基于BP神经网络和遗传算法的库尔勒香梨挥发性物质萃取条件的优化

doi:10.3864/j.issn.0578-1752.2018.23.012

摘要/Abstract

摘要：

【目的】通过BP神经网络结合遗传算法对固相微萃取条件（萃取头种类、样品量、萃取温度、萃取时间和盐离子添加量）的优化,建立高效的库尔勒香梨汁香气物质的萃取分析方法,为库尔勒香梨相关产品天然风味的拟合及调控奠定基础。【方法】通过单因素试验确定最佳萃取纤维头种类和样品量、萃取温度、萃取时间和盐离子添加量的最佳萃取参数。在单因素试验设计的基础上,以库尔勒香梨汁中关键香气物质（己醛、丁酸乙酯、壬醛、乙酸乙酯、(E)-2-己烯醛、己醇、(Z)-2己烯醛、丙酸乙酯和己酸乙酯）的含量为评价指标进行中央复合设计试验进一步考察样品量、萃取温度、萃取时间和盐离子添加量对库尔勒香梨汁固相微萃取效果的影响。最后以中央复合设计的试验结果为初始种群,以样品量、萃取温度、萃取时间和盐离子添加量为输入值,以库尔勒香梨汁中关键香气物质含量为其函数的输出值,通过BP神经网络模型来调试函数适应度。使用中央复合设计数据以外的数据集对BP神经网络外推能力进行验证。利用遗传算法（算法参数：最大进化代数100、初始种群数20、变异概率0.2、交叉概率0.4）在试验水平范围内预测全局最优的萃取条件。【结果】统计分析结果表明,65 μm PDMS/DVB萃取头对库尔勒香梨汁中关键香气物质的萃取效果最好,样品量、萃取温度、萃取时间和盐离子添加量对固相微萃取效果具有显著影响（P<0.05）。BP神经网络的拓扑结构为‘4-15-1’。验证数据和训练数据的MSE均大于0.017;训练、测试和验证数据的相关系数分别为0.990、0.951和0.973,表明BP神经网络预测模型有很好的准确性,可以用于库尔勒香梨汁香气物质固相微萃取结果的预测。BP神经网络外推能力验证试验的结果与BP神经网络模型预测值的拟合度（R ²）为0.992,表明试验建立的BP神经网络模型有很好的外推能力,能够准确地预测不在训练集内的数据集的输出值。遗传算法迭代100代后确定的最优个体为3.41 μg·g ^-1,最佳固相微萃取条件是样品量5.33 g、萃取温度44.70℃、萃取时间25.22 min和盐离子添加量0.63 g。在尽可能接近此条件下进行验证试验（试验条件：样品量5.33 g、萃取温度45℃、萃取时间25 min和盐离子添加量0.63 g）,测得库尔勒香梨汁中9种关键香气物质的含量为（3.37±0.23）μg·g ^-1,与最优个体预测值相比,误差仅为-1.19%。【结论】上述结果表明BP神经网络模型结合遗传算法是一种准确度较好的优化固相微萃取参数的新方法,同时也为解决非线性模型工艺优化的问题提供了新思路。库尔勒香梨汁香气物质固相微萃取的最佳萃取条件为：样品量5.33 g,萃取温度45℃,萃取时间 25 min,盐离子添加量0.63 g。

关键词: 库尔勒香梨, 挥发性物质, 固相微萃取（SPME）, BP神经网络, 遗传算法（GA）

Abstract:

【Objective】 In order to lay the foundation for fitting and regulation of natural flavors of Korla fragrant pear-related products, the extraction and analytical method for Korla pear aroma compounds was established by the optimization of solid phase microextraction (SPME) conditions (extraction fiber type, sample amount, extraction temperature, extraction time, and salt addition amount) with the back propagation neural network coupled genetic algorithm (BPNN-GA). 【Method】The type of extraction fiber and the best extraction parameters of sample amount, extraction temperature, extraction time, and salt addition amount were determined by single factor test. On the basis of the single factor test, the central composite design was applied to further investigate the effects of sample amount, extraction temperature, extraction time, and salt addition amount on the SPME of Korla pear juice by using the content of the key aroma substances in Korla pear juice (1-Hexanal, butanoic acid ethyl ester, 1-nonanal, hexanoic acid ethyl ester, acetic acid ethyl ester, (E)-2-hexenal, 1-hexanol, (Z)-2-hexenal, and propanoic acid ethyl ester) as evaluation indicators. The model of BPNN was used to debug fitness of the function based on the results of central composite design, and then the sample amount, extraction temperature, extraction time, salt addition amount and the content of the key aroma substances in Korla pear juice were used as initial population, input values, and output value, respectively. The generalization ability of BPNN was validated using the data set outside the central composite design. The global optimal extraction conditions within the experimental level were predicted using the genetic algorithm (algorithm parameter setting: maximum evolution algebra 100, population size 20, mutation probability 0.2, and crossover probability 0.4).【Result】The result of statistical analysis showed that the extraction fibers had a significant (P<0.05) effect on the extraction efficiency of SPME, and the 65 μm PDMS/DVB fiber was the best extraction fiber of SPME for volatile compounds of Korla pear juice. The sample amount, extraction temperature, extraction time, and salt addition had significant (P<0.05) effect on the SPME efficiency. The topology of the BPNN was ‘4-15-1’. The Root Mean Square Error (MSE) of the verification data and the training data were both greater than 0.017, and the correlation coefficients of training, test and verification data were 0.990, 0.951 and 0.973, respectively, indicating that the BPNN prediction model had good accuracy and could be used for the prediction of the result of SPME of the aroma substances in Korla pear juice. The fitting degree of the prediction value of BPNN and the verification experiment result validating the generalization ability of the BPNN was 0.992, which indicated the established BPNN model had good generalization ability and could accurately predict the output value of datasets outside the training datasets. Using genetic algorithm after evaluation of data for 100 generations determined the best individual (3.41 μg?g ^-1) and optimum condition: sample amount 5.33 g, extraction temperature 44.7℃, extraction time 25.22 min and salt addition 0.63 g. The verification test were conducted under these conditions as closely as possible (sample amount 5.33 g, extraction temperature 45℃, extraction time 25 min and salt addition 0.63 g), and the content of 9 key volatile substances in Korla pear juice was (3.37±0.23) μg?g ^-1, which compared with the predicted value, and the error was -1.19%.【Conclusion】 The above results indicated that BPNN-GA was a new method with better accuracy to optimize the solid-phase microextraction parameters, and the method of BPNN-GA provided a new way to solve the nonlinear problem. And the optimum extraction conditions for SPME of Koral pear juice were as follows: sample amount 5.33 g, extraction temperature 45℃, extraction time 25 min and salt addition 0.63 g.

Key words: Korla pear, volatile compounds, solid phase microextraction (SPME), back propagation neural network, genetic algorithm (GA)

张芳,未志胜,王鹏,李凯旋,詹萍,田洪磊. 基于BP神经网络和遗传算法的库尔勒香梨挥发性物质萃取条件的优化[J]. 中国农业科学, 2018, 51(23): 4535-4547.

ZHANG Fang,WEI ZhiSheng,WANG Peng,LI KaiXuan,ZHAN Ping,TIAN HongLei. Using Neural Network Coupled Genetic Algorithm to Optimize the SPME Conditions of Volatile Compounds in Korla Pear[J]. Scientia Agricultura Sinica, 2018, 51(23): 4535-4547.

0
/ / 推荐

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: https://www.chinaagrisci.com/CN/10.3864/j.issn.0578-1752.2018.23.012

https://www.chinaagrisci.com/CN/Y2018/V51/I23/4535

图/表 11

表1

图1

图2

图3

图4

图5

图6

图7

表2

图8

图9

参考文献 42

[1]	ZHAO L, WANG Y T, HU X T, SUN Z J, LIAO X J . Korla pear juice treated by ultrafiltration followed by high pressure processing or high temperature short time. LWT-Food Science and Technology, 2016,65:283-289. doi: 10.1016/j.lwt.2015.08.011
[2]	TIAN H L, ZHAN P, DENG Z Y, YAN H Y, ZHU X R . Development of a flavour fingerprint by GC-MS and GC-O combined with chemometric methods for the quality control of Korla pear (Pyrus serotina Reld). International Journal of Food Science & Technology, 2015,49(12):2546-2552.
[3]	陈计峦, 江英, 吴继红, 胡小松 . 固相微萃取GC-MS技术在梨香气成分分析中的应用研究. 食品与发酵工业, 2007,33(3):107-110. doi: 10.3321/j.issn:0253-990X.2007.03.028
	CHEN J L, JIANG Y, WU J H, HU X S . Research on the aromatic compounds of pear fruit by GC-MS with solid phase microextraction. Food and Fermentation Industries,, 2007,33(3):107-110. (in Chinese) doi: 10.3321/j.issn:0253-990X.2007.03.028
[4]	BALASUBRAMANIAN S . Solid-phase microextraction (SPME) techniques for quality characterization of food products: A review. Food & Bioprocess Technology, 2011,4(1):1-26. doi: 10.1007/s11947-009-0299-3
[5]	YU A N, SUN B G, TIAN D T, QU W Y . Analysis of volatile compounds in traditional smoke-cured bacon(CSCB) with different fiber coatings using SPME. Food Chemistry, 2008,110(1):233-238. doi: 10.1016/j.foodchem.2008.01.040 pmid: 26050188
[6]	ZHANG Z, PAWLISZYN J . Headspace solid-phase microextraction. Analytical Chemistry, 1993,65(14):1843-1852. doi: 10.1021/ac00062a008
[7]	皋香, 施瑞城, 谷风林, 张彦军, 吕岱竹 . 固相微萃取结合气相色谱-质谱测定海南番木瓜香气成分. 食品工业科技, 2013,34(14):148-151.
	GAO X, SHI R C, GU F L, ZHANG Y J, LV D Z . Volatile composition of Hainan solo papaya by headspace solid-phase microextraction/gas chromatography-mass spectrometry. Food Industry Technology, 2013,34(14):148-151. (in Chinese)
[8]	韩素芳, 钟冬莲, 丁明 . 固相微萃取-气相色谱-质谱测定杨梅香气. 分析试验室, 2009,28(s1):99-101.
	HAN S F, ZHONG D L, DING M . Determination of aroma of Yangmei by Solid Phase Microextraction-Gas Chromatography-Mass Spectrometry. Analytical Laboratory, 2009,28(s1):99-101. (in Chinese)
[9]	王永倩, 刘卫光, 高杰, 钟昔阳 . 油条风味物质的顶空固相微萃取条件优化及成分分析. 农产品加工月刊, 2017(4):30-36. doi: 10.16693/j.cnki.1671-9646(X).2017.04.037
	WANG Y Q, LIU W G, Gao J, ZHONG X Y . Optimization of HS- SPME to study the odorant extracts and analysis of flavor composition from Youtiao.Agricultural Product Processing Monthly, 2017(4):30-36. (in Chinese) doi: 10.16693/j.cnki.1671-9646(X).2017.04.037
[10]	LIANG R J . Orthogonal test design for optimization of the extraction of polysaccharides from Phascolosoma esulenta and evaluation of its immunity activity. Carbohydrate Polymers, 2008,73(4):558. doi: 10.1016/j.carbpol.2007.12.026 pmid: 26048221
[11]	BURIN V M, MARCHAND S, DE R G, BORDIGNONLUIZ M T . Development and validation of method for heterocyclic compounds in wine: optimization of HS-SPME conditions applying a response surface methodology. Talanta, 2013,117(22):87-93. doi: 10.1016/j.talanta.2013.08.037 pmid: 24209315
[12]	DESAI K M, SURVASE S A, SAUDAGAR P S, LELE S S . Comparison of artificial neural network (ANN) and response surface methodology (RSM) in fermentation media optimization: Case study of fermentative production of scleroglucan. Biochemical Engineering Journal, 2008,41(3):266-273. doi: 10.1016/j.bej.2008.05.009
[13]	MARAN J P, PRIYA B . Comparison of response surface methodology and artificial neural network approach towards efficient ultrasound- assisted biodiesel production from muskmelon oil. Ultrasonics Sonochemistry, 2015,23(123):192-200. doi: 10.1016/j.ultsonch.2014.10.019 pmid: 25457517
[14]	SORIA M A, GONZALEZ FUNES J L G, GARCIA A F . A simulation study comparing the impact of experimental error on the performance of experimental designs and artificial neural networks used for process screening. Journal of Industrial Microbiology & Biotechnology, 2004,31(10):469-474. doi: 10.1007/s10295-004-0171-4 pmid: 15480940
[15]	MOGHADDAM M G, KHAJEH M . Comparison of response surface methodology and artificial neural network in predicting the microwave-assisted extraction procedure to determine zinc in fish muscles. Food & Nutrition Sciences, 2011,2(8):803-808.
[16]	SHAN H, FEI Y, HUAN Y, FENG G, FEI Q . Quantitative near-infrared spectroscopic analysis of trimethoprim by artificial neural networks combined with modified genetic algorithm. Chemical Research in Chinese Universities, 2014,30(4):582-586. doi: 10.1007/s40242-014-3410-x
[17]	LAN Z, ZHAO C, GUO W Q, GUAN X, ZHANG X L . Optimization of culture medium for maximal production of spinosad using an artificial neural network-genetic algorithm modeling. Journal of Molecular Microbiology & Biotechnology, 2015,25(4):253.
[18]	BAS D, ISMAIL H . BOYACI. Modeling and optimization II: Comparison of estimation capabilities of response surface methodology with artificial neural networks in a biochemical reaction. Journal of Food Engineering, 2007,78(3):846-854. doi: 10.1016/j.jfoodeng.2005.11.025
[19]	AGATONOVIC-KUSTRIN S, ZECEVIC M, ZIVANOVIC L, TUCKER I G . Application of artificial neural networks in HPLC method development. Journal of Pharmaceutical & Biomedical Analysis, 1998,17(1):69-76. doi: 10.1016/S0731-7085(97)00170-2 pmid: 9608428
[20]	LOU W, NAKAI S . Application of artificial neural networks for predicting the thermal inactivation of bacteria: A combined effect of temperature, pH and water activity. Food Research International, 2001,34(7):573-579. doi: 10.1016/S0963-9969(01)00074-6
[21]	ENTESHARI M, MODAMMADI A, NAYEBZADEH K, AZADNIYA E . Optimization of headspace single-drop microextraction coupled with gas chromatography-mass spectrometry for determining volatile oxidation compounds in mayonnaise by response surface methodology. Food Analytical Methods, 2014,7(2):438-448. doi: 10.1007/s12161-013-9643-3
[22]	PAKRAVAN P, AKHBARI A, MORADI H, AZANDARYANI A H, MANSOURI A M, SAFARI M . Process modeling and evaluation of petroleum refinery wastewater treatment through response surface methodology and artificial neural network in a photocatalytic reactor using poly ethyleneimine (PEI)/titania (TiO₂) multilayer film on quartz tube. Applied Petrochemical Research, 2015,5(1):47-59. doi: 10.1007/s13203-014-0077-7
[23]	周文杰, 王鹏, 詹萍, 田洪磊 . 香气活度值法结合PLS1用于梨酒特征香气物质筛选与鉴定. 食品科学, 2017,38(14):138-143. doi: 10.7506/spkx1002-6630-201714021
	ZHOU W J, WANG P, ZHAN P, TIAN H L . Aroma activity value method combined with PLS1 for the screening and identification of flavor characteristics of pear wine. Food Science, 2017,38(14):138-143. (in Chinese) doi: 10.7506/spkx1002-6630-201714021
[24]	何小燕, 田洪磊, 詹萍, 王晓玲, 杨富彬, 雷银凤, 宁孔卵, 马友梅 . 鸡肉香精最适风味前体PLSR分析与研究. 现代食品科技, 2016(10):268-275.
	HE X Y, TIAN H L, ZHAN P, WANG X L, YANG F B, LEI Y F, NING K L, MA Y M . Analysis and study of the desirable precursor of the chicken flavor based on PLSR.Modern Food Science and Technology, 2016(10):268-275. (in Chinese)
[25]	李媛媛, 常庆瑞, 刘秀英, 严林, 罗丹, 王烁 . 基于高光谱和BP神经网络的玉米叶片SPAD值遥感估算. 农业工程学报, 2016,32(16):135-142. doi: 10.11975/j.issn.1002-6819.2016.16.019
	LI Y Y, CHANG Q R, LIU X Y, YAN L, LUO D, WANG S . Estimation of maize leaf SPAD value based on hyperspectrum and BP neural network. Journal of Agricultural Engineering, 2016,32(16):135-142. (in Chinese) doi: 10.11975/j.issn.1002-6819.2016.16.019
[26]	WHITLEY D . An executable model of a simple genetic algorithm. Foundations of Genetic Algorithms, 1993,2:45-62. doi: 10.1016/B978-0-08-094832-4.50009-X
[27]	雷英杰 . MATLAB遗传算法工具箱及应用. 西安: 西安电子科技大学出版社, 2005.
	LEI Y J. MATLAB Genetic Algorithm Toolbox and Application. Xi. an: Xi’an University of Electronic Science and Technology Press, 2005. ( in Chinese)
[28]	LIN L, ZHUANG M, LEI F, YANG B, ZHAO M . GC/MS analysis of volatiles obtained by headspace solid-phase microextraction and simultaneous-distillation extraction from Rabdosia serr.(MAXIM.) HARA leaf and stem. Food Chemistry, 2013,136(2):555-562. doi: 10.1016/j.foodchem.2012.08.048 pmid: 23122097
[29]	PLUTOWSKA B, CHMIEL T, DYMERSKI T, WARDENCKI W . A headspace solid-phase microextraction method development and its application in the determination of volatiles in honeys by gas chromatography. Food Chemistry, 2011,126(3):1288-1298. doi: 10.1016/j.foodchem.2010.11.079
[30]	TANKIEWICZ M, MORRISON C, BIZIUK M . Application and optimization of headspace solid-phase microextraction (HS-SPME) coupled with gas chromatography-flame-ionization detector (GC-FID) to determine products of the petroleum industry in aqueous samples. Microchemical Journal, 2013,108(1):117-123. doi: 10.1016/j.microc.2012.10.010
[31]	HU G F, ZHU Y, HEENANDEZ M, KOUTCHMA T, SHAO S Q . An efficient method for the simultaneous determination of furan, 2-methylfuran and 2-pentylfuran in fruit juices by headspace solid phase microextraction and gas chromatography-flame ionisation detector. Food Chemistry, 2016,192(192):9-14. doi: 10.1016/j.foodchem.2015.06.100 pmid: 26304314
[32]	于静, 李景明, 吴继红, 葛毅强 . 顶空固相微萃取(SPME)在红葡萄酒香气成分测定中的应用研究. 中外葡萄与葡萄酒, 2006(3):4-9. doi: 10.3969/j.issn.1004-7360.2006.03.001
	YU J, LI J M, WU J H, GE Y Q . Rapid determination of aroma components in Cabernet sauvignon wine by HS--SPME and GC-MS. Chinese and Foreign Grapes and Wine, 2006(3):4-9. (in Chinese) doi: 10.3969/j.issn.1004-7360.2006.03.001
[33]	CHAKRABORTY R, SAHU H . Intensification of biodiesel production from waste goat tallow using infrared radiation: process evaluation through response surface methodology and artificial neural network. Applied Energy, 2014,114(2):827-836. doi: 10.1016/j.apenergy.2013.04.025
[34]	张纪兴, 陈燕忠 . 人工神经网络建模结合遗传算法优化岗松油环糊精包合物制备工艺参数. 中国药科大学学报, 2011,42(4):324-328.
	ZHANG J X, CHEN Y Z . Optimization of the process parameters of Baeckeae oil-β-cyclodextrin inclusion complex was conducted by artificial neural network and genetic algorithm. Journal of China Pharmaceutical University, 2011,42(4):324-328. (in Chinese)
[35]	VAS G., VEKEY K . Solid-phase microextraction: a powerful sample preparation tool prior to mass spectrometric analysis. Journal of Mass Spectrometry, 2004,39(3):233-254. doi: 10.1002/(ISSN)1096-9888
[36]	WARDENCKI W, MICHULEC M, CURYLO J . A review of theoretical and practical aspects of solid-phase microextraction in food analysis. International Journal of Food Science & Technology, 2004,39(7):703-717. doi: 10.1111/j.1365-2621.2004.00839.x
[37]	CHEN M X, CHEN X S, WANG X G, CI Z J, HE T M, ZHANG L J . Comparison of headspace solid-phase microextraction with simultaneous steam distillation extraction for the analysis of the volatile constituents in Chinese apricot. Agricultural Sciences in China, 2006,5(11):879-884. doi: 10.1016/S1671-2927(06)60139-9
[38]	周志, 徐永霞, 胡昊, 程超, 潘思铁 . 顶空固相微萃取和同时蒸馏萃取应用于GC-MS分析野生刺梨汁挥发性成分的比较研究. 食品科学, 2011,32(16):279-282.
	ZHOU Z, XU Y X, CHENG C, PAN S T . Comparative application of head space-solid phase micro-extraction and simultaneous distillation extraction for GC-MS analysis of volatile components in Chestnut Rose (Rosa roxburghii Tratt) juice. Food Science, 2011,32(16):279-282. (in Chinese)
[39]	LI G P, LIN L J, ZHAO J, LI J H . The optimization and HS-SPME/ GC-MS analysis in pear fruit. Advanced Materials Research, 2014, 1033- 1034:726-766. doi: 10.4028/www.scientific.net/AMR.1033-1034.762
[40]	薛宏坤, 刘成海, 刘钗, 徐浩, 秦庆雨, 沈柳杨, 郑先哲 . 响应面法和遗传算法-神经网络模型优化微波萃取蓝莓中花青素工艺. 食品科学, 2018,39(16):280-288.
	XUE H K, LIU C H, LIU C, XU H, QIN Q Y, SHEN L Y, ZHENG X Z . A comparative study of the optimization of microwave extraction of anthocyanins from blueberry by response surface methodology and genetic algorithm-artificial neural network. Food Science, 2018,39(16):280-288. (in Chinese)
[41]	王慧竹, 陈盟杰, 石琳, 朱文静, 杨鑫 . BP神经网络结合响应面优化罗布麻总黄酮的提取工艺. 中国酿造, 2017,36(6):153-157.
	WANG H Z, CHEN M J, SHI L, ZHU W J, YANG X . Optimization of extraction process of total flavonoids from Apocynum venetum by BP neural network and response surface methodology. Chinese Brewing, 2017,36(6):153-157. (in Chinese)
[42]	周梦舟, 吴珊, 柳念, 汪超, 徐宁, 胡勇, 史文博, 李冬生 . 基于神经网络对枯草芽孢杆菌富硒过程的建模研究. 中国食品学报, 2016,16(12):66-74. doi: 10.16429/j.1009-7848.2016.12.010
	ZHOU M Z, WU S, LIU N, WANG C, XU N, HU Y, SHI W B, LI D S . Modelling research of the selenium enrichment by Bacillus subtilis base on artificial neural network. Journal of Chinese Institute of Food Science and Technology, 2016,16(12):66-74. (in Chinese) doi: 10.16429/j.1009-7848.2016.12.010

序列 Sequence	样品量 Sample amount (g)	萃取温度 Extraction temperature (℃)	萃取时间 Extraction time (min)	盐添加量 Salt addition (g)	实验值 Observed value (μg?g^-1)	预测值 Predictive value (μg?g^-1)	相对误差 Relative error (%)
1	7	50	30	0.8	1.52	1.54	-1.60
2	4	45	15	0.5	2.88	2.81	2.43
3	6	40	15	0.7	2.24	2.27	-1.34
4	5	45	25	0.9	3.10	3.00	3.23
5	6	35	20	1.0	2.05	2.11	-2.93
6	5	50	25	0.5	1.96	2.02	-3.06
7	3	35	25	0.8	1.64	1.59	1.24