A novel time-domain pattern-based approach for trending is presented. The approach is based on an interval-based qualitative and semi-quantitative representation capable of transforming the time-records of data into meaningful and explicit descriptions of trends at different time scales. The representation of trends is based on the triangular representation , whereas the detection of trends from data records is based on qualitative scaling. Together, they provide a unified formal framework for the consistent detection and representation of trends from arbitrary noisy data in a compact and natural manner. Because the representation is generic to trends and intuitive to humans, it can provide useful multi-scale dynamic data models for different process applications such as data compression, data reconciliation and rectification, fault diagnosis, trend interpretation, adaptive control, quality control, learning patterns from historical data, economic evaluation, process modelling, monitoring, planning and optimization.
|Number of pages
|Advances in Instrumentation, Proceedings
|Published - Dec 1 1990
|Proceedings of the ISA '90 International Conference and Exhibition Part 4 (of 4) - New Orleans, LA, USA
Duration: Oct 14 1990 → Oct 18 1990
ASJC Scopus subject areas
- General Engineering