부산대학교 도서관

Aiming at the non-linear, multi-stage and high dimension characteristics of the plasticizing and molding process of single-based gun propellant, a quality prediction method based on GG-KECA-RVM multi-stage model fusion is proposed. The method is based on Gath-Geva dynamic fuzzy clustering to identify the stages of the plasticizing and molding process. KECA is introduced for deep feature extraction in each stage, and the local latent variable regression models based on KECA-RVM are established for each sub-stage. Finally, the fuzzy membership degree of Gath-Geva clustering is used to fuse the prediction results of multiple local models, which reflects the difference and cumulative characteristics of each stage on the quality, and realizes the accurate prediction of stage quality and process endpoint quality. The experimental results of the plasticizing and molding process show the effectiveness of the proposed method.