Study On The Mathematical Model Of Hydrocyclones For Oil-water Separation Based On Artificial Neural Network

The hydrocyclone for oil-water separation is a new kind of centrifugal separator in petroleum industry. The separation process in hydrocyclones is quite complex. The relation between influence factors and target parameters is a typical multidimensional nonlinear system. It is hard to describe it by simple linear mathematical tools. The artificial neural network (ANN) is a kind of nonlinear dynamic system, and it is very preponderant in simulating the multidimensional nonlinear system and unknown model system. In this paper, a mathematical model had been built up by using BP artificial neural network (BPANN) to predict the separation performance based on the physical property parameters, structure parameters and operation parameters of hydrocyclones. Based on the practical operation situation of hydrocyclones, an optimal three-layer BPANN which was 18-30-2 had been designed after synthetically consideration of training speed, training errors and predicting errors of the network. Because of the limitation of BPANN, PRP conjugate gradient method and genetic algorithm had been used to optimize the training of the ANN, which made the training speed and prediction precision improve markedly. The model had been used to predict the separation performance of hydrocyclones. The model had been used to predict the separation performance of hydrocyclones. The result showed that it is feasible to simulate the separation process of hydrocyclones by BPANN. By means of the BPANN model, the influence factors of separation efficiency and grade efficiency line had been reviewed, the result tallied with the practical experiment very well. That meant the model was quite useful. The results in this paper are very helpful to spreading the application of oil-water separation hydrocyclones both in theory and in practice. The software was implemented by using Visual Basic 6.0 and Microsoft Access in this paper.