Research On Model Migration And Its Application On Coating Weight Prediction Of Galvanized Steel Sheet

In the industrial production, the original process model built on one particular process becomes invalid when working conditions change. Model reconstruction is not only needs human, material consumption, financial resources, but also takes a long time, which will lead to prediction and optimization control of the production process difficult to conduct during this period. Model migration is an effective way to solve this problem. Galvanizing is a kind of important methods for anti-corrosion, meanwhile, control coating weight of galvanized steel sheet precisely based on market demands is significance for improving the quality of products and reducing production costs, researching on the issue of galvanizing process modeling and model tuning has theoretical and practical significance.This paper first introduces the technological process of galvanizing and then establishes a mechanism model for galvanizing process by going into galvanizing process and studying kinetics of reaction> mechanics and so on. Make use of mechanism model to simulate practical galvanizing process and produce production data which is used to modeling and simulation study for model tuning. Study on galvanizing according to the important process parameters such as silicon, carbon and aluminum content change which have significant impact on actual galvanizing process. For the content change of silicon, carbon and aluminum has characteristic of similarity, propose the idea of modeling and model tuning for practical galvanizing process base on model migration technology. Then the feasibility of model migration, the types of process similarity and migration strategy based on several typical kinds of process similarity are briefly introduced. On that basis, for the change of a single process parameter, the family similarity is defined and the appropriate strategy is designed to develop static and dynamic predicting model of galvanizing process, simulation results verify the effectiveness of the method. For the changes of multiple process parameters, the inclusive similarity in galvanizing process is defined, and develops new models based on integrated training algorithm. Simulation results verify the effectiveness of the method. Finally, aim at model migration that discussed above, online application of model migration is studied. On the base of defining condition attributes, propose online modeling and model tuning of multiple model on account of case. Simulation results verify the effectiveness of the method.