A Hybrid Bond Graph Model-based - Data Driven Method for Failure Prognostic
- Failure prognostic builds up on constant data acquisition and processing and fault diagnosis and is an essential part of predictive maintenance of smart manufacturing systems enabling condition based maintenance, optimised use of plant equipment, improved uptime and yield and to prevent safety problems. Given known control inputs into a plant and real sensor outputs or simulated measurements, the model-based part of the proposed hybrid method provides numerical values of unknown parameter degradation functions at sampling time points by the evaluation of equations that have been derived offline from a bicausal diagnostic bond graph. These numerical values are computed concurrently to the constant monitoring of a system and are stored in a buffer of fixed length. The data-driven part of the method provides a sequence of remaining useful life estimates by repeated projection of the parameter degradation into the future based on the use of values in a sliding time window. Existing software can be used to determine the best fitting function and can account for its random parameters. The continuous parameter estimation and their projection into the future can be performed in parallel for multiple isolated simultaneous parametric faults on a multicore, multiprocessor computer. The proposed hybrid bond graph model-based, data-driven method is verified by an offline simulation case study of a typical power electronic circuit. It can be used to implement embedded systems that enable cooperating machines in smart manufacturing to perform prognostic themselves.
Document Type: | Article |
---|---|
Language: | English |
Author: | Wolfgang Borutzky |
Parent Title (English): | Procedia Manufacturing |
Volume: | 42 |
Number of pages: | 9 |
First Page: | 188 |
Last Page: | 196 |
ISSN: | 2351-9789 |
URN: | urn:nbn:de:hbz:1044-opus-48733 |
DOI: | https://doi.org/10.1016/j.promfg.2020.02.069 |
Publisher: | Elsevier |
Place of publication: | Amsterdam |
Publishing Institution: | Hochschule Bonn-Rhein-Sieg |
Date of first publication: | 2020/04/06 |
Note: | Part of special issue: International Conference on Industry 4.0 and Smart Manufacturing (ISM 2019). Edited by Francesco Longo, Feng Qiao, Antonio Padovano |
Keyword: | Unknown parameter degradation; bicausal diagnostic Bond Graphs; failure prognostic; parameter estimation; predictive maintenance; remaining useful life; repeated trend projection; uncertainties |
Departments, institutes and facilities: | Fachbereich Informatik |
Dewey Decimal Classification (DDC): | 6 Technik, Medizin, angewandte Wissenschaften / 62 Ingenieurwissenschaften / 620 Ingenieurwissenschaften und zugeordnete Tätigkeiten |
Entry in this database: | 2020/04/09 |
Licence (German): | Creative Commons - CC BY-NC-ND - Namensnennung - Nicht kommerziell - Keine Bearbeitungen 4.0 International |