Related Experiment Videos
Online monitoring by dynamically refining imprecise models.
Bernhard Rinner1, Ulrich Weiss
1Institute for Technical Informatics, Graz University of Technology, A-8010 Graz, Austria. b.rinner@computer.org
Summary
MOSES, a model-based monitoring system, detects faults by comparing system data with imprecise models. It refines uncertainty spaces to identify system failures, ensuring timely and accurate fault detection.
Area of Science:
- Engineering
- Computer Science
- Control Systems
Background:
- Model-based monitoring is crucial for fault detection in technical systems.
- Existing systems struggle with incomplete knowledge, noisy data, and time constraints.
Purpose of the Study:
- To present MOSES, a novel model-based monitoring system.
- To enable fault detection using imprecise models with interval parameters.
Main Methods:
- Utilizes imprecise models with known structure and interval-specified parameters.
- Computes trajectory bounds using numerical integration from uncertainty space boundaries.
- Refines the model's uncertainty space by refuting inconsistent measurements.
Main Results:
- MOSES detects faults when the entire model uncertainty space is refuted.
- The system effectively processes noisy observations and reasons with incomplete knowledge.
- Demonstrates performance through examples and online monitoring of a heating system.
Conclusions:
- MOSES offers a robust approach to model-based fault detection.
- It integrates methodologies from Fault Detection and Isolation (FDI) and Diagnosis (DX) communities.
- The system provides conservative uncertainty space refinement and early exploitation of measurements for online monitoring.