基于BP神经网络的烟熏香肠色泽预测研究

Prediction of Smoked Sausage Color Based on BP Neural Network

以传统烟熏方式加工的香肠为研究对象,利用反向传播(Back-Propagation,BP)神经网络建立烟熏香肠色泽的预测模型。通过试验获得不同烟熏温度、烟熏时间和肥瘦比条件的烟熏香肠,测定其L*、a*、b*和△E值,并对BP神经网络算法、隐含层神经元个数、学习速率和动量系数进行优化,获得最佳的BP神经网络预测模型结构。基于Levenberg-Marquardt算法建立精确的L*、b*和△E预测模型,性能测试显示L*、b*和△E预测模型的相关系数(R2)分别为0.847、0.825和0.924。相应的均方根误差(root mean square error,RMSE)分别为4.609、3.564和5.012。基于拟牛顿BFGS算法建立精确的a*值预测模型,性能测试显示模型的R2和RMSE分别为0.905和2.237。

Processed in a conventional manner smoked sausages as the research object, back-propagation（BP） neural network prediction model is used to predict the color of smoked sausage. Used the smoked sausage with different smoked temperature, smoked time and fineness ratio, the L＊, a＊, b＊and AE value were determined, and the BP neural network algorithm, hidden layer neuron number, learning rate and momentum coefficient were optimized, and the best BP neural network prediction model structure. Based on Levenberg-Marquardt algorithm, the accurate L＊, b＊ and △E prediction model are established. The performance test shows that the correlation coefficient （R2） of L＊, b＇and AE prediction model are 0.847, 0.825 and 0.924, respectively. The corresponding root mean square error （RMSE） are 4.609, 3.564 and 5.012, respectively. Based on the Quasi-Newton BFGS algorithm, an accurate a＊ prediction model is established, the performance test shows that the R2 and RMSE of the model are 0.905 and 2.237, respectively.