Research On Strip Floating Height Serial Prediction Model Based On Neural Network And Meta-heuristic Optimization Algorithm In Air Cushion Furnace

With the fast development of the economy,the high quality mental strips are extensively used in automobile industry,national defense industry and electric power industry.Different from traditional heat treatment technology,the strip is suspended in the air to complete continuous heat treatment in air cushion furnace.This special work mode greatly improves the efficiency of heat treatment and guarantees the surface quality of the production.However,the floating height is a key process parameter which greatly affects the surface and heat treatment of the strip,and is difficult to obtain in a complex production environment.Therefore,accurately predicting the floating height of the strip is of great significance for ensuring the quality of strip processing.The main research content is carried out from the nozzle structure of air cushion furnace,the influence of floating process parameters and the construction method of floating height prediction model.The main content of this article can be summarized as follows:A new type of mixing nozzle that combines a double-slit nozzle and a round-hole nozzle,which can provide higher lift power.In order to accurately predict the strip floating height under the new mixing nozzle in this paper,firstly,a novel serial hybrid model is constructed,which consists of a floating height mechanism model and two SGO-RBF neural network data models.Then,by abstracting the lift force received by the strip as together provided by the circular hole nozzle and the slit nozzle.Therefore,this paper derives and obtains a mechanism model including the jet impingement angle and other floating process parameters.Finally,in order to solve the problem that the jet impingement angle depends on artificial setting,a SGO-RBF neural network model is proposed to predict the jet impingement angle.Compare to the mixing nozzle,the double-slit nozzle is the most common nozzle type.According the impact of jet impact angle on the strip floating height in the air cushion furnace based on the dual-slit nozzle,a mechanism model is proposed based on the wall jet theory and the gravity balance equation,which can more comprehensively describe relationship between the floating height and the many parameters in the strip floating process.Due to the existence of fluid field and coupling field,the jet impingement angle cannot be measured directly.Consequently,a low discrepancy heuristic evolution ELM(LDHEELM)is proposed as the data driven model to predict the jet impinging angle.In the LDHEELM,a novel double-mutation collaborative differential evolution algorithm(NDMCDE)is proposed,which can optimize model parameters while ensuring low variance of model parameters.In order to accurately predict the strip floating height,a serial hybrid soft-sensing model is constructed by serially connecting the LDHEELM data model to the floating height mechanism model.The robustness of the soft-sensing model is improved by introducing of the mechanism model and the flexibility and applicability of the soft-sensing model is also increased by using the data driven model.In this paper,the models proposed are all tested on the air cushion platform.The experimental results show that the RMSE and MAE of prediction model based on SGO-RBF and the mechanism model are 6.1400 and 5.7795,and the RMSE and MAE of SHSSM model based on low discrepancy heuristic evolution ELM and ground effect theory are 4.4561 ? 4.2169.