CN102058012A - Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system - Google Patents

Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system Download PDF

Info

Publication number
CN102058012A
CN102058012A CN2010105425955A CN201010542595A CN102058012A CN 102058012 A CN102058012 A CN 102058012A CN 2010105425955 A CN2010105425955 A CN 2010105425955A CN 201010542595 A CN201010542595 A CN 201010542595A CN 102058012 A CN102058012 A CN 102058012A
Authority
CN
China
Prior art keywords
oyster
enzymatic hydrolysis
network system
gustin
meant
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.)
Pending
Application number
CN2010105425955A
Other languages
Chinese (zh)
Inventor
秦小明
林华娟
章超桦
侯清娥
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.)
Guangdong Ocean University
Original Assignee
Guangdong Ocean University
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 Guangdong Ocean University filed Critical Guangdong Ocean University
Priority to CN2010105425955A priority Critical patent/CN102058012A/en
Publication of CN102058012A publication Critical patent/CN102058012A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Meat, Egg Or Seafood Products (AREA)

Abstract

本发明涉及食品技术领域,特别是一种基于神经网络系统优化的牡蛎呈味肽可控酶解工艺。本发明基于神经网络系统优化的牡蛎呈味肽可控酶解工艺,BP神经网络学习过程由输入数据的正向传播和误差的反向传播的过程中经过隐含层的处理,最终实现了复杂的非确定性酶解问题的模拟。该方法模拟人的大脑判断系统,用高精度的实时模拟处理数据,在不需要精确地数学模型的基础实现了酶解因素与肽含量及感官分值之间的非线性映射关系;并有精确地预测能力,避免减少了一些实际工艺的实施以及在人为感官评定中的一些弊端和局限性,直接达到快速准确的预测仿真效果。采用本发明得到的酶解工艺更为科学,因此可提高工艺生产效率和产品质量,降低生产成本。

Figure 201010542595

The invention relates to the field of food technology, in particular to a controllable enzymatic hydrolysis process of oyster taste peptides optimized based on a neural network system. The present invention is based on the controllable enzymatic hydrolysis process of oyster flavor peptides optimized by the neural network system. The BP neural network learning process is processed by the hidden layer in the process of forward propagation of input data and reverse propagation of errors, and finally realizes complex Simulation of the nondeterministic enzymatic hydrolysis problem. This method simulates the human brain judgment system, uses high-precision real-time simulation to process data, and realizes the nonlinear mapping relationship between enzymatic hydrolysis factors, peptide content and sensory scores without the need for precise mathematical models; and has accurate It avoids reducing the implementation of some actual processes and some drawbacks and limitations in human sensory evaluation, and directly achieves fast and accurate prediction simulation results. The enzymolysis process obtained by adopting the invention is more scientific, so the process production efficiency and product quality can be improved, and the production cost can be reduced.

Figure 201010542595

Description

一种基于神经网络系统优化的牡蛎呈味肽可控酶解工艺 A controllable enzymatic hydrolysis process of oyster taste peptide based on neural network system optimization

技术领域technical field

本发明涉及食品技术领域,特别是一种基于神经网络系统优化的牡蛎呈味肽可控酶解工艺。The invention relates to the field of food technology, in particular to a controllable enzymatic hydrolysis process of oyster taste peptides optimized based on a neural network system.

背景技术Background technique

牡蛎及其牡蛎制品以其独特风味和营养价值倍受消费者喜爱,尤其是经过发酵后的牡蛎味道极其鲜美,自古以来古人就将牡蛎加工成蚝油、蚝豉等发酵调味品。现代科学研究结果显示,蚝油、蚝豉等牡蛎发酵调味品的优良风味与牡蛎发酵过程中产生的呈味肽有密切相关。由于传统发酵工艺机械化程度低,耗时长,生产效率低下,目前现代化的酶工程技术正在逐步取代传统的发酵工艺。但是如何最大限度提高呈味肽产出率,并实现呈味肽可控酶解是牡蛎调味精深加工中亟待解决的关键技术问题。牡蛎呈味肽酶解工艺研究结果表明,加酶量、料水比、酶解温度以及酶解时间等多种因素均较大程度,并且呈非线性地影响呈味肽及其风味的产生。但是目前许多传统的酶解工艺条件优化均采取正交实验方法获得的最佳工艺方案。正交实验虽然具备“均匀分散,齐整可比”的特点,但是由于蛋白酶在酶动力学的复杂性以及非线性,导致正交实验获得的最佳因素条件不一定是最佳因素条件。人工神经网络由于其特有的非线性适应性信息处理能力,具有自学习功能、联想存储功能和高速寻找优化解能力等优势而在自动控制领域、处理组合优化、图像处理等方面得到广泛应用。可以推测人工神经网络在酶工程技术中同样有广阔的应用前景。但是目前神经网络在酶工程技术中的研究仍处于初期阶段,在牡蛎呈味肽酶解工艺优化应用中尚无任何报道。Oysters and oyster products are loved by consumers for their unique flavor and nutritional value, especially the fermented oysters are extremely delicious. Since ancient times, the ancients have processed oysters into fermented condiments such as oyster sauce and oyster sauce. Modern scientific research results show that the excellent flavor of oyster fermented condiments such as oyster sauce and dried oyster is closely related to the taste peptides produced during the fermentation process of oysters. Due to the low degree of mechanization, long time consumption and low production efficiency of the traditional fermentation process, the current modern enzyme engineering technology is gradually replacing the traditional fermentation process. However, how to maximize the yield of taste peptides and realize the controllable enzymatic hydrolysis of taste peptides is a key technical problem to be solved in the deep processing of oyster seasoning. The research results of the enzymatic hydrolysis process of oyster flavor peptides showed that the amount of enzyme added, the ratio of material to water, the enzymatic hydrolysis temperature and the enzymatic hydrolysis time and other factors were all to a great extent and non-linearly affected the production of flavor peptides and their flavor. But at present, many traditional enzymatic hydrolysis process conditions are optimized by adopting the best process plan obtained by orthogonal experiment method. Although the orthogonal experiment has the characteristics of "uniform dispersion, neat and comparable", due to the complexity and nonlinearity of protease kinetics, the optimal factor conditions obtained by the orthogonal experiment are not necessarily the optimal factor conditions. Due to its unique nonlinear adaptive information processing ability, artificial neural network has the advantages of self-learning function, associative storage function and high-speed search for optimal solution, and has been widely used in the field of automatic control, processing combination optimization, image processing, etc. It can be speculated that the artificial neural network also has broad application prospects in enzyme engineering technology. However, the research of neural network in enzyme engineering technology is still in the initial stage, and there is no report on the application of oyster flavor peptide enzymatic hydrolysis process optimization.

发明内容Contents of the invention

本发明的目的是为了弥补上述现有技术存在的不足,提供一种基于神经网络系统优化牡蛎呈味肽可控酶解工艺技术方法,使酶解工艺更为科学,提高工艺生产效率和产品质量,降低生产成本。The purpose of the present invention is to make up for the deficiencies in the above-mentioned prior art, and provide a controllable enzymolysis process technology method based on a neural network system to optimize the oyster flavor peptide, so that the enzymolysis process is more scientific, and the process production efficiency and product quality are improved. ,reduce manufacturing cost.

为实现上述发明目的,本发明采取的技术方案是该基于神经网络系统优化的牡蛎呈味肽可控酶解工艺包括如下步骤:In order to achieve the purpose of the above invention, the technical solution adopted by the present invention is that the controllable enzymolysis process of oyster taste peptide based on neural network system optimization includes the following steps:

(1)以牡蛎肉为原料,以料比水、蛋白酶添加量、酶反应温度和酶反应时间等4因素为酶解工艺条件参数,以肽比例和感官评分作为牡蛎酶解液呈味肽评价指标,采用正交实验和随机酶解实验获得酶解工艺参数与评价指标的一一对应关系;(1) Using oyster meat as raw material, four factors including material ratio water, protease addition amount, enzyme reaction temperature and enzyme reaction time were used as enzymatic hydrolysis process parameters, and peptide ratio and sensory score were used as the evaluation of taste peptides in oyster enzymatic hydrolyzate Index, using orthogonal experiments and random enzymatic hydrolysis experiments to obtain the one-to-one correspondence between enzymatic hydrolysis process parameters and evaluation indicators;

(2)以料比水、蛋白酶添加量、酶反应温度和酶反应时间等4个工艺参数作为输入值,以牡蛎酶解液的肽比例和感观评分作为输出信号,建立牡蛎呈味肽可控酶解的BP神经网络结构模型训练样本;(2) Taking four process parameters as input values, such as material ratio water, protease addition amount, enzyme reaction temperature and enzyme reaction time, and taking the peptide ratio and sensory score of oyster enzymatic hydrolyzate as output signals, the oyster flavor peptide can be established. Enzyme-controlled BP neural network structure model training samples;

(3)对BP神经网络结构模型进行训练、仿真,最终实现酶解工艺参数到酶解指标的良好映射关系;(3) Train and simulate the BP neural network structure model, and finally realize a good mapping relationship between the enzymatic hydrolysis process parameters and the enzymatic hydrolysis index;

(4)在BP神经网络的基础上,以神经网络所得的函数作为遗传算法的适应度函数,分别以酶解液肽比例最大值和感官评分最高值为目标采用遗传算法进一步求得酶解工艺参数的最佳组合。(4) On the basis of the BP neural network, the function obtained by the neural network is used as the fitness function of the genetic algorithm, and the maximum value of the peptide ratio of the enzymatic hydrolysis solution and the highest sensory score are used as the targets to further obtain the enzymatic hydrolysis process by genetic algorithm the best combination of parameters.

所述牡蛎酶解工艺是指牡蛎肉的匀浆,调整料水比,添加蛋白酶,在一定条件下酶反应,沸水浴加热10分钟,冷却,在4000转/分钟条件下离心20分钟,取酶解上清液得牡蛎酶解液;其中料水比、蛋白酶添加量、酶反应温度和时间等4因素是牡蛎酶解工艺的变量参数。The oyster enzymatic hydrolysis process refers to the homogenization of oyster meat, adjusting the ratio of feed to water, adding protease, enzymatic reaction under certain conditions, heating in a boiling water bath for 10 minutes, cooling, centrifuging at 4000 rpm for 20 minutes, and taking the enzyme The oyster enzymatic hydrolyzate was obtained by decomposing the supernatant, and the four factors including the ratio of material to water, the amount of protease added, the temperature and time of enzymatic reaction were the variable parameters of the oyster enzymatic hydrolysis process.

所述匀浆是指牡蛎肉经过自来水清洗、并沥干表面水分后,进行组织破碎、匀浆,得到牡蛎肉匀浆原液;The homogenization refers to that the oyster meat is washed with tap water and drained of surface water, then the tissue is broken and homogenized to obtain the oyster meat homogenate stock solution;

所述调整料水比是指牡蛎肉匀浆原液与添加水的份数比例;The ratio of adjusted material to water refers to the ratio of oyster meat homogenate stock solution and added water;

所述蛋白酶添加量是指每100克牡蛎肉匀浆原液添加蛋白酶的重量百分比例;The added amount of protease refers to the percentage by weight of protease added per 100 grams of oyster meat homogenate stoste;

所述酶反应时间是指牡蛎肉经匀浆、调整料水比等预处理后,从添加蛋白酶时刻开始进行酶反应的时间。The enzyme reaction time refers to the time for the oyster meat to undergo the enzyme reaction from the moment when the protease is added after pretreatment such as homogenization and adjustment of the feed-to-water ratio.

所述牡蛎酶解液的肽比例是指对牡蛎酶解液中的肽氮和总氮含量分别进行测定,肽氮含量占总氮含量的百分比即为肽比例。The peptide ratio of the oyster enzymatic hydrolyzate refers to the determination of the peptide nitrogen and total nitrogen contents in the oyster enzymatic hydrolyzate respectively, and the percentage of the peptide nitrogen content in the total nitrogen content is the peptide ratio.

所述牡蛎酶解液的感官评分是指通过感官评定方法对牡蛎酶解液进行感官评价得到的分值。The sensory score of the oyster hydrolyzate refers to the score obtained by sensory evaluation of the oyster hydrolyzate by a sensory evaluation method.

所述感官评定是指以不同浓度(0.2~1.6g/L)的谷氨酸钠溶液所对应的鲜度作为标准,谷氨酸钠溶液浓度为1.6g/L的鲜度为最高值(8分),谷氨酸钠溶液浓度为0.2g/L的鲜度为最低值(1分),其他各浓度相对应的分值范围为1~8。感官评定员均经过基本滋味培训,并至少由15人以上组成。牡蛎酶解液通过以上方法进行感官评分,评分的原始数据采用狄克松法进行检验处理,最终取平均值。Described sensory evaluation refers to the freshness corresponding to the sodium glutamate solution of different concentrations (0.2~1.6g/L) as a standard, and the freshness of the sodium glutamate solution concentration is 1.6g/L is the highest value (8 points), the freshness of sodium glutamate solution with a concentration of 0.2g/L is the lowest value (1 point), and the corresponding scores for other concentrations range from 1 to 8. The sensory assessors have all received basic taste training and are composed of at least 15 people. The oyster enzymatic hydrolyzate was subjected to sensory evaluation by the above methods, and the raw data of the evaluation were inspected and processed by the Dixon method, and finally the average value was taken.

所述步骤(2)BP神经网络的结构模型构建是指应用MATLAB7.8版软件,在建立神经网络模型时以料水比、蛋白酶添加量、酶反应时间以及酶反应温度等4因素为输入量(神经元个数为4),分别以肽比例或感官评分值为输出层(神经元个数为1),构建四输入单输出的三层BP神经网络,并以输入层和隐含层之间的传递函数为正切S形函数(tansig),隐含层和输出层之间的传递函数为线性函数(purelin)。The structural model construction of described step (2) BP neural network refers to application MATLAB7.8 version software, is input quantity with 4 factors such as material-water ratio, protease addition, enzyme reaction time and enzyme reaction temperature when setting up neural network model (the number of neurons is 4), and the peptide ratio or sensory score is used as the output layer respectively (the number of neurons is 1), and a three-layer BP neural network with four inputs and one output is constructed, and the difference between the input layer and the hidden layer The transfer function between them is a tangent S-shaped function (tansig), and the transfer function between the hidden layer and the output layer is a linear function (purelin).

所述步骤(3)结构模型训练是指采用trainlm函数为网络训练函数,设定训练目标误差为0.0001,进行仿真训练。使BP神经网络通过正向传播和反向传播的不断学习过程,逐一修改各神经元连接的权值,这种过程不断迭代,最后实现误差信号达到允许的范围之内0.0001时,就停止训练。The step (3) structural model training refers to adopting the trainlm function as the network training function, setting the training target error as 0.0001, and performing simulation training. Make the BP neural network modify the weights of each neuron connection one by one through the continuous learning process of forward propagation and back propagation. This process is iterated continuously, and finally the training is stopped when the error signal reaches 0.0001 within the allowable range.

所述步骤(4)遗传算法是指以建立的神经网络模型为遗传算法的适应度函数,以各酶解参数水平范围为约束条件,以酶解液肽比例最大值和感官评分最高值为优化目标建立优化模型,确定可行解的编码方法及解码方法,同时设定遗传算法相关参数。The step (4) genetic algorithm refers to taking the established neural network model as the fitness function of the genetic algorithm, taking the level range of each enzymatic hydrolysis parameter as a constraint condition, and taking the maximum value of the enzymatic hydrolysis liquid peptide ratio and the highest sensory score as the optimal value The goal is to establish an optimization model, determine the coding method and decoding method of the feasible solution, and set the relevant parameters of the genetic algorithm at the same time.

所述设定遗传算法相关参数是指设定生成初始种群、交叉概率和变异概率等参数。生成初始种群,即群体中所含个数的数量,一般取20~160,其过小将影响搜索范围,从而得不到最优解,过大则搜索时间长,效率低;交叉概率和变异概率,取0~1之间,两者越大,则算法探测能力越强,越容易探测到新的超平面,但个体的平均适应度波动较大,相反越小则算法的开发能力越强,使得较优个体不易被破坏,个体的平均适应度平衡。The setting of related parameters of the genetic algorithm refers to setting parameters such as initial population generation, crossover probability and mutation probability. Generate the initial population, that is, the number of people contained in the population, generally 20 to 160, if it is too small, it will affect the search range, so that the optimal solution cannot be obtained, if it is too large, the search time will be long and the efficiency will be low; crossover probability and mutation probability , between 0 and 1, the larger the two, the stronger the detection ability of the algorithm, and the easier it is to detect new hyperplanes, but the average fitness of individuals fluctuates greatly, on the contrary, the smaller the algorithm is, the stronger the development ability is. The better individual is not easy to be destroyed, and the average fitness of the individual is balanced.

将设定好的参数及程序放在MATLAB软件中运行,由多个个体组成的一个初始种群开始最优搜索过程,并对这个群体进行的选择、交叉、变异等运算,产生出新一代的群体,继续多点的搜索,经过多次的试算后及合理参数的设定下,最终由遗传算法得出一个稳定的最优因素组合。Run the set parameters and programs in MATLAB software, start the optimal search process with an initial population composed of multiple individuals, and perform operations such as selection, crossover, and mutation on this population to generate a new generation of population , continue to search at multiple points, after several trial calculations and reasonable parameter settings, a stable optimal factor combination is finally obtained by the genetic algorithm.

本发明基于神经网络系统优化的牡蛎呈味肽可控酶解工艺,BP神经网络学习过程由输入数据的正向传播和误差的反向传播的过程中经过隐含层的处理,最终实现了复杂的非确定性酶解问题的模拟。该方法模拟人的大脑判断系统,用高精度的实时模拟处理数据,在不需要精确地数学模型的基础实现了酶解因素与肽含量及感官分值之间的非线性映射关系;并有精确地预测能力,避免减少了一些实际工艺的实施以及在人为感官评定中的一些弊端和局限性,直接达到快速准确的预测仿真效果。采用本发明得到的酶解工艺更为科学,因此可提高工艺生产效率和产品质量,降低生产成本。The present invention is based on the controllable enzymatic hydrolysis process of oyster flavor peptides optimized by the neural network system. The BP neural network learning process is processed by the hidden layer in the process of forward propagation of input data and reverse propagation of errors, and finally realizes complex Simulation of the nondeterministic enzymatic hydrolysis problem. This method simulates the human brain judgment system, uses high-precision real-time simulation to process data, and realizes the nonlinear mapping relationship between enzymatic hydrolysis factors, peptide content and sensory scores without the need for precise mathematical models; and has accurate It avoids reducing the implementation of some actual processes and some drawbacks and limitations in human sensory evaluation, and directly achieves fast and accurate prediction simulation results. The enzymolysis process obtained by adopting the invention is more scientific, so the process production efficiency and product quality can be improved, and the production cost can be reduced.

附图说明Description of drawings

图1为本发明的神经网络结构图Fig. 1 is the neural network structural diagram of the present invention

图2为本发明的神经网络及遗传算法结合的流程图Fig. 2 is the flow chart that neural network of the present invention and genetic algorithm combine

具体实施方式Detailed ways

下面结合实施例对本发明基于神经网络系统优化的牡蛎呈味肽可控酶解工艺方法作出详细说明。The controllable enzymatic hydrolysis process of oyster taste peptide based on neural network system optimization of the present invention will be described in detail below in conjunction with the examples.

(1)获取神经网络学习样本(1) Obtain neural network learning samples

牡蛎肉按照以下工艺流程制备牡蛎酶解液,然后测定牡蛎酶解液的肽氮含量和总氮含量,并计算牡蛎酶解液的肽比例,同时对牡蛎酶解液进行感官评定。获得酶解工艺参数与评价指标的一一对应关系,作为神经网络学习样本。Oyster meat was prepared according to the following process, and then the peptide nitrogen content and total nitrogen content of the oyster hydrolyzate were determined, and the peptide ratio of the oyster hydrolyzate was calculated, and the sensory evaluation of the oyster hydrolyzate was carried out at the same time. Obtain the one-to-one correspondence between enzymatic hydrolysis process parameters and evaluation indicators, and use them as neural network learning samples.

牡蛎酶解工艺流程:牡蛎肉(新鲜或冷冻)→清洗→沥干→匀浆→调整料水比→添加蛋白酶→在一定条件下酶反应(温度、时间)→沸水浴加热10min→冷却→离心(4000转/分钟、20min)→取酶解上清液→牡蛎酶解液。Oyster enzymatic hydrolysis process: oyster meat (fresh or frozen)→cleaning→draining→homogenization→adjusting material-water ratio→adding protease→enzyme reaction under certain conditions (temperature, time)→heating in boiling water bath for 10min→cooling→centrifugation (4000 rpm, 20min) → take the enzymatic hydrolysis supernatant → oyster enzymatic hydrolyzate.

操作要点如下:The main points of operation are as follows:

1)匀浆。牡蛎肉用自来水清洗、沥干表面水分后,用组织捣碎机或匀浆机进行组织破碎、匀浆,得到牡蛎肉匀浆原液。1) homogenate. After the oyster meat is washed with tap water and the surface water is drained, the tissue is crushed and homogenized by a tissue masher or a homogenizer to obtain the oyster meat homogenate stock solution.

2)调整料水比。按照不同料水比(1∶2,1∶3,1∶4,1∶5等比例),用纯净自来水调整牡蛎肉匀浆液浓度。2) Adjust the ratio of material to water. According to different material-water ratios (1:2, 1:3, 1:4, 1:5, etc.), the concentration of oyster meat homogenate was adjusted with pure tap water.

3)添加蛋白酶。以牡蛎肉匀浆原液的重量百分比计,按照不同酶添加量(0.3%、0.6%、0.9%、1.2%等),向牡蛎浆液中添加蛋白酶。3) Add protease. Add protease to oyster slurry according to different enzyme addition amounts (0.3%, 0.6%, 0.9%, 1.2%, etc.) based on the weight percentage of oyster meat homogenate stock solution.

4)酶反应温度。添加蛋白酶后的牡蛎浆液,放置于恒温水浴锅中,按照不同酶反应温度(45℃、50℃、55℃、60℃等)条件,调整酶反应温度。4) Enzyme reaction temperature. The oyster slurry after adding protease is placed in a constant temperature water bath, and the enzyme reaction temperature is adjusted according to different enzyme reaction temperatures (45°C, 50°C, 55°C, 60°C, etc.).

5)酶反应时间。从添加蛋白酶时刻开始计时,按照不同酶反应时间要求(4小时、5小时、6小时、7小时等)对牡蛎浆液进行酶解。5) Enzyme reaction time. Start timing from the moment of adding protease, and enzymatically hydrolyze the oyster slurry according to different enzyme reaction time requirements (4 hours, 5 hours, 6 hours, 7 hours, etc.).

6)牡蛎酶解液。牡蛎浆液按照不同工艺条件进行的酶解反应结束后,放入沸水浴中充分加热10min,冷却、离心,取其上清液即得到牡蛎酶解液。6) Oyster enzymatic hydrolyzate. After the enzymatic hydrolysis reaction of the oyster slurry is carried out according to different process conditions, it is placed in a boiling water bath and fully heated for 10 minutes, cooled and centrifuged, and the supernatant is taken to obtain the oyster enzymatic hydrolysis solution.

(2)建立神经网络模型(2) Establish a neural network model

以料比水、蛋白酶添加量、酶反应温度和酶反应时间等4个工艺参数作为输入值,以牡蛎酶解液的肽比例和感观评分作为输出信号,采用MATLAB7.8版软件,建立BP神经网络。Taking four process parameters as input values, such as material ratio water, protease addition amount, enzyme reaction temperature and enzyme reaction time, and taking the peptide ratio and sensory score of oyster enzymatic hydrolyzate as output signals, the BP was established by using MATLAB version 7.8 software. Neural Networks.

按照图1所示,将BP神经网络设计成3层网络:一个输入层,一个隐含层和一个输出层。输入层设定四个神经元:料比水、蛋白酶添加量、酶反应温度和酶反应时间。隐含层设定为13个神经元,牡蛎酶解液的肽比例和感观评分设定为输出层。输入层和隐含层之间的传递函数为正切S形函数(tansig),隐含层和输出层之间的传递函数为线性函数(purelin),网络训练函数采用trainlm函数。As shown in Figure 1, the BP neural network is designed as a 3-layer network: an input layer, a hidden layer and an output layer. The input layer sets four neurons: material ratio water, protease addition amount, enzyme reaction temperature and enzyme reaction time. The hidden layer was set as 13 neurons, and the peptide ratio and sensory score of oyster hydrolyzate were set as the output layer. The transfer function between the input layer and the hidden layer is a tangent S-shaped function (tansig), the transfer function between the hidden layer and the output layer is a linear function (purelin), and the network training function uses the trainlm function.

(3)神经网络训练(3) Neural network training

按照图2所示,用训练样本总量的80%数据对神经网络进行训练,设定一个牡蛎酶解液的肽比例期望值和目标精密度,然后让BP神经网络模型开始运行。通过调整训练步数、网络学习速率以及目标精密度,不断对神经网络进行训练,直至网络肽比例输出值与期望值之间的误差降低至5%以内。神经网络经过不断学习训练,获得最优的神经网络模型参数:最大训练步数为100,网络学习速率为0.1,网络性能目标误差为0.0001。As shown in Figure 2, use 80% of the total training sample data to train the neural network, set an expected value of the peptide ratio and target precision of the oyster hydrolyzate, and then let the BP neural network model start running. By adjusting the number of training steps, network learning rate and target precision, the neural network is continuously trained until the error between the network peptide ratio output value and the expected value is reduced to within 5%. After continuous learning and training, the neural network obtains the optimal neural network model parameters: the maximum number of training steps is 100, the network learning rate is 0.1, and the network performance target error is 0.0001.

(4)对BP神经网络进行遗传运算,获取最佳酶解工艺条件(4) Perform genetic calculation on BP neural network to obtain the best enzymatic hydrolysis process conditions

根据图2所示,以建立的神经网络模型为遗传算法的适应度函数,以各酶解参数水平范围为约束条件,以酶解液肽比例最大值和感官评分最高值为优化目标建立优化模型,确定可行解的编码方法及解码方法,同时设定生成初始种群、交叉概率和变异概率等相关参数。As shown in Figure 2, the established neural network model is used as the fitness function of the genetic algorithm, the level range of each enzymatic hydrolysis parameter is used as the constraint condition, and the maximum value of the enzymatic hydrolyzed peptide ratio and the highest sensory score are the optimization goals to establish an optimization model , to determine the encoding method and decoding method of the feasible solution, and set related parameters such as initial population generation, crossover probability and mutation probability.

生成初始种群、交叉概率和变异概率等3个参数分别设定为24、0.3和0.1,由24个个体组成的一个初始种群开始,分别以牡蛎酶解液肽比例最大值和感官评分最高值作为输出信号,通过选择、交叉、变异等运算,产生出新一代的群体,经过反复运算搜索最佳酶解工艺条件。The three parameters of initial population generation, crossover probability and mutation probability were set to 24, 0.3 and 0.1, respectively. An initial population consisting of 24 individuals was started, and the maximum ratio of oyster enzymatic hydrolyzed peptide and the highest sensory score were used as the parameters respectively. The output signal, through operations such as selection, crossover, and mutation, generates a new generation of groups, and searches for the best enzymatic hydrolysis process conditions through repeated operations.

以牡蛎酶解液肽比例最大值作为输出信号,经过以上遗传运算后,获得的最佳酶解工艺条件为:料水比1∶2.8、蛋白酶添加量1.03%、酶反应温度58.6℃,酶反应时间5.4小时,牡蛎酶解液的肽比例预测值为80.81%。按照上述工艺对牡蛎肉进行验证试验,实验结果显示,牡蛎酶解液的肽比例达78.35%,感官评分值达6.58,相对误差均保持在±5%以内,预测值和实际值没有明显差异。另外,与正交实验获得的最佳工艺相比,以神经网络获得的最佳工艺获得的牡蛎酶解液,其肽比例和感官评分均明显优于正交实验(肽比例和感官评分值分别为75.34%和5.51)。Taking the maximum peptide ratio of oyster enzymatic hydrolyzate as the output signal, after the above genetic calculations, the optimal enzymolysis process conditions were: material-water ratio 1:2.8, protease addition 1.03%, enzyme reaction temperature 58.6°C, enzyme reaction After 5.4 hours, the predicted value of peptide ratio in oyster hydrolyzate was 80.81%. According to the above-mentioned process, the verification test of oyster meat was carried out. The experimental results showed that the peptide ratio of oyster enzymatic hydrolyzate reached 78.35%, and the sensory score value reached 6.58. The relative error was kept within ±5%, and there was no significant difference between the predicted value and the actual value. In addition, compared with the optimal process obtained by the orthogonal experiment, the peptide ratio and sensory score of the oyster hydrolyzate obtained by the optimal process obtained by the neural network were significantly better than those obtained by the orthogonal experiment (peptide ratio and sensory score values were respectively 75.34% and 5.51).

以感官评分最高值作为输出信号,经过以上遗传运算后,获得的最佳酶解工艺条件为:料水比1∶2.1、蛋白酶添加量0.95%、酶反应温度53.8℃,酶反应时间6.0小时,牡蛎酶解液的感观评分预测值为6.67分。按照上述工艺对牡蛎肉进行验证试验,实验结果显示,牡蛎酶解液的感官评分值达6.39,预测值和实际值没有明显差异。Taking the highest value of sensory score as the output signal, after the above genetic calculations, the optimal enzymatic hydrolysis process conditions obtained are: material-water ratio 1:2.1, protease addition 0.95%, enzyme reaction temperature 53.8°C, enzyme reaction time 6.0 hours, The predicted sensory score of oyster hydrolyzate was 6.67 points. According to the above process, the verification test of oyster meat was carried out. The experimental results showed that the sensory score value of the oyster enzymatic hydrolyzate reached 6.39, and there was no significant difference between the predicted value and the actual value.

Claims (10)

1. the oyster based on nerve network system optimization is gustin Controlled-enzymatic Hydrolysis technology, it is characterized in that comprising the steps:
(1) with the oyster meat is raw material, is the enzymolysis process conditional parameter with material than 4 factors such as water, protease addition, enzyme reaction temperature and enzyme reaction times, be the gustin evaluation index with peptide ratio and sensory evaluation scores as the oyster enzymolysis liquid, adopt the orthogonal experiment and the one-to-one relationship of enzymolysis experiment acquisition enzymolysis process parameter and evaluation index at random;
(2) to expect than 4 technological parameters such as water, protease addition, enzyme reaction temperature and enzyme reaction times as input value, peptide ratio and sense organ with the oyster enzymolysis liquid are marked as output signal, set up the BP neural network structure model training sample that oyster is the gustin Controlled-enzymatic Hydrolysis;
(3), finally realize the good mapping relations of enzymolysis process parameter to the enzymolysis index to the training of BP neural network structure model, emulation;
(4) on the basis of BP neutral net, with the function of neutral net gained fitness function, be that target adopts genetic algorithm further to try to achieve the best of breed of enzymolysis process parameter with enzymolysis liquid peptide ratio maximum and sensory evaluation scores peak respectively as genetic algorithm.
2. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 1 based on nerve network system optimization, it is characterized in that: described oyster enzymolysis process is meant the homogenate of oyster meat, adjust material-water ratio, add protease, enzyme reaction under certain condition, boiling water bath heating 10 minutes, cooling, under 4000 rev/mins of conditions centrifugal 20 minutes, get the enzymolysis supernatant and get the oyster enzymolysis liquid; Wherein 4 factors such as material-water ratio, protease addition, enzyme reaction temperature and time are the variable parameters of oyster enzymolysis process.
3. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 2 based on nerve network system optimization, it is characterized in that: described homogenate is meant that oyster meat is through after the running water cleaning and draining surface moisture, carry out historrhexis, homogenate, obtain oyster meat homogenate stoste;
Described adjustment material-water ratio is meant oyster meat homogenate stoste and the umber ratio of adding water;
Described protease addition is meant the weight percent of per 100 gram oyster meat homogenate stostes interpolation protease;
The described enzyme reaction time is meant that oyster meat is after preliminary treatment such as homogenate, adjustment material-water ratio, from adding the time that protease begins to carry out enzyme reaction constantly.
4. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 1 based on nerve network system optimization, it is characterized in that: the peptide ratio of described oyster enzymolysis liquid is meant to be measured respectively peptide nitrogen and total nitrogen content in the oyster enzymolysis liquid, and the percentage that peptide nitrogen content accounts for total nitrogen content is the peptide ratio.
5. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster based on nerve network system optimization of claim 1, it is characterized in that: the sensory evaluation scores of described oyster enzymolysis liquid is meant by the subjective appreciation method carries out the score value that sensory evaluation obtains to the oyster enzymolysis liquid.
6. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 5 based on nerve network system optimization, it is characterized in that: described subjective appreciation is meant with the pairing freshness of the monosodium glutamate solution of 0.2~1.6g/L variable concentrations as standard, monosodium glutamate solution concentration is that the freshness of 1.6g/L is peak 8 minutes, monosodium glutamate solution concentration is that the freshness of 0.2g/L is minimum 1 minute, and the corresponding score value scope of other each concentration is 1~8; Subjective appreciation person all passes through basic flavour training, and at least by forming more than 15 people; The oyster enzymolysis liquid carries out sensory evaluation scores by above method, and the initial data of scoring adopts the processing of testing of Dixon method, finally averages.
7. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 1 based on nerve network system optimization, it is characterized in that: the structural model of described step (2) BP neutral net makes up and is meant application MATLAB7.8 version software, when setting up neural network model with material-water ratio, the protease addition, 4 factors such as enzyme reaction time and enzyme reaction temperature are input quantity, be output layer with peptide ratio or sensory evaluation scores value respectively, make up three layers of BP neutral net of the single output of four inputs, and be tangent sigmoid function tansig with the transfer function between input layer and the hidden layer, the transfer function between hidden layer and the output layer is linear function purelin.
8. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 1 based on nerve network system optimization, it is characterized in that: the training of described step (3) structural model is meant that adopting the trainlm function is the network training function, setting the training objective error is 0.0001, carry out simulation training, make the continuous learning process of BP neutral net by forward-propagating and backpropagation, revise the weights that each neuron connects one by one, the continuous iteration of this process, realize that at last error signal reaches within the scope of permission at 0.0001 o'clock, just stops training.
9. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster of claim 1 based on nerve network system optimization, it is characterized in that: described step (4) genetic algorithm is meant that the neural network model to set up is the fitness function of genetic algorithm, with each enzymolysis parameter horizontal extent is constraints, with enzymolysis liquid peptide ratio maximum and sensory evaluation scores peak is that optimization aim is set up the optimization model, determine the coding method and the coding/decoding method of feasible solution, set the genetic algorithm relevant parameter simultaneously.
10. be gustin Controlled-enzymatic Hydrolysis technology according to the described oyster based on nerve network system optimization of claim 9, it is characterized in that: described setting genetic algorithm relevant parameter is meant sets parameters such as generating initial population, crossover probability and variation probability; Generate initial population, promptly the quantity of contained number in the colony generally gets 20~160, and it crosses the young pathbreaker influences the hunting zone, thereby can not get optimal solution, and excessive then search time is long, and efficient is low; Crossover probability and variation probability, get between 0~1, both are big more, then the algorithm detectivity is strong more, detect new hyperplane easily more, but individual average fitness fluctuation is bigger, the development ability of opposite more little then algorithm is strong more, it is destroyed to make that more excellent individuality is difficult for, the average fitness balance of individuality;
The parameter that configures and program be placed in the MATLAB software move, an initial population of being made up of a plurality of individualities begins optimal search procedure, and computing such as selection that this colony is carried out, intersection, variation, produce the colony of a new generation, continue the search of multiple spot, through after the tentative calculation repeatedly and under the setting of Reasonable Parameters, finally draw a stable optimum factor combination by genetic algorithm.
CN2010105425955A 2010-11-03 2010-11-03 Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system Pending CN102058012A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2010105425955A CN102058012A (en) 2010-11-03 2010-11-03 Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2010105425955A CN102058012A (en) 2010-11-03 2010-11-03 Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system

Publications (1)

Publication Number Publication Date
CN102058012A true CN102058012A (en) 2011-05-18

Family

ID=43993683

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2010105425955A Pending CN102058012A (en) 2010-11-03 2010-11-03 Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system

Country Status (1)

Country Link
CN (1) CN102058012A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107227329A (en) * 2017-06-06 2017-10-03 威海温喜生物科技有限公司 Oyster Protein beam system for glycoside hydrolase inhibitor method
CN109287687A (en) * 2018-09-29 2019-02-01 广东科学技术职业学院 Intelligent baking device and method based on deep learning
CN109886500A (en) * 2019-03-05 2019-06-14 北京百度网讯科技有限公司 Method and apparatus for determining processing technology information
CN110210680A (en) * 2019-06-11 2019-09-06 北京农业信息技术研究中心 A kind of fish body Noninvasive Measuring Method of Freshness and device based on temperature change
CN112926052A (en) * 2021-02-23 2021-06-08 尚蝉(浙江)科技有限公司 Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm
CN113807030A (en) * 2021-10-29 2021-12-17 重庆科炬企业孵化器有限公司 A user-experience-oriented intelligent production method for soybean milk and its storage device
CN117519054A (en) * 2023-12-11 2024-02-06 广州智业节能科技有限公司 High-efficient cold station control system

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107227329A (en) * 2017-06-06 2017-10-03 威海温喜生物科技有限公司 Oyster Protein beam system for glycoside hydrolase inhibitor method
CN109287687A (en) * 2018-09-29 2019-02-01 广东科学技术职业学院 Intelligent baking device and method based on deep learning
CN109287687B (en) * 2018-09-29 2021-04-13 广东科学技术职业学院 An intelligent baking device and method based on deep learning
CN109886500A (en) * 2019-03-05 2019-06-14 北京百度网讯科技有限公司 Method and apparatus for determining processing technology information
CN110210680A (en) * 2019-06-11 2019-09-06 北京农业信息技术研究中心 A kind of fish body Noninvasive Measuring Method of Freshness and device based on temperature change
CN112926052A (en) * 2021-02-23 2021-06-08 尚蝉(浙江)科技有限公司 Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm
CN112926052B (en) * 2021-02-23 2024-03-29 尚蝉(浙江)科技有限公司 Deep learning model security hole testing and repairing method, device and system based on genetic algorithm
CN113807030A (en) * 2021-10-29 2021-12-17 重庆科炬企业孵化器有限公司 A user-experience-oriented intelligent production method for soybean milk and its storage device
CN117519054A (en) * 2023-12-11 2024-02-06 广州智业节能科技有限公司 High-efficient cold station control system
CN117519054B (en) * 2023-12-11 2024-06-11 广州智业节能科技有限公司 High-efficient cold station control system

Similar Documents

Publication Publication Date Title
CN102058012A (en) Oyster sapidity peptide controllable enzymolysis process based on optimization of nerve network system
CN107909149A (en) A kind of Temperature in Greenhouse Forecasting Methodology based on Genetic BP Neutral Network
Liu et al. Automatic and accurate prediction of key water quality parameters based on SRU deep learning in mariculture
CN101243885B (en) Method for preparing low-salt non-saltpetre dry cured goose
Nordhaus Climate Club Futures: On the Effectiveness of Future Climate Clubs
Lv [Retracted] Martial Arts Competitive Decision‐Making Algorithm Based on Improved BP Neural Network
Zhao et al. Based on vis-NIR combined with ANN for on-line detection of bacterial concentration during kombucha fermentation
Adeyi et al. Adaptive neuro fuzzy inference system modeling of Synsepalum dulcificum L. drying characteristics and sensitivity analysis of the drying factors
He et al. Making wines smarter: Evidence from an interpretable learning paradigm
Zhu et al. Optimization of operating conditions in rice heat blast process for Chinese rice wine production by combinational utilization of neural network and genetic algorithms
CN101706488A (en) Method for establishing infant dairy product sensory evaluation expert system
Li et al. Drying characteristics and neural network models of contact ultrasound strengthened cold air drying on yam
Sun et al. Prediction method of wheat moisture content in the hot air drying process based on backpropagation neural network optimized by genetic algorithms
Qi et al. Parameter optimization of double‐blade normal milk processing and mixing performance based on RSM and BP‐GA
Magdaong et al. Developing an artificial neural network model for predicting the growth of Chlorella sorokiniana in a photobioreactor
El-Shafeiy et al. Estimation of coconut maturity based on fuzzy neural network and sperm whale optimization
CN111899805A (en) A method for obtaining the fermentation process model of rice wine based on the improved ant-lion optimization algorithm
CN101706489A (en) Method for establishing raw milk product sensory evaluation expert system
CN102018166B (en) Production process of noodles
Chen et al. Prediction of Nitrite Content of Pickled Okra with Lactobacillus plantarum Using Sensors and by Numerical Simulation Based on Artificial Intelligence.
Ma et al. Simulation of crop evaportranspiration based on BP neural network model and grey relational analysis
CN109588230A (en) A kind of easy-to-use Chinese cabbage method for increasing and test method
Wei et al. Research on Pepper Disease Diagnosis Model Based on Genetic Neural network
Jiang et al. Construction of Elman-Ann Model Based on Virtual Sample Generation and its Application in Fermentation Parameters Optimization of Low-Salt Shrimp Paste
AKLILU Artificial Neural Networks (ANNs) and Response Surface Methodology (RSM) Approach for Modeling and Optimization of Pectin Extraction from Banana Peel

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20110518