Research On Parallel Computing Method For Dynamic Optimization Of Reheating Furnace Temperature With Operating Mode’s Switch

As an important link in the steel production process,the reheating furnace is mainly used to reheat the solidified slab from the continuous casting process or the slab from the holding pit,so that the temperature of the slab meets the subsequent rolling requirements of the hot rolling line.However,the reheating furnace control system is a non-linear,strongly coupled,multivariable and multi-constrained complex control system,and it is difficult to accurately measure the heating temperature distribution of the slab in the furnace.One or more parameters such as the type of billet,the model size,the initial temperature of entering the furnace,and the target temperature of exiting the furnace will change.When the production conditions are switched,in order to ensure the current technological production requirements of the slab,it is necessary to timely and effectively adjust the furnace temperature settings of the different heating sections.However,the adjustment of the furnace temperature setting value at the actual production site is often based on manual experience.Although manual experience can ensure real-time control,it is difficult to ensure the overall dynamic optimization and adjustment of the entire reheating furnace control system.Considering the influence of the change of working conditions on the optimization of furnace temperature,how to establish the model of the dynamic optimization control system of the reheating furnace temperature is the focus of this article.The specific research progress made in this thesis is as follows:(1)A linear interpolation method is used to construct a furnace temperature distribution model for different heating sections.Then,based on the mechanism model of heat exchange in the furnace,comprehensively considering factors such as water-cooling black mark and physical parameter changes,a more accurate three-dimensional billet temperature prediction model is constructed under the premise of ensuring combustion efficiency and quality.(2)In order to ensure the accuracy of the calculation results of the billet temperature prediction model,a parameter identification model for the total heat exchange factor is constructed based on the "black box" measurement data,and the parameters can still be accurately identified even when the noise data is included.Here,an adaptive weighted least squares model(WLS)is introduced.WLS can adaptively reduce the influence of noise on the identification results,and avoid the "residual pollution" and "residual flooding" that may occur in the traditional fixed bandwidth kernel density estimation method.Finally,a model combining conjugate gradient and gradient projection is used to solve the improved inverse thermal problem,which ensures the stability of the process and improves the identification accuracy.(3)In order to improve computing efficiency,a solution strategy based on GPU parallel computing is proposed.Considering the actual production conditions of the billet heating process,using the finite difference method to discretize the three-dimensional nonlinear billet temperature prediction model,and finally comparing the calculation results of the temperature field with the CPU calculation results to verify the efficiency and accuracy of the GPU parallel calculation.(4)Based on the optimal control problem of the reheating furnace temperature in the steady-state production state,an objective function is constructed to minimize the difference between the surface and core temperature of the billet and the material loss during the heating process.Different model optimization strategies are used to verify the experimental effect,and finally the ideal temperature rise curve of the billet under this specification is obtained as the set value of the dynamically optimized billet temperature information.In the process of solving the temperature field,the model solving strategy based on GPU parallel computing greatly improves the optimization efficiency and meets the efficiency requirements of actual industrial production.Finally,on the basis of the steady-state optimization results,based on the basic idea of the model predictive control algorithm,the structure is constructed to minimize the difference between the calculated temperature value and the desired temperature,minimize the difference between the upper and lower surface temperatures and the center temperature,minimize the energy consumption of the reheating furnace and minimize the difference between the temperature rise curve and the ideal temperature rise curve during the billet heating process.When the billet type,billet thickness,initial temperature and other conditions change,the current furnace temperature setting value is re-optimized,and finally compared with the empirical method results of the actual industrial production process,it is verified that the simulation optimization results can be better adapted to changes in working conditions.