, 3, obtaining real-time parameter information of the wine brewing process, calculating parameter optimization results of all links of the wine brewing process based on all models according to the real-time parameter information, and performing variable control on corresponding links of the wine brewing process according to the parameter optimization results.
The optimization algorithm runs in an optimization server, required historical parameter information is obtained from an MES System, required real-time parameter information is obtained from a DCS System, a generated optimization result is sent to the DCS System to serve as a set value of a control parameter of a wine brewing production process, variable control is carried out according to the set value to improve the utilization rate of raw grains, the proportion of high quality (proportion of two-section wine and three-section wine) and energy consumption, and further achieve an expected optimization target, wherein RBF is called Radial basic function completely, a Chinese name Radial basis function is used as an RBF model for saccharification, the RBF model is a neural Network model taking the Radial basis function as an activation function, ANFIS is called Adaptive Network-based Fuzzy basis System, a Chinese name Adaptive neural Fuzzy System is used for providing a tank, the ANFIS model is used for combining a neural Network with Fuzzy organic reasoning so that the neural Network and the Fuzzy basis are developed in the self-Adaptive, self-organizing and self-learning directions, L S-SVM is called L full support of supply, the concept of the Chinese Network and the feedback algorithm is called a minimum of the feedback gradient Vector propagation of the multi-layer neural Network, the minimum of the feedback error, the minimum of the feedback algorithm, the minimum of the feedback, and the minimum of the feedback algorithm, the minimum of the feedback gradient error, the minimum of the feedback, the minimum of the feedback, the minimum of the.
Examples
The brewing process optimization method based on big data analysis, as shown in fig. 1, includes the following steps:
s1, obtaining historical parameter information of a wine brewing process, wherein the wine brewing process at least comprises a saccharification link, a vinasse preparation link, a fermentation link and a distillation link;
the historical parameter information is obtained from an MES system, namely a production process management system in the wine brewing process.
S2, respectively establishing an RBF model for a saccharification link, an ANFIS model for a vinasse preparation link, an L S-SVM model for a fermentation link and a BP neural network model for a distillation link according to the historical parameter information;
specifically, the historical parameter information may include: the method comprises the following steps of establishing an ANFIS model for a grain blending link, wherein the ANFIS model comprises the following steps of grain saccharification picture information, mother grain water content, grain temperature, mother grain starch content and grain blending proportion:
acquiring gray level co-occurrence matrix contrast characteristics and HSV information in the grain saccharified picture information, wherein the formula is as follows;
Figure BDA0002454034530000061
wherein F represents the contrast characteristic of gray level co-occurrence matrix in the picture information after grain saccharification, and PijThe number of times of occurrence of each (i, j) value in the picture information after grain saccharification.
And taking the gray level co-occurrence matrix contrast characteristic, HSV information, the water content of the mother grains, the grain temperature and the starch content of the mother grains as the input of an ANFIS model, and taking the grain preparation proportion as the output of the ANFIS model for training to obtain the ANFIS model for the grain preparation link.
Wherein. The ANFIS model is divided into five layers, including: a fuzzy layer, a rule reasoning layer, a normalization layer, a de-fuzzy layer and an output layer, wherein a membership function u is customized for each fuzzy subsetA(x，θ1) Wherein, theta1＝(a1i，a2i，b1i，b2i，c1i，c2i) Setting initial value of parameter theta for former coefficient according to artificial experience. Common membership functions include a triangular function, a trapezoidal function, a gaussian function, a bell-shaped function, and the like, and when the membership function is the bell-shaped function, the output formula of the fuzzy layer is as follows:
Figure BDA0002454034530000062
wherein i is 1, 2;
each node of the rule inference layer corresponds to a fuzzy rule and is used for calculating the excitation strength of each fuzzy rule, namely the product of membership degrees of each precondition corresponding to a fuzzy subset in the rule, and the calculation formula is as follows:
Figure BDA0002454034530000063
wherein i is 1, 2;
the normalization layer is used for normalizing the excitation intensity, calculating to obtain a weighting coefficient output by each fuzzy rule, and calculating the formula as follows:
Figure BDA0002454034530000064
the de-blurring layer is used for blurring the image according to a function f (x, R, theta) in a blurring rule2) Converting the fuzzy output to a precise output, wherein θ2(p, q, r) are the parameters of the back piece, and the conversion formula is as follows:
Figure BDA0002454034530000071
the output layer is used for summing output results of the de-fuzzy layer, and the summation formula is as follows:
Figure BDA0002454034530000072
the learning algorithm of the ANFIS model is a BP and L SM mixed algorithm, and the learning steps are as follows:
A. fixing the front piece coefficient to propagate forward to obtain an output for adjusting the back piece parameter, wherein the output is a linear combination of the back piece parameter, and an equation is as follows:
Figure BDA0002454034530000073
the equation y is a × θ1The least squares solution of (c) is:
θ1＝A+y+(I-A+A)z；
in the formula, A+Moore-Penrose generalized inverse of A, z is any real number, and the minimum two-times solution calculation formula is as follows:
x＝A+b；
B. and adjusting the parameters of the back piece according to the output, calculating errors according to the output after the parameters of the back piece are adjusted, reversely propagating the errors to the fuzzy layer, and updating the parameters of the front piece by using a gradient descent method.
It is to be understood that the historical parameter information may further include: the method comprises the following steps of saccharifying time, water adding temperature in a saccharifying link, grain preparation proportion, mother grain starch content and pit entering grain starch content, wherein the method for establishing the RBF model for the saccharifying link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link, the ratio of the prepared grains and the starch content of the mother grains as the input of an RBF model, and taking the starch content of the fermented grains entering the cellar as the output of the RBF model for training to obtain the RBF model for the saccharification link.
For the RBF model, model deviation of the PBF model can be calculated based on historical parameter information, a correction model is established according to the model deviation, and the RBF model is corrected according to the correction model. Further reducing parameter optimization errors and improving optimization effect of saccharification link.
It is understood that the historical parameter information can also comprise saccharification time, water adding temperature of a saccharification link, fermentation time, cellaring alcohol concentration, butyl butyrate concentration and ethanol concentration, and the method for establishing the L S-SVM model for the fermentation link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link and the fermentation time as the input of an L S-SVM model, and taking the cellaring alcohol concentration, the butyl butyrate concentration and the ethanol concentration as the output of a L S-SVM model for training to obtain a L S-SVM model for the fermentation link.
The specific establishing steps of the L S-SVM model for the fermentation model are as follows:
1. establishing a dynamic model based on L logistic equation:
Figure BDA0002454034530000081
wherein x is the concentration of the cellaring alcohol, the concentration of butyl butyrate and the concentration of ethanol, s is the concentration of starch, u is the growth rate, y is the yield coefficient, and u ismaxTo maximize growth rate, KsIs the saturation coefficient, KipIs the product inhibition constant;
2. the mechanism model and the L S-SVM model form a mixed model, and the final output of the mixed model is
Figure BDA0002454034530000082
Figure BDA0002454034530000083
For the purpose of the output of the mechanism model,
Figure BDA0002454034530000084
selecting error estimation of L S-SVM model to actual value and mechanism model output
Figure BDA0002454034530000085
Is a sample set, where xi＝x，eiThe expression is as follows:
Figure BDA0002454034530000086
wherein the content of the first and second substances,
Figure BDA0002454034530000087
indicating that the output error of the last L S-SVM model is within a preset range,
Figure BDA0002454034530000088
indicating that the output error of the L S-SVM model at the last time exceeds a preset range.
L the expression of the optimization problem of the S-SVM model is as follows:
Figure BDA0002454034530000089
Figure BDA00024540345300000810
where ξ i is the deviation and r is the penalty factor, its lagrangian function is defined as follows:
Figure BDA00024540345300000811
where ai is the Lagrangian multiplier, which is obtained under the KKT condition:
Figure BDA0002454034530000091
eliminating the variables w and ξ i, the following equation can be obtained:
Figure BDA0002454034530000092
wherein e ═ e1，e2，...，el]T；Z＝[1，1，...，1]；a＝[a1，a2，...，an]T；
For the elements in the matrix Ω are represented by gaussian radial basis kernel functions:
Figure BDA0002454034530000093
the values of ai and b can be found for a given r-sum, resulting in an error estimation model:
Figure BDA0002454034530000094
l, the selection of a penalty factor r and a kernel function parameter sigma in the S-SVM model is particularly important, in actual use, the sum of squares of prediction errors of a whole sample is minimized as a target, and the r and the sigma are selected by an optimization algorithm based on gradient.
It is to be understood that the historical parameter information may further include: the method for establishing the BP neural network model for the distillation link comprises the following steps of:
and taking the steam flow and the distillation time as the input of a BP neural network model, and taking the two-stage liquor yield and the three-stage liquor yield as the output of the BP neural network model for training to obtain the BP neural network model for the distillation link.
And S3, acquiring real-time parameter information of the wine brewing process, calculating parameter optimization results of all links of the wine brewing process based on all models according to the real-time parameter information, and performing variable control on the corresponding links of the wine brewing process according to the parameter optimization results.
It can be understood that after all models are built, 3 comprehensive indexes of the utilization rate of raw grains, the high-quality ratio (the second-stage wine yield and the third-stage wine yield) and the energy consumption are taken as targets, the parameter optimization results corresponding to all links are obtained according to the real-time parameter information of the wine brewing process obtained from the DCS and the models of all links, and the production process control parameters corresponding to all links are adjusted according to the parameter optimization results to achieve the expected optimization target.
Specifically, after the L S-SVM model is built, the L S-SVM model can be optimized through a genetic algorithm, the corresponding saccharifying time, the water adding temperature in the saccharifying link and the fermenting time when the cellar-out alcohol concentration, the butyl butyrate concentration and the ethanol concentration are the maximum are obtained, and the corresponding saccharifying time, the corresponding water adding temperature in the saccharifying link and the fermenting time when the corresponding cellar-out alcohol concentration, the butyl butyrate concentration and the ethanol concentration are the maximum are used as parameter optimization results to carry out variable control on the fermenting link, so that the maximum cellar-out alcohol concentration, the butyl butyrate concentration and the ethanol concentration are obtained.
After the BP neural network model is established, optimizing the BP neural network model through a particle swarm algorithm to obtain the steam flow and the distillation time corresponding to the maximum two-section wine yield and the maximum three-section wine yield, and performing variable control on the distillation link by taking the corresponding steam flow and the corresponding distillation time as parameter optimization results to obtain the maximum two-section wine yield and the maximum three-section wine yield.

Claims (10)

1. The brewing process optimization method based on big data analysis is characterized by comprising the following steps:
step 1, obtaining historical parameter information of a wine brewing process, wherein the wine brewing process at least comprises a saccharification link, a vinasse preparation link, a fermentation link and a distillation link;
step 2, respectively establishing an RBF model for a saccharification link, an ANFIS model for a vinasse preparation link, an L S-SVM model for a fermentation link and a BP neural network model for a distillation link according to the historical parameter information;
and 3, acquiring real-time parameter information of the brewing process, calculating parameter optimization results of all links of the brewing process based on all models according to the real-time parameter information, and performing variable control on corresponding links of the brewing process according to the parameter optimization results.
2. The big data analysis-based brewing process optimization method according to claim 1, wherein the historical parameter information comprises: the method comprises the following steps of establishing an ANFIS model for a grain blending link, wherein the ANFIS model comprises the following steps of grain saccharification picture information, mother grain water content, grain temperature, mother grain starch content and grain blending proportion:
acquiring gray level co-occurrence matrix contrast characteristics and HSV information in the grain saccharified picture information;
and taking the gray level co-occurrence matrix contrast characteristic, HSV information, the water content of the mother grains, the grain temperature and the starch content of the mother grains as the input of an ANFIS model, and taking the grain preparation proportion as the output of the ANFIS model for training to obtain the ANFIS model for the grain preparation link.
3. The big data analysis-based brewing process optimization method according to claim 2, wherein the ANFIS model is divided into five layers, including: a fuzzy layer, a rule reasoning layer, a normalization layer, a de-fuzzy layer and an output layer, wherein a membership function u is customized for each fuzzy subsetA(x，θ1) Wherein, theta1＝(a1i，a2i，b1i，b2i，c1i，c2i) When the membership function is a bell-shaped function, the output formula of the fuzzy layer is as follows:
Figure FDA0002454034520000011
wherein i is 1, 2;
each node of the rule inference layer corresponds to a fuzzy rule and is used for calculating the excitation strength of each fuzzy rule, and the calculation formula is as follows:
Figure FDA0002454034520000012
wherein i is 1, 2;
the normalization layer is used for normalizing the excitation intensity, calculating to obtain a weighting coefficient output by each fuzzy rule, and calculating the formula as follows:
Figure FDA0002454034520000021
the de-blurring layer is used for blurring the image according to a function f (x, R, theta) in a blurring rule2) Converting the fuzzy output to a precise output, wherein θ2(p, q, r) are the parameters of the back piece, and the conversion formula is as follows:
Figure FDA0002454034520000022
the output layer is used for summing output results of the de-fuzzy layer, and the summation formula is as follows:
Figure FDA0002454034520000023
4. the big data analysis-based brewing process optimization method according to claim 3, wherein the learning algorithm of the ANFIS model is a BP and L SM hybrid algorithm, and the learning steps are as follows:
A. fixing the front piece coefficient to propagate forward to obtain an output for adjusting the back piece parameter, wherein the output is a linear combination of the back piece parameter, and an equation is as follows:
Figure FDA0002454034520000024
the equation y is a × θ1The least squares solution of (c) is:
θ1＝A+y+(I-A+A)z；
in the formula, A+Moore-Penrose generalized inverse of A, z is any real number, and the minimum two-times solution calculation formula is as follows:
x＝A+b；
B. and adjusting the parameters of the back piece according to the output, calculating errors according to the output after the parameters of the back piece are adjusted, reversely propagating the errors to the fuzzy layer, and updating the parameters of the front piece by using a gradient descent method.
5. The big data analysis-based brewing process optimization method according to claim 1, wherein the historical parameter information comprises: the method comprises the following steps of saccharifying time, water adding temperature in a saccharifying link, grain preparation proportion, mother grain starch content and pit entering grain starch content, wherein the method for establishing the RBF model for the saccharifying link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link, the ratio of the prepared grains and the starch content of the mother grains as the input of an RBF model, and taking the starch content of the fermented grains entering the cellar as the output of the RBF model for training to obtain the RBF model for the saccharification link.
6. The big data analysis-based brewing process optimization method according to claim 5, further comprising:
and calculating the model deviation of the PBF model based on historical parameter information, establishing a correction model according to the model deviation, and correcting the RBF model according to the correction model.
7. The big data analysis-based wine brewing process optimization method as claimed in claim 1, wherein the historical parameter information comprises saccharification time, water adding temperature in the saccharification link, fermentation time, cellaring alcohol concentration, butyl butyrate concentration and ethanol concentration, and the method for establishing the L S-SVM model for the fermentation link comprises the following steps:
and taking the saccharification time, the water adding temperature in the saccharification link and the fermentation time as the input of an L S-SVM model, and taking the cellaring alcohol concentration, the butyl butyrate concentration and the ethanol concentration as the output of a L S-SVM model for training to obtain a L S-SVM model for the fermentation link.
8. The big data analysis-based wine brewing process optimization method as claimed in claim 7, further comprising optimizing the L S-SVM model by a genetic algorithm to obtain the saccharifying time, the water adding temperature in the saccharifying link and the fermentation time corresponding to the maximum pit-out alcohol concentration, the maximum butyl butyrate concentration and the maximum ethanol concentration, and performing variable control on the fermentation link by using the saccharifying time, the water adding temperature in the saccharifying link and the fermentation time as parameter optimization results.
9. The big data analysis-based brewing process optimization method according to claim 1, wherein the historical parameter information comprises: the method for establishing the BP neural network model for the distillation link comprises the following steps of:
and taking the steam flow and the distillation time as the input of a BP neural network model, and taking the two-stage liquor yield and the three-stage liquor yield as the output of the BP neural network model for training to obtain the BP neural network model for the distillation link.
10. The big data analysis-based brewing process optimization method according to claim 9, further comprising: and optimizing the BP neural network model by a particle swarm algorithm to obtain the steam flow and the distillation time corresponding to the maximum two-stage liquor yield and the maximum three-stage liquor yield, and performing variable control on the distillation link by taking the steam flow and the distillation time as parameter optimization results.
CN202010301141.2A 2020-04-16 2020-04-16 Big data analysis-based brewing process optimization method Active CN111476428B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010301141.2A CN111476428B (en) 2020-04-16 2020-04-16 Big data analysis-based brewing process optimization method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010301141.2A CN111476428B (en) 2020-04-16 2020-04-16 Big data analysis-based brewing process optimization method

Publications (2)

Publication Number Publication Date
CN111476428A true CN111476428A (en) 2020-07-31
CN111476428B CN111476428B (en) 2022-07-08

Family

ID=71753686

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010301141.2A Active CN111476428B (en) 2020-04-16 2020-04-16 Big data analysis-based brewing process optimization method

Country Status (1)

Country Link
CN (1) CN111476428B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112544949A (en) * 2020-12-10 2021-03-26 广西顶俏食品科技集团有限公司 Environment-friendly treatment method and device for precipitate in soy sauce production

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01257466A (en) * 1988-01-23 1989-10-13 Kyowa Hakko Kogyo Co Ltd Preparation of liquors
CN101763036A (en) * 2009-12-29 2010-06-30 江苏大学 Lysine fermentation process feeding prediction control system and method based on fuzzy neural network
CN105259827A (en) * 2015-10-16 2016-01-20 江苏大学 Real-time monitoring system and real-time monitoring method for solid fermentation process condition
CN106052753A (en) * 2016-05-18 2016-10-26 江苏大学 Straw fermentation fuel ethanol production process key state variable soft measuring method based on fuzzy support vector machine
CN106803023A (en) * 2017-02-07 2017-06-06 秦皇岛华恒生物工程有限公司 The method that alanine is produced based on the feed supplement of neural network combined with genetic algorithm real-time optimization
CN107063991A (en) * 2017-04-14 2017-08-18 中国石油天然气股份有限公司 A kind of conveyance conduit internal corrosion defect dynamic security assessment method and device
CN107325930A (en) * 2017-09-08 2017-11-07 泸州老窖集团有限责任公司 A kind of brewing method of Chinese yeast fen-flavor type white spirit
CN108090500A (en) * 2017-11-22 2018-05-29 江苏大学 The flexible measurement method of edible fungus fermented production process key parameters
CN109376805A (en) * 2018-12-21 2019-02-22 四川理工学院 A kind of classification method based on white wine base liquor Fingerprints
CN109652265A (en) * 2019-02-20 2019-04-19 泸州品创科技有限公司 Pure and sweet type white wine and its brewing method
CN110415762A (en) * 2019-08-06 2019-11-05 苏州大学 A kind of system and method based on sequence prediction temperature of protein denaturation

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01257466A (en) * 1988-01-23 1989-10-13 Kyowa Hakko Kogyo Co Ltd Preparation of liquors
CN101763036A (en) * 2009-12-29 2010-06-30 江苏大学 Lysine fermentation process feeding prediction control system and method based on fuzzy neural network
CN105259827A (en) * 2015-10-16 2016-01-20 江苏大学 Real-time monitoring system and real-time monitoring method for solid fermentation process condition
CN106052753A (en) * 2016-05-18 2016-10-26 江苏大学 Straw fermentation fuel ethanol production process key state variable soft measuring method based on fuzzy support vector machine
CN106803023A (en) * 2017-02-07 2017-06-06 秦皇岛华恒生物工程有限公司 The method that alanine is produced based on the feed supplement of neural network combined with genetic algorithm real-time optimization
CN107063991A (en) * 2017-04-14 2017-08-18 中国石油天然气股份有限公司 A kind of conveyance conduit internal corrosion defect dynamic security assessment method and device
CN107325930A (en) * 2017-09-08 2017-11-07 泸州老窖集团有限责任公司 A kind of brewing method of Chinese yeast fen-flavor type white spirit
CN108090500A (en) * 2017-11-22 2018-05-29 江苏大学 The flexible measurement method of edible fungus fermented production process key parameters
CN109376805A (en) * 2018-12-21 2019-02-22 四川理工学院 A kind of classification method based on white wine base liquor Fingerprints
CN109652265A (en) * 2019-02-20 2019-04-19 泸州品创科技有限公司 Pure and sweet type white wine and its brewing method
CN110415762A (en) * 2019-08-06 2019-11-05 苏州大学 A kind of system and method based on sequence prediction temperature of protein denaturation

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
AHMAD JAHANBAKHSHI等: "Processing watermelon waste using Saccharomyces cerevisiae yeast and the fermentation method for bioethanol production", 《JOURNAL OF FOOD PROCESS ENGINEERING》 *
HAN YAN等: "Adaptive neuro fuzzy inference system for classification of water quality status", 《JOURNAL OF ENVIRONMENTAL SCIENCES》 *
吕美: "黄酒发酵过程酸败预测及相关参数检测技术研究", 《中国优秀硕士学位论文全文数据库 工程科技I辑》 *
张良等: "浓香型白酒固态发酵规律模型的预测研究", 《酿酒》 *
王晓等: "果酒品质评价的自适应量子粒子群LS-SVM模型", 《科学技术与工程》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112544949A (en) * 2020-12-10 2021-03-26 广西顶俏食品科技集团有限公司 Environment-friendly treatment method and device for precipitate in soy sauce production
CN112544949B (en) * 2020-12-10 2022-10-11 广西顶俏食品科技集团有限公司 Environment-friendly treatment method and device for precipitate in soy sauce production

Also Published As

Publication number Publication date
CN111476428B (en) 2022-07-08

Similar Documents

Publication Publication Date Title
Dodić et al. Kinetic modelling of batch ethanol production from sugar beet raw juice
CN101689045B (en) Nonlinear model predictive control of a biofuel fermentation process
US7933849B2 (en) Integrated model predictive control of batch and continuous processes in a biofuel production process
Veloso et al. Fed-batch ethanol fermentation at low temperature as a way to obtain highly concentrated alcoholic wines: Modeling and optimization
CN101539776B (en) Method for realizing computer
Carreño-Zagarra et al. Linear active disturbance rejection control for a raceway photobioreactor
Ochoa et al. Integrating real-time optimization and control for optimal operation: Application to the bio-ethanol process
Vikash et al. Conceptual design of a lignocellulosic biorefinery and its supply chain for ethanol production in India
CN111476428B (en) Big data analysis-based brewing process optimization method
Murthy et al. A simultaneous saccharification and fermentation model for dynamic growth environments
Shokrkar et al. Model‐based evaluation of continuous bioethanol production plant
Chen et al. Optimal trade-off design of integrated fermentation processes for ethanol production using genetically engineered yeast
Teissier et al. A hybrid recurrent neural network model for yeast production monitoring and control in a wine base medium
Liu et al. Fermentation process modeling with Levenberg‐Marquardt algorithm and Runge‐Kutta method on ethanol production by Saccharomyces cerevisiae
David et al. Nitrogen-backboned modeling of wine-making in standard and nitrogen-added fermentations
Niu et al. Modeling and parameter updating for nosiheptide fed-batch fermentation process
Boareto et al. A hybrid neural model (HNM) for the on‐line monitoring of lipase production by Candida rugosa
CN105740599B (en) A kind of Gene regulation, which is continuously fermented, mixes dynamical system
David et al. Modeling of the aromatic profile in wine-making fermentation: the backbone equations
Câmara et al. On-line identification of fermentation processes for ethanol production
Kiran et al. Control of continuous fed-batch fermentation process using neural network based model predictive controller
Sari et al. Reuse of wheat flour liquid waste for enzymatic hydrolysis to yield glucose-derived bioethanol
CN115389703A (en) Prediction method of ethanol discharge concentration of fermentation tank, control device and storage medium
Zhang et al. Reliable multi-objective on-line re-optimisation control of a fed-batch fermentation process using bootstrap aggregated neural networks
Murthy et al. Design and evaluation of an optimal controller for simultaneous saccharification and fermentation process

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant