Strategies for synchronizing chocolate conching batch process data using dynamic time warping

In batch processing, process control is typically carried out comparing trajectories of process variables with those in an in-control set of batches that yielded products within specifications. However, one strong assumption of these schemes is that all batches have equal duration and are synchroniz...

Full description

Saved in:
Bibliographic Details
Published in:Journal of food science and technology 2020-01, Vol.57 (1), p.122-133
Main Authors: Peres, Fernanda Araujo Pimentel, Peres, Thiago Neves, Fogliatto, Flávio Sanson, Anzanello, Michel Jose
Format: Article
Language:English
Subjects:
Alignment
Batch processing
Chemistry
Chemistry and Materials Science
Chemistry/Food Science
Chocolate
Classification
Food Science
Nutrition
Original
Original Article
Process control
Process controls
Process variables
Synchronism
Synchronization
Warping
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In batch processing, process control is typically carried out comparing trajectories of process variables with those in an in-control set of batches that yielded products within specifications. However, one strong assumption of these schemes is that all batches have equal duration and are synchronized, which is often not satisfied in practice. To overcome that, dynamic time warping (DTW) methods may be used to synchronize stages and align the duration of batches. In this paper, three DTW methods are compared using supervised classification through the k -nearest neighbor technique to determine the in-control set in a milk chocolate conching process. Four variables were monitored over time and a set of 62 batches with durations between 495 and 1170 min was considered; 53% of the batches were known to be conforming based on lab test results and experts’ evaluations. All three DTW methods were able to promote the alignment and synchronization of batches; however, the KMT method (Kassidas et al. in AIChE J 44(4):864–875, 1998) outperformed the others, presenting 93.7% accuracy, 97.2% sensitivity, and 90.3% specificity in batch classification as conforming and non-conforming. The drive current of the main motor was the most consistent variable from batch to batch, being deemed the most important to promote alignment and synchronization of the chocolate conching dataset.
ISSN:0022-1155
0975-8402
DOI:10.1007/s13197-019-04037-5