本文已被：浏览 2490次 下载 88次
投稿时间：2020-07-17
投稿时间：2020-07-17
中文摘要: 以大西洋鲭鱼为原料，研究不同烘烤条件下，大西洋鲭鱼水分含量和色度值（L*、a*、b*和ΔE）的变化情况，并利用试验所得的数据以大西洋鲭鱼烘烤温度和烘烤时间作为模型输入值，水分含量和色度值（L*、a*、b*和ΔE）同时作为输出值，建立人工神经网络（artificial neural network，ANN）模型，并对模型性能进行测试。结果表明，随着烘烤时间的增加，水分含量和L*值逐渐下降，烘烤温度越高，下降越迅速。而a*和ΔE 与水分含量和L*的趋势正好相反。b*值先升高后下降。通过试验可知，当隐含层神经元个数为14 时，ANN 模型的均方根误差（root mean square error，RMSE）为0.07，R2 全部大于0.98，模型整体拟合程度最高。因此选择2-14-5 作为ANN 模型最佳拓扑结构。
Abstract:These experiments found that baking conditions influenced moisture content and colorimetric values（L*，a*，b* and ΔE）of Atlantic mackerel. Atlantic mackerel baking times and temperatures used in these experiments were provided as model input values，and measured moisture content and colorimetric values（L*，a*，b*and ΔE）were used as output values to establish an artificial neural network（ANN）based model.Model performance was subsequently tested. The results showed that proportional to increases in heating temperature and decreases in moisture content，the value of L*decreases with increasing heating time. By contrast，a*and ΔE displayed an opposite trend to that of moisture content and L*.The value of b*was raised and then lowered.Throughout the experiment，when the number of hidden layer neurons was 14，the root mean square（RMSE）value of the ANN model was 0.07，and R2 was always greater than 0.98 in the model with the highest overall fit.Therefore，2-14-5 was selected as the best topology for the ANN model.
keywords: Atlantic mackerel artificial neural network（ANN） moisture content baking colorimetric values
文章编号：202107003 中图分类号： 文献标志码：
基金项目:国家重点研发计划“蓝色粮仓科技创新”重点专项（2019YFD0902000）
引用文本：
津ICP备20000603号-1