Multi-step-ahead prediction of cold rolling chatter state based on the combination of Functional Data Analysis and General Autoregression Model

Abstract

Mill chatter seriously restricts the improvement of production efficiency and the development of new products. Thus, a method of cold rolling chatter monitoring and early warning is proposed based on the combination of Functional Data Analysis (FDA) and General Autoregression Model (GAM). Firstly, the multi-source cold rolling data are preprocessed by FDA to realize the smooth fitting of the unequally sampled data and the sample space is constructed based on chatter mechanism. Then, the multi-step-ahead prediction of cold rolling chatter is executed through different machine learning algorithms based on GAM, and the prediction effects of different algorithms are compared and evaluated. Finally, the maximum prediction step is defined to select the optimal algorithm, and the conclusion is drawn that Extra Tree Regression (ETR) has the best prediction effect. Therefore, the proposed method for cold rolling chatter monitoring can effectively solve the problems of delay, false alarm and distortion in prediction.