基于遗传算法进化的人工神经网络(GA-ANN)对葡萄糖发酵生产普鲁兰多糖的条件优化  

Optimization of glucose fermentation conditions for pullulan polysaccharide production by artificial neural network based on genetic algorithm

作　　者：陈世伟 罗嘉伟 王舸楠 赵廷彬 殷海松[3] 郑志强 郑国保[5] 乔长晟[1,2,6,7] CHEN Shiwei;LUO Jiawei;WANG Genan;ZHAO Tingbin;YIN Haisong;ZHENG Zhiqiang;ZHENG Guobao;QIAO Changsheng(College of Bioengineering,Tianjin University of Science and Technology,Tianjin 300457,China;Tianjin Huizhi Baichuan Bioengineering Co.Ltd.,Tianjin 300457,China;College of Bioengineering,Tianjin Modern Vocational Technology College,Tianjin 300457,China;Institute of Systems Engineering,Academy of Military Sciences,Beijing 101300,China;Agricultural Biotechnology Research Center,Ningxia Academy of Agriculture and Forestry Sciences,Yinchuan 750002,China;Key Laboratory of Industrial Fermentation Microbiology,Ministry of Education and Tianjin Key Laboratory of Industrial Microbiology,Tianjin 300457,China;Tianjin Engineering Center of Microbial Metabolism and Fermentation Process Control,Tianjin 300457,China)

机构地区：[1]天津科技大学生物工程学院,天津300457 [2]天津慧智百川生物工程有限公司,天津300457 [3]天津现代职业技术学院生物工程学院,天津300457 [4]军事科学院系统工程研究所军需工程技术研究所,北京101300 [5]宁夏农林科学院农业生物技术研究中心,宁夏银川750002 [6]工业发酵微生物教育部重点实验室暨天津市工业微生物重点实验室,天津300457 [7]天津市微生物代谢与发酵过程控制技术工程中心,天津300457

出　　处：《食品与发酵工业》2023年第8期60-66,共7页Food and Fermentation Industries

基　　金：天津市科委基金项目(21YDTPJC00140);宁夏回族自治区重点研发计划(2022BEG02006);工业发酵微生物教育部重点实验室暨天津市工业微生物重点实验室开放课题项目(2020KF006)。

摘　　要：基于遗传算法进化的人工神经网络,以葡萄糖为原料,对出芽短梗霉产普鲁兰多糖的发酵培养条件进行优化。首先通过单因素试验和Plackett-Burman实验筛选显著因素,再进行Box-Behnken实验建立数据样本,最后利用Matlab建立神经网络模型寻找最优解。结果表明,葡萄糖和酵母抽提物对普鲁兰多糖的合成具有显著的正效应,K_(2)HPO_(4)对普鲁兰多糖的合成具有显著的负效应。遗传算法-人工神经网络的决定系数与相对误差分别为0.9988与1.72%。最终优化获得普鲁兰多糖发酵的最佳培养基组分为葡萄糖150 g/L,酵母抽提物7.1 g/L,MgSO_(4)·7H_(2)O 1.4 g/L,K 2HPO 47 g/L,NaCl7 g/L,自然pH。在此条件下,普鲁兰多糖的产量为83.25 g/L,较优化前提高了79.73%。经济分析表示优化后的培养基成本较优化前降低了约70%。该研究结果为普鲁兰多糖的工业化生产提供了数据支撑,有助于提升普鲁兰多糖在行业中的竞争力。Using glucose as raw material,the fermentation conditions of pullulan polysaccharide produced by Aureobasidium pullulans were optimized by artificial neural network based on genetic algorithm(GA-ANN).Firstly,single factor test and Plackett-Burman(PB)experiment were used to screen significant factors,then Box-Behnken(BBD)experiment was conducted to establish data samples,and finally Matlab was used to establish model optimization to find the optimal solution.Results showed that glucose and yeast extracts had a significant positive effect on the synthesis of pullulan polysaccharide,while K_(2)HPO_(4) had a significant negative effect on the synthesis of pullulan polysaccharide.The coefficient and relative error of GA-ANN were 0.9988 and 1.67%respectively.Finally,the optimal fermentation conditions of pullulan polysaccharide were 150 g/L of glucose,7.1 g/L of yeast extract,1.4 g/L of MgSO_(4)7H_(2)O,7 g/L of K 2HPO 4,7 g/L of NaCl,and natural pH.Under these conditions,the yield of pullulan polysaccharide was 83.25 g/L,79.73%higher than that before optimization.The cost of optimized medium was reduced by about 70%compared with that before optimization.The results provide data support for the industrial production of pullulan polysaccharide and help to improve the competitiveness of pullulan polysaccharide in the industry.

关 键 词：普鲁兰多糖 遗传算法 人工神经网络 非线性关系 模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TQ929.2[自动化与计算机技术—控制科学与工程]