Title: Fault diagnosis for a milk pasteurisation plant with missing data

Authors: Ouahab Kadri; L.H. Mouss; Adel Abdelhadi

Addresses: Department of Computer Science and Math, University of Khenchela, Algeria ' Automatic and Computer Integrated Manufacturing Laboratory, University of Batna 2, Algeria ' Automatic and Computer Integrated Manufacturing Laboratory, University of Batna 2, Algeria

Abstract: This paper addresses the problem of fault diagnosis from observed data containing missing values amongst the inputs. In order to provide good classification accuracy for the decision function, a novel approach based on support vector machine and extreme learning machine is developed. SVM mixture model is used to model the data distribution, which is adapted to handle missing values, while extreme learning machine enables to devise a multiple imputation strategy for final estimation. In order to prove the efficiency of our proposed method, we have developed the classifier using the condition monitoring observations from milk pasteurisation data. The experiments show that the proposed algorithm gives improved results compared to recent methods, essentially if the number of missing data is significant. The results show that our approach can perfectly detect dysfunction, identify the fault, and is strong in unsupervised process monitoring.

Keywords: fault diagnosis; missing data; support vector machine mixture; extreme learning machine; ELM; multiple imputation; least squares; process monitoring; milk pasteurisation; classification; multiclass problem.

DOI: 10.1504/IJQET.2017.10009786

International Journal of Quality Engineering and Technology, 2017 Vol.6 No.3, pp.123 - 136

Received: 16 Aug 2016
Accepted: 11 May 2017

Published online: 18 Dec 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article