Aluminum Strip Crown Prediction in Hot Rolling Process Based on Data-driven Methods

Abstract

Aluminum process parameters and crown are two important factors that determine the product performance in hot rolling process of aluminum strip. In this paper, we propose a data-driven model fitting method for the relationship between process parameters and crown of aluminum strip hot rolling process. The method includes two parts: the data preprocessing and the relational model fitting. In data preprocessing, we use feature selection algorithm, outlier handling algorithm and missing value padding algorithm to preprocess the given data and obtain high-quality data for analysis. In the relationship model fitting, we use six typical machine learning methods to fit the relationship model between process parameters and crown. Based on the relationship model, we can accurately predict the crown by given process parameters. In order to verify the effectiveness of the proposed algorithm, we construct the dataset of aluminum strip hot rolling process. A large number of experimental results show that this proposed method can be used to build an accurate relationship model between process parameters and crown, and realize the automatic prediction of crown.