Study On The Rolling Schedule Optimization And Model Self-Learning Of 1370 Tandem Cold Rolling Mill

The production of steel rolling is a systems engineering. There are manyworking procedures before the material being finished product and it is neededto organize these procedures in an optimized way. The procedure for steelproduction is not unique, the connection and the match for all the stands is alsoan optimization problem. In the production procedure of the cold rolling, it isan important condition to offer a rational rolling schedule and program theproduce capability of every stand for the high quality, high yield and lowenergy consumption. Presently, the procedure based on the experience is notsatisfying. To offer a more rational schedule, the field experience should beconcluded into theory for more extensive significance.Firstly, based on the analysis of the reconstruct 1370 five-stand coldrolling control system, we propose the main parameter models for its specialtechnology and performance. On considering the uncertainty in rolling forceset-up of the new steel roller, the Neural Network(NN) is used to predict therolling force instead. The rolling schedule can be revised on-line based onmodels' self-learning off-line.Secondly, more rational optimization method is offered based on theanalysis of the characters of the cold rolling. Different object functions andconstrain conditions are proposed for different requirements. Further more,corresponding programs are given for different optimization schemes. Withthese, we optimized the field rolling schedule and show the results of the afterand before the optimization.Lastly, we analyze the variables online disturbance and the influence tothe rolling procedure, then, propose the corresponding control methods. Sincethe disturbance can not be considered in the rolling schedule enactment, weadopt the rolling schedule with adaptive function. We propose the adaptive andself-learn algorithm for the main rolling parameters. The simulation resultsshow the satisfying performance.