Research On The Operation Optimization Of Cold Rolling Production Process Based On Physical And Data Driven Model

Based on scientifically and reasonably setting up the operation parameters for cold rolling mill in accordance with the production technical requirements, the operation optimization for cold rolling production process is expected to minimize the energy consumption for the same cold rolling strip of the same quality assurance, which makes the whole production process more economical and effective.A series of operation optimization problems derived from the real production practice are studied in this dissertation, including the modeling and solutions of steady state operation optimization, robust operation optimization, and real-time online operation optimization. For the steady state operation optimization problems under normal production condition, we built the mathematical programming model based on physical model and then proposed the improved particle swarm algorithm to solve it when the physical model is available, and otherwise the data modeling method based on subspace is studied and the multi-population particle swarm algorithm with negative gradient search is explored when the mechanism or model parameters are hard to identified; for the robust operation optimization problem considering that the parameters such as the rolled piece thickness are unable to accurately measured or that disturbance accompanies with these parameters, we built the robust operation optimization model and then proposed the robust particle swarm algorithm to solve it; or the real-time operation optimization problem under dynamic status condition,we built the real-time operation optimization model and then proposed the improved DE particle swarm algorithm to solve it. The details are given as follows:1) Concerning the steady state operation optimization problem when the physical model is available, the physical modeling is adopted with the physical laws of energy, mass and momentum conservation based on the steel strip rolling principle. On the premise of the steel quality assurance and with the goal of energy saving, the steady state operation optimization model for the cold rolling production process is built based on physical model taking into account the constraint conditions of the rolling force balance and rolling power balance. Furthermore, the improved particle swarm algorithm specific to this model is proposed considering its characteristics. The real data collected from cold rolling production is used to conduct the performance test for the model and algorithm, whose effectiveness are proved by the result.2) Concerning the condition that is too complicated to build the correct physical mathematic model or under which the physical model is unavailable during the production process, the operation optimization modeling and the optimization method based on data-driven are researched. The modeling method and mathematic model based on subspace are proposed upon the history data of cold rolling production, and also the operation optimization model based on data-driven is built considering with the constraint conditions such as rolling force and rolling power for the purpose of quality assurance and energy saving. Through the comparison with operation setup method based on the experience and principle from production practice, the algorithm’s effectiveness is positively recognized.3) Concerning the universal existing problems such as the inaccurate measurement results and disturbance during the production process, the robust operation optimization problem is researched and the robust optimization model is built for the rolling force during the cold rolling production process. When the disturbance happens to the convexity degree, thickness and horizontal rigidity and other parameters of the strip steel, the model is able to ensure the reasonable, scientific and improved set up of the parameters of the rolling force, thus minimizing the energy consumption and ensure the favorable shape. Based on the particle swarm algorithm, the robust particle swarm optimization algorithm is proposed for the model. As a matter of course, the effectiveness of the model and the algorithm are proved by the experiments.4) Concerning the real-time continuous cold rolling process, the real-time online operation optimization problem is researched. The steady state optimization model and model prediction control are combined to build the steady state operation optimization model whose target is energy-saving and performance-improving. In order to provide the expectation curve for prediction control, the improved particle swarm algorithm would work out the best set points of the various parameters in real time on line. Besides, with the minimum deviation between the expectation curve and future control output as the objective, and with the real production constraint condition as the constraint function, the prediction control model is built with rolling optimization futures. For this problem, the improved DE particle swarm algorithm is proposed to provide fast solution for prediction control model with constraint conditions, thus meeting the quality and speed requirements of real-time control.5) With the above-mentioned cold rolling operation optimization models and methods, the operation optimization software platform for the production process is developed to meet the practical production requirements. The functions cover production process monitoring interface, production parameters setup interface, operation optimization interface, model management interface, optimization algorithm interface and data management interface, etc. It works as a friendly system operation interface, providing scientific support for the setting up of the operation parameters such as rolling force, exit thickness and rolling reduction, etc.