[目的]食品3D打印技术是食品领域具有发展前景的新技术,但是打印过程影响因素多,存在打印参数确定困难、打印精度预测能力差等问题.寻找有效建模方法,对杏鲍菇粉 3D打印参数进行寻优,以确定最佳 3D打印条件.[方法]本研究采用杏鲍菇粉和刺槐豆胶为 3D打印原料,以单因素试验为基础,通过中心组合试验设计,研究喷嘴直径、打印高度、喷嘴移动速度和填充率4个关键的工艺参数对杏鲍菇粉3D打印精度的影响,并在此基础上采用响应面法和神经网络-遗传算法分别建模分析,确定 3D打印的工艺参数.[结果]单因素试验及中心组合试验结果表明,影响 3D打印精度的主要因素从大到小顺序为填充率、喷嘴直径、喷嘴移动速度、打印高度.响应面法和神经网络-遗传算法均可用于杏鲍菇粉3D打印参数优化,但是优化效果不同.响应面法的决定系数R 2 值、均方根误差、相对误差、预测最优值分别为0.8817、0.2314、72.73%、0.148;神经网络-遗传算法的决定系数R 2 值、均方根误差、相对误差、预测最优值分别为0.9389、0.2269、33.85%、0.215.比较模型参数可得,神经网络-遗传算法的决定系数R 2 值较高,均方根误差、相对误差较低,比响应面法拟合能力更好,同时其预测最优值较高,具有更好的预测能力.神经网络-遗传算法比响应面法更适合于杏鲍菇粉3D打印参数工艺的优化.采用神经网络-遗传算法获得以杏鲍菇为原料的 3D打印最佳工艺参数条件为:喷嘴直径 1.2 mm、打印高度 1.1 mm、喷嘴移动速度24 mm·s-1、填充率84%.经过试验验证,神经网络-遗传算法确定的最优参数打印样品偏差为0.325,优于响应面的实际打印偏差0.550.[结论]本研究结果表明神经网络-遗传算法可以有效确定3D打印过程最优工艺参数,准确预测食品3D打印产品的精度,可作为农产品及食品个性化3D打印工艺参数优化的一种有效便捷方法.
[Objective]Food 3D printing technology,a promising technology in the field of food,can be affected by multiple factors and thus has problems,such as difficulty in determining printing parameters and poor ability of predicting printing accuracy.This paper aimed to seek out an effective modeling method to optimize 3D printing parameters of Pleurotus eryngii powder and to determine the optimal conditions for 3D printing.[Method]Pleurotus eryngii powder and locust bean gum were adopted as 3D printing ink.Then,based on single-factor experiments,the central composite experimental design was performed to study the influence of four key process parameters-nozzle diameter,printing height,nozzle movement speed and fill density-on the accuracy of 3D printing.In order to optimize 3D printing parameters of Pleurotus eryngii powder,response surface methodology(RSM)and artificial neural network and genetic algorithm(ANN-GA)were employed to achieve different effects.[Result]The determination coefficient(R2),root mean square error(RMSE),relative error(RE),and optimal value of prediction(VOP)of RSM model were 0.8817,0.2314,72.73%,and 0.148,respectively;the R2,RMSE,RE,and optimal VOP of ANN-GA model were 0.9389,0.2269,33.85%,and 0.215,respectively.The ANN-GA model obtained higher R2,lower RMSE and RE,and was better fitting ability,and higher optimal VOP than RSM model,so ANN-GA model possessed better prediction ability.Compared with RSM,ANN-GA was more suitable for optimization of 3D printing parameters of Pleurotus eryngii powder.The optimal process parameters of 3D printing obtained by ANN-GA,with Pleurotus eryngii as printing ink,included nozzle diameter 1.2 mm,printing height 1.1 mm,nozzle movement speed 24 mm·s-1,and fill density 84%.Experimental verification suggested that the deviation of printed samples by ANN-GA was 0.325,which was superior to the actual printing deviation 0.550 by RSM.[Conclusion]ANN-GA was effective in determining the optimal process parameters of 3D printing and accurate in predicting the accuracy of food 3D printing products.Therefore,ANN-GA could serve as an effective and convenient method for optimizing personalized 3D printing parameters of agricultural products and food.
苏安祥;贺安琪;马高兴;赵立艳;杨文建;胡秋辉
南京财经大学食品科学与工程学院/江苏省现代粮食流通与安全协同创新中心/江苏省食用菌保鲜与深加工工程研究中心,南京 210023南京农业大学食品科学技术学院,南京 210095
3D食品打印;杏鲍菇;神经网络;遗传算法;工艺优化
3D food printing;Pleurotus eryngii;neural network;genetic algorithm;process optimization
《中国农业科学》 2024 (003)
584-596 / 13
江苏省科技成果转化项目(BA2021062)、江苏高校优势学科建设工程资助项目(PAPD)、江苏高校品牌专业建设工程资助项目(TAPP)
10.3864/j.issn.0578-1752.2024.03.012
评论