油水分离水力旋流器的人工神经网络模拟

Artificial neural network simulation of hydrocyclones for oil-water separation

采用改进的BP神经网络模型模拟水力旋流器的油水分离过程。根据水力旋流器的实际生产条件 ,确定旋流器模型设计中的优化神经网络结构为 8- 12 - 12型神经网络。对神经网络的结构进行了优选和分析 ,提出了自适应调节学习率和动量因子的快速BP算法 ,并对激励函数进行了改进。结果表明 ,用改进的BP神经网络模型模拟水力旋流器的分离过程是切实可行的 ,通过旋流器的人工神经网络模拟 ,能够根据旋流器的物性参数和分离期望值 ,预测旋流器的结构参数和操作参数 ,实现旋流器结构参数和操作参数的优化。

The process of oil-water separation in hydrocyclones was simulated by improved BP neural network (BPNN). According to the actual production condition, the optimized neural network structure is determined as 8-12-12 network in the model design of the hydrocyclone. A kind of quick BP algorithm of self-adaptive adjusting learn rate and moment factor is put forward. And the stimulating function of BPNN was also improved during the designation of BPNN. It shows that the simulation of the process of oil-water separation in hydrocyclones by means of improved BPNN is feasible. By the simulation of BPNN, the geometric parameters and the operational parameters of the gydrocyclone can be predicted according to the different physical property and separation requirement, therefore the optimization of the geometric parameters and the operational parameters can be realized.