Research On Soft Density Measurement Of Dust Concentration Based On Improved Neural Network

In this paper, a new method of dust concentration detection based on neural network soft sensor is proposed to solve the problems existing in traditional dust detection, which is of great practical significance to the research of dust.Firstly, a new variable step size LMS adaptive algorithm is proposed to modify the filter coefficients and it is improved the convergence speed and the steady performance under low SNR. The experimental and simulation results show that the algorithm can filter out the random noise of electrostatic signal.Secondly, by analyzing the time-frequency of the filtered electrostatic signal, a new method of EEMD decomposition based on the improved similar extreme value is presented. Simulations and contrasts with method of EMD are given. By analyzing the energy and energy entropy of the electrostatic signal, there is a positive correlation between the change of electrostatic signal and the trend of dust concentrationThen, based on the soft measurement method, two kinds of dust concentration soft sensing models of BP and RBF are established. The parameters of the two models are determined based on the combination of experiment and heuristic methods. The experiment and simulation results show that the BP network model is slightly better than that in the training accuracy and generalization ability of RBF network model.Finally, the IGA is used to optimize and improve the BP neural network model. Experiments and simulation results show that the convergence performance, run time and mean square error of the model are improved after optimization, which improves the accuracy and efficiency of the model. It is proved that the soft sensor model based on neural network can monitor the dust concentration.