基于SSA-BP神经网络优化姜粉辛料制备工艺  

Optimization of ginger powder preparation process based on SSA-BP neural network

作　　者：王雷[1] 张煜 陆宏伟 胡书旭 肖波 WANG Lei;ZHANG Yu;LU Hongwei;HU Shuxu;XIAO Bo(Agricultural Machinery Equipment Research and Design Institute,Hubei University of Technology,Wuhan 430068;Guangdong Institute of Modern Agricultural Equipment,Guangzhou 510630)

机构地区：[1]湖北工业大学农机工程研究设计院,武汉430068 [2]广东省现代农业装备研究所,广州510630

出　　处：《中国食品添加剂》2024年第9期73-82,共10页China Food Additives

基　　金：广东省重点领域研发计划项目(2018B020241003);广东省省级乡村振兴战略专项项目(2022KJ101);广东省科技创新战略资金项目(粤科资字[2021]133号)。

摘　　要：为优化姜粉的热风干燥制备工艺,研究不同干燥温度、切片厚度、干燥时间对制得姜粉感官风味的影响。以感官评分为响应值,先进行Box-Behnken响应面法设计再通过SSA-BP神经网络算法对响应面法进行验证。SSA-BP算法优化后所得的最优参数为:干燥温度62.978℃,切片厚度2.734 mm,干燥时间5.773 h。SSA-BP神经网络预测的感官评分的均方根误差RMSE为0.033457,小于BP神经网络的均方根误差0.054574;决定系数R2为0.99094,大于BP神经网络的决定系数R2=0.97064,证明优化模型的预测性能较BP神经网络更高。响应面法寻得的最优姜粉制备条件下样品的感官值为37.705分,而SSA-BP神经网络模型的预测分值为37.280分,二者误差仅1.1%,模型拟合度较高,SSA-BP神经网络模型可以优化姜粉辛料制备工艺。本研究可为姜粉制备提供科学参考。In order to optimize the hot air-drying process of ginger powder preparation,the effects of different drying temperatures,slice thicknesses,and drying times on the sensory flavor of the resulting ginger powder were investigated.The Box-Behnken response surface method was designed and then validated by SSA-BP neural network algorithm with sensory score as the response value.The optimal parameters after optimization were as follows:drying temperature 62.978℃,slice thickness 2.734 mm,drying time 5.773 h.The root mean square error(RMSE)of sensory score predicted by the SSA-BP neural network was 0.033457,which was smaller than RMSE of 0.054574 for the BP neural network.The coefficient of determination R²was 0.99094,which was greater than R²of 0.97064 for the BP neural network.This indicated that the optimization model had a higher prediction performance than that of the BP neural network.The sensory score of the product under the optimal preparation conditions found by the RSM was 37.705 points,while the predicted score of SSA-BP neural network model was 37.280.The error between the two was only 1.1%,indicating a high degree of model fit.The SSA-BP neural network model is suitable for optimizing the ginger powder preparation process.The study provides a scientific reference for ginger powder preparation.

关 键 词：姜粉 BP神经网络 麻雀搜索算法 响应面法 感官评价 

分 类 号：TS202.3[轻工技术与工程—食品科学]