Research On Rolling Force Prediction And Load Distribution Optimization Of Hot Rolling Process Based On Swarm Intelligent Optimization Algorithm

As a typical tandem industrial production process,the hot strip rolling process is nonlinear,strongly correlated,deeply coupled and non-stationary.Accurate prediction of rolling forces in actual production plays an important role in improving product quality and fully utilizing the rolling capacity of the mill;while the accuracy of the rolling force prediction model has a strong correlation with the effective allocation of the stand load.In this paper,an offline simulation study is conducted on a hot strip rolling line with machine learning and Swarm intelligence optimization algorithm as tools to predict rolling forces and optimize load distribution in a real production line based on field data.The main research elements of this dissertation are as follows:(1)Based on the original particle swarm algorithm,the learning factor of the basic particle swarm algorithm is improved with the number of iterations by drawing on and referring to the relevant literature,and three improved particle swarm algorithms are proposed and tested.The iteration speed and convergence accuracy of the adaptive acceleration particle swarm algorithm(TACPSO)and the exponential time-varying acceleration particle swarm algorithm(IPSO)are significantly improved compared with the original particle swarm algorithm;(2)Relying on actual production data in the field,two improved particle swarm algorithms were established to optimize the rolling force forecasting models of least squares support vector machines based on the analysis and screening of data,TACPSO-LSSVM model and IPSO-LSSVM model.The principle is to optimize the penalty parameters of least squares support vector machines and the kernel parameters of radial basis kernel functions.Combined with statistical analysis theory,the prediction models were evaluated by considering two common analysis indexes,root mean square error and mean absolute error,and the effectiveness of the optimized models was verified;(3)Several improvement strategies are proposed in combination with the robust and adaptable sparrow search algorithm(SSA),considering the positive cosine algorithm for updating individual sparrow joiners,the perturbation strategy of fireflies for iterating sparrow populations,and three improved sparrow search algorithms using the improved Circle chaotic mapping,and a multi-strategy sparrow search algorithm(CSFSSA).To test the four algorithms using singlepeak and multi-peak test functions,the CSFSSA has significantly improved its performance compared with the other algorithms;(4)In order to verify the practical application of CSFSSA algorithm,the established three objective functions of load balance,good plate shape and energy consumption reduction are optimized by CSFSSA based on the actual production process of hot strip rolling,aiming to produce a load allocation scheme that takes into account the above objectives.The results show that the optimized load allocation scheme is more fully considered than the original one and provides multiple alternatives for decision makers.