Research On Soft Sensor Methods For Overflow Particle Size Estimation Of Hydrocyclone

Hydrocyclone is an efficient equipment which is using centrifugal to separate different particle size mixture, it has been widely used in mineral processing. The overflow particle size of hydrocyclone is an important parameter in evaluating the operating performance in practical production. Determinate the overflow particle size distribution of the hydrocyclone, return it to the control circuit, ensure the optimal composition of the partical size, which can improve the metal recoveries, reduce the energy consumption and increase productivity.Because of the complexity of the hydrocyclone’s internal flow field, influence factor and the relationship between the parameters is a typical multidimensional nonlinear system, it is difficult to use simple linear mathematical tools to describe, so it is difficult to put forward a systematic and complete, simple, accurate and has a general formula of certain depth. So to establish an accurate model has a certain difficulty.Based on the analysis of grinding-classification processes, the mathematical models and operating principle of hydrocyclone, we select some variables as auxiliary variable which the soft sensor model needs. And this paper established the soft-sensing model based on BP neural network to predict overflow particle size distribution of the hydrocyclone. To solve the problem of easily enter into the local minimal point and slow convergent speed in the BP algorithm, the global search ability swarm intelligent optimization method—the particle swarm optimization algorithm (PSO) was introduced to optimize the weights and thresholds of the BP neural network, based on this foundation I adopt BP algorithm for modeling so as to improve the training effect. To the precocious phenomena of the particle swarm optimization (PSO) algorithm, this paper adopts a nonlinear adjustment inertia weight strategy to improve the PSO algorithm, the performance of algorithm testing shows that the nonlinear adjustment inertia weight strategy PSO algorithm converges quickly and search precision has been improved. In order to test the nonlinear adjustment inertia weight strategy PSO-BP algorithm, the problem of M-G chaotic time series prediction is simulated in this paper, the simulated experimental results show that the improved algorithm, in a certain extent, improved network generalization ability.Finally, the established soft sensor model is simulated, simulation results indicate that the BP neural network model can predict overflow particle size distribution of the hydrocyclone better.