遗传算法结合反向传播神经网络优化牡丹花粉硬糖制备工艺  

Optimization of preparation technology of peony pollen hard candy by genetic algorithm combined with back propagation neural network

作　　者：彭腾腾 范彬[2] 尹盼盼 李海燕 王新娣[2] 石晓峰 PENG Tengteng;FAN Bin;YIN Panpan;LI Haiyan;WANG Xindi;SHI Xiaofeng(School of Pharmacy,Gansu University of Traditional Chinese Medicine,Lanzhou 730000;Gansu Academy of Medical Sciences,Lanzhou 730000)

机构地区：[1]甘肃中医药大学药学院,兰州730000 [2]甘肃省医学科学研究院,兰州730000

出　　处：《中国食品添加剂》2023年第6期235-244,共10页China Food Additives

基　　金：兰州市人才创新项目(2021-RC-112);兰州市科技发展指导性计划项目(2019-ZD-132)。

摘　　要：目的:利用反向传播神经网络(back propagation neural network,BP-NN)结合遗传算法(genetic algorithm,GA)优化牡丹花粉硬糖制备工艺,为牡丹花粉硬糖品质的改善提供最优的工艺参数。方法:以综合评分为指标,通过单因素试验遴选影响硬糖品质的主要因素,采用正交试验考察糖醇配比、牡丹花粉用量、熬糖温度对牡丹花粉硬糖品质的影响,在此基础上,利用反向传播神经网络建立预测模型,再结合遗传算法寻优得出各因素的最佳复配结果。结果:糖醇配比、牡丹花粉用量和熬糖温度对硬糖品质具有显著性影响(P<0.05)且彼此交互作用明显;反向传播神经网络模型的训练、测试和预测集的相关系数均大于0.95,表明模型准确度高、拟合程度好;经遗传算法寻优,得出牡丹花粉硬糖的最佳制备工艺为糖醇配比4∶5(w/w),牡丹花粉用量0.94%,柠檬酸用量0.6%,熬糖温度175℃,调和温度90℃。结论:反向传播神经网络结合遗传算法优化牡丹花粉硬糖制备工艺合理可行,为牡丹花粉相关产品的制备提供了新思路。Objective:To optimize the preparation process of peony pollen hard candy by genetic algorithm(GA)combined with back propagation neural network(BP-NN),so as to provide optimal process parameters for the improvement of peony pollen hard candy quality.Method:Taking the comprehensive score as the index,the main factors affecting the quality of hard candy were selected by single-factor experiment,and the influence of sugar alcohol ratio,peony pollen dosage and boiling sugar temperature on the quality of peony pollen was investigated by orthogonal experiment,and on this basis,a prediction model was established by using BP-NN,and then combined with genetic algorithm to find the optimal compounding results of each factor.Results:The sugar alcohol ratio,peony pollen dosage and boiling sugar temperature had a significant effect on the quality of hard candy(P<0.05),and interacted with each other significantly.The correlation coefficients of the training,testing and prediction sets of the BP-NN model were all greater than 0.95,indicating that the model had high accuracy and good fit.After genetic algorithm optimization,it was concluded that the optimal preparation process of peony pollen hard candy was 4∶5(w/w)of sugar alcohol ratio,0.94%of peony pollen,0.6%of citric acid,175℃of boiling sugar,and 90℃of blending temperature.Conclusion:The BP-NN combined with GA to optimize the preparation process of peony pollen hard candy was reasonable and feasible.The study provides a new idea for the preparation of peony pollenrelated products.

关 键 词：牡丹花粉 硬糖 遗传算法 反向传播神经网络 工艺优化 

分 类 号：TS201.1[轻工技术与工程—食品科学]