Application Research Of Sulfuric Acid Concentration Soft-sensing Technique Based On Neural Network

One principal parameter of zinc smelter off-gas is the concentration ofacid. Mainly by using the imported instruments, the present concentrationdetection of acid still lies at the instrument measuring stage which has manymeasuring errors and a longer time-lag. Therefore a new type of detectionmethod has been proposed in this paper which integrated the artificial neuralnetwork of the Particle Swarm Optimization (PSO) and the Soft-SensingTechnology (SST).The following researching tasks will be included:1、The overseas and domestic research results and significanceconcerning the smelter off-gas has been introduced. Then a general review hasbeen made with respect to the applications of Soft-Sensing Technology (SST)in the field of the modern chemical industry.2、Adopted the SST theory Factors which influenced the concentrationdetection of acid has been summarized after a comprehensive analysis of thesmelter off-gas technology. Furthermore primary and supplementary variableshas been deduced which are beneficial to the establishment of theSoft-Sensing Model (SSM) of the acid concentration.3、The mathematical model of the RBF neural network, the learningalgorithm and Soft-sensing Model (SSM) based on the RBF network havebeen founded on account of the applications of the neural network. Through asimulation comparison with the BP network, the sulfuric acid concentrationsoft-sensing model of the RBF neural network based on the orthogonal leastsquares (OLS) can constantly react according to the changing situation and bewidely generalized. Nevertheless with the further research, we uncover thatthe selection of RBF network parameters have influenced the soft-sensingmodel most.4、Another soft-sensing model of the RBF network acid concentrationdetection based on the particle swarm optimization (PSO) has been presentedso as to acquire optimized parameters more efficiently. With a Multiple inputand single output RBF network system, optimized by the Particle SwarmOptimization (PSO) to its network parameters, the great advantage of thismodel is in simple structure, short time training and high learning efficiency which are appropriate for the concentration detection of acid.Integrated with the neural network algorithm and Soft-SensingTechnology (SST), this paper investigates the smelter off-gas in a certainsmelter. The scientific and efficiency of the model is finally validated by asimulated model building，and at the same time, a new method and means ofsulfuric acid concentration detection provided by Soft-Sensing Technology(SST).