摘要
根据粗集理论和神经网络的结合,对浓香型白酒的成分做了定量的研究。由于BP神经网络在解决实际问题时,其模型结构的确定、每层神经元个数的选择无现成的规律可遵循,必须由实验数据来确定。为此通过对浓香型白酒的实验数据分析,确定所需要的元素,然后对其建模。经过多次的实验仿真,训练了1个BP网络,并对该网络进行误差分析,并使结果一目了然。
In this paper, the composition of strong-flavor liquor was quantitively evaluated by combining fuzzy set theory and neural network. But the application of BP neural network needs support with experimental data, because several parameters such as model structure and number of neuron were difficult to be determined. Therefore we used the data generated from strong-flavor liquor to carry on our study. By analyzing these data, some compulsory, factors can be determined firstly and then added to the model. After evaluation with trial experiments, 1 BP neutral network was established and it has been estimated by error analysis.
出处
《中国酿造》
CAS
北大核心
2010年第1期94-96,共3页
China Brewing
基金
四川省重点实验室研究课题基金(2008RK008)
关键词
浓香
白酒
BP神经网络
建模
strong flavor
liquor
neural network of BP
modeling
