为探究南极磷虾酶解过程的影响参数,以水解度为考查指标,实施多因素作用下的酶解实验,建立了基于BP神经网络的南极磷虾酶解工艺模型。使用78个样本对该神经网络模型进行83次迭代后,得到了准确度最优的拟合模型(MSE达到最小值0.002 242,样本相关系数达到最大值0.956 9);使用9个样本对模型进行测试发现,9组数据的MSE=0.003 889,R=0.985 6,表明该工艺模型可以准确地描述和预测不同工艺参数下南极磷虾酶解反应的结果;使用神经网络模型求解水解度的极大值和最优工艺条件,在酶添加量为4.73%,pH 6.99,温度54.0 ℃,时间201.0 min时,水解度最大,该条件下实验实测值为41.20%,与预测值41.36%无显著差异,该模型工艺优化结果准确。与响应面相比,BP神经网络具有避免舍去高次交互项所引起的误差,拟合模型纠偏性强,拟合结果更加精准等方面的优势。
To study the influence of process parameters on the enzymatic hydrolysis of Antarctic krill, enzymatic hydrolysis experiments under multiple factors were carried out and a process model based on BP neural network was built. In the model, after 83 epochs with 78 training samples, the mean square error (MSE) of model reached a minimum value of 0.002 242, and the correlation coefficient of the model samples reached a maximum value of 0.956 9, which confirmed that the accuracy of the model was optimal. MSE of 0.003 889 and R of 0.985 6 was obtained for nine tested data sets, which indicated that the model could accurately describe and predict the results of enzymatic hydrolysis of Antarctic krill under different process parameters. Finally, the optimal parameters were found by solving the maximum degree of hydrolysis in the model, i.e., enzyme addition amount 4.73%, pH 6.99, enzyme hydrolysis time 201.0 min, and enzyme hydrolysis temperature 54.0 ℃. With the above parameters, the degree of hydrolysis determined by experiment was 41.20%. which was close to the predicted value of 41.36%.
[1] 刘志东,陈雪忠,黄洪亮.南极磷虾粉的营养成分分析及评价[J]. 中国海洋药物,2012,31(2):43-48.
[2] 全沁果,段伟文,曾雪鸽.南极磷虾粉成分分析及营养学评价[J]. 食品与机械,2018,34(9):68-72;76.
[3] 袁玥,李学英,杨宪时.南极磷虾粉营养成分的分析与比较[J]. 海洋渔业,2012,34(4):57-63.
[8] 楼乔明,王玉明,杨文鸽,等.南极磷虾粉脂质及脂肪酸组成分析[J]. 水产学报,2012,36(8):56-62.
[5] FÄRBER-LORDA J,BEIER E,MAYZAUD P. Morphological and biochemical differentiation in Antarctic krill[J]. Journal of Marine Systems,2009,78(4):525-535.
[6] WEN J,ZHAO J L,LUO S W,et.al.The improvements of BP neural network learning algorithm[C].International Conference on Signal Processing:IEEE,2000.
[7] 白宝光,朱洪磊,范清秀.基于遗传优化BP神经网络的乳制品质量安全风险预警[J/OL].食品科学:1-12[2020-04-13]. http://kns.cnki.net/kcms/detail/11.2206.TS.20200323.0935.002.html.
[8] 周桢,张胜军.基于BP神经网络模型的食品安全供给分析[J].价值工程,2019,38(35):72-73.
[9] 靳淑祎.基于BP神经网络算法的食品烘干机温度控制[J].食品工业,2019,40(9):229-231.
[10] 郭敏强.基于BP人工神经网络建立巴沙鱼片干燥动力学模型[C].中国食品科学技术学会第十五届年会论文摘要集.北京:中国食品科学技术学会,2018:736.
[11] WANG M,ZHANG J,ZHANG Z,et al. Simultaneous ultraviolet spectrophotometric determination of sodium benzoate and potassium sorbate by BP-neural network algorithm and partial least squares[J]. Optik,2020,201:163-166.
[12] JIANG X,XUE H,ZHANG L,et al. Nondestructive detection of chilled mutton freshness based on multi-label information fusion and adaptive BP neural network[J]. Computers and Electronics in Agriculture,2018,155:371-377.
[13] SUYKENS J A K. Artificial Neural Networks for Modelling and Control of Non-Linear Systems[M]. Boston:Spring-Verlag New York Inc,2010.
[14] ROUAT J,PICHEVAR R,LOISELLE S. Perceptive,Non-linear Speech Processing and Spiking Neural Networks[M].Nonlinear Speech Modeling and Applications,2005.
[15] 张华丹,张玲云,张国玉,等.响应面法优化南极磷虾蛋白酶解工艺条件[J].食品工业,2019,40(7):94-98.
[16] 满洋,陈舜胜,曲映红.南极磷虾副产物酶解蛋白胨工艺优化及实际应用[J].食品工业科技,2019,40(21):198-202;224.
[17] 王林林.酶解法提取南极磷虾虾油的研究[D].上海:上海海洋大学,2019.
[18] 薛勇,赵明明,王超,等.响应面法优化南极磷虾蛋白自溶工艺的研究[J].食品工业科技,2012,33 (4):346-348;373.
[19] 中华人民共和国卫生部.GB 5009.235—2016 食品中氨基酸态氮的测定[S].北京:中国标准出版社,2016.
[20] 中华人民共和国卫生部.GB 5009.5—2016 食品中蛋白质的测定[S].北京:中国标准出版社,2016.
[21] 刘璐,曹文红,李佳蔚,等.南极磷虾酶解工艺优化及酶解产物对牡蛎肉的冷冻保护作用[J].食品工业科技,2020,41(1):252-257;265.
[22] 燕梦雅,刘宝林,刘志东,等.响应面法优化南极磷虾蛋白酶解物溶解性工艺[J].食品工业科技,2018,39(5):151-156.
