Optimization of Hot Mill Production Parameters Based on RF-GA
The hot mill is one of the most important pieces of equipment in fiberboard production;The comprehensive energy consumption of hot mills accounts for a significant portion of the overall energy consumption of fiberboard pro-duction,and optimizing the production parameters to reduce the energy consumption of the hot mill has consistently been a requirement in fiberboard production.Based on the data collected from a fiberboard enterprise in North China,for several production parameters that affect specific energy consumption(SEC)and fiber quality(QF),the SEC and QF prediction models were established with BP neural network,support vector machine(SVM)and random forest(RF)optimized by particle swarm algorithm,and the best SEC and QF prediction models were deter-mined by comparison;then the production parameter optimization model was constructed based on the best prediction model,and the best production parameters were obtained by solving it with genetic algorithm(GA),so as to propose the optimization method of production parameters of the hot mill based on RF-GA model.The results show that there is a strong coupling relationship between each production parameter of the hot mill,and it is nonlinear with SEC and QF;the PSO algorithm is necessary for the optimization of hyperparameters in the machine learning algorithm,which can effectively reduce the error of the machine learning algorithm;five sets of data with different fiber quality were selected and optimized using the established hot mill production parameter optimization method,and the optimized specific energy consumption was reduced by 6.9kW·h/t and 5.22%on average.This study verifies the feasibility of the RF-GA production parameter optimization method for hot mills,which can be applied to optimize production pa-rameters,effectively achieve the goal of energy saving and consumption reduction,and provide a basis for the setting of production parameters of hot mills.
hot millenergy savingoptimizationproduction parametersPSORF-GA
NETL
NSTL
万方数据
