Title :
A DKPLS reconstruction algorithm for fault diagnosis
Author :
Yingwei Zhang ; Yunpeng Fan ; Rongrong Sun
Author_Institution :
Key Lab. of Integrated Autom. of Process Ind., Northeastern Univ., Shenyang, China
Abstract :
In this paper, a directional kernel partial least squares (DKPLS) reconstruction method for process monitoring is proposed. Firstly, in order to build a more direct relationship between the input and output variables, a new KPLS algorithm which is called DKPLS algorithm is proposed to extract the output-relevant variation. And then, the fault direction is determined by calculating fault magnitude of every principal component. At last, the fault is effectively diagnosed compared to the conventional KPLS method. The proposed method is applied to electro-fused magnesia furnace and is compared to KPLS method. Experiment results show that the selection of fault direction is more accurately and the proposed method can more effectively diagnose the fault.
Keywords :
fault diagnosis; furnaces; least squares approximations; principal component analysis; process monitoring; DKPLS reconstruction algorithm; directional kernel partial least squares; electro-fused magnesia furnace; fault diagnosis; principal component analysis; process monitoring; Electrodes; Fault diagnosis; Furnaces; Kernel; Monitoring; Principal component analysis; Reconstruction algorithms; Directional Partial Least Squares (DKPLS); Fault Diagnosis; Fault Reconstruction; Output-relevant Variation;
Conference_Titel :
Control and Decision Conference (CCDC), 2015 27th Chinese
Conference_Location :
Qingdao
Print_ISBN :
978-1-4799-7016-2
DOI :
10.1109/CCDC.2015.7162600
