张彦粉,魏华,葛纪者,邹洋.基于遗传算法优化BP神经网络的可食用油墨粘度的预测[J].包装工程,2021,42(19):49-54.
ZHANG Yan-fen,WEI Hua,GE Ji-zhe,ZOU Yang.Prediction of the Viscosity of Edible Ink Based on BP Neural Network Optimized with Genetic Algorithm[J].Packaging Engineering,2021,42(19):49-54. |
| 基于遗传算法优化BP神经网络的可食用油墨粘度的预测 |
| Prediction of the Viscosity of Edible Ink Based on BP Neural Network Optimized with Genetic Algorithm |
| |
| DOI：10.19554/j.cnki.1001-3563.2021.19.007 |
| 中文关键词: 壳聚糖 粘度 可食用油墨 BP神经网络 遗传算法 |
| 英文关键词: chitosan viscosity edible ink Back Propagation neural network genetic algorithm |
| 基金项目:广东省教育厅普通高校青年创新人才类项目（2019GKQNCX005） |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 通过研究遗传算法优化BP神经网络建立自变量与因变量之间的关系，从而对可食用油墨的粘度进行预测和模拟。方法 在前期关于可食用油墨的研究基础上，以醋酸浓度、壳聚糖用量、酒精用量、研磨速度为自变量，以配制得到的油墨粘度作为因变量，利用正交实验设计实验，运用BP神经网络结合遗传算法对可食用油墨的粘度进行预测和模拟。结果 以正交实验设计得到30组实验数据，利用Matlab 2018a软件中GAOT遗传算法工具箱，经过38次迭代训练，得到收敛精度为10−4的神经网络，粘度的预测值与对应的真实值相对误差介于0.05%~3.7%，拟合度R2值为0.8672，表明该神经网络对可食用油墨的粘度具有较好的预测能力和较高的预测精度。结论 遗传算法优化BP神经网络可以用来预测和模拟可食用油墨的粘度，可以将神经网络拓展到可食用油墨其他性能的评价体系中，从而对可食用油墨的生产和应用提供指导性的建议。 |
| 英文摘要: |
| The work aims to predict and simulate the viscosity of edible ink by studying the genetic algorithm to optimize the Back Propagation (BP) neural network to establish the relationship between the independent variable and the dependent variable. Based on the previous researches on edible inks, the acetic acid concentration, chitosan dosage, alcohol dosage, and grinding speed were used as independent variables, and the obtained viscosity of edible ink as the dependent variable. The experiment was first designed by orthogonal experiments, and then the BP neural network with genetic algorithm was used to predict and simulate the viscosity of edible ink. 30 sets of experimental data were obtained by orthogonal experimental design: by using Genetic Algorithm Optimization Toolbox (GAOT) in Matlab 2018a software, a neural network with a convergence accuracy of 104 was obtained after 38 iterative trainings. The relative error between the predicted value of viscosity and the corresponding true value of viscosity was between 0.05% and 3.7%, and the R2 value of the fit was 0.8672, indicating that the BP neural network had excellent predictive ability and high predictive accuracy to predict the viscosity of edible inks. The BP neural network optimized with genetic algorithm can be used to predict and simulate the viscosity of edible inks, and extend the neural network into the evaluation system of other performances of the edible ink, thereby providing guidance for the production and application of the edible ink. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501716号