基于人工神经网络的油-水乳状液粘度智能预测研究

Study on Oil-water Emulsion Viscosity Prediction by Using Artificial Neural Network

油水混合粘度受温度、剪切速率、含水率3种因素协同影响，很难用常规方程准确计算。提出了采用人工神经网络进行油水乳状液粘度预测的新方法。建立了三层结构BP神经网络模型，输入层有3个神经元，分别代表温度、剪切速率、含水率，输出层有一个神经元，代表油水混合物粘度，隐层神经元数目为30个。在实验室配置一定比例的油水乳状液，通过流变性测量获得学习训练样本。试验温度范围24．3～46．1℃，剪切速率范围10～400s^-1，含水率范围10％～60％。结果表明，神经网络预测值与实测值符合良好。最大误差7．37％。

The determination of viscosity of oil-water emulsion by using temperature,shear rate,water cut and their relationships were difficult to be expressed with traditional mathematic equations.A new method based on artificial neural network was proposed for viscosity prediction.The BP network model with three layers was established.The input layer had three neurons representing temperature,water cut and shear rate respectively.The output layer had one neurons,representing viscosity.The hidden layer had 30 neurons experiments were conducted to obtain the learning and testing samples from rheologic test.The experimental temperature was ranged from 24.3 to 46.1℃,shear rate was ranged from 10% to 60% and water cut was ranged from 10% to 60%.The results show that the Ann prediction value agrees well with the experimental data and the maximum error is 7.37%.