Evaluation of novelty detection methods for condition monitoring applied to an electromechanical system
Dealing with industrial applications, the implementation of condition monitoring schemes must overcome a critical limitation, that is, the lack of a priori information about fault patterns of the system under analysis. Indeed, classical diagnosis schemes, in general, outdo the membership probability...
| Autores: | , , , , , |
|---|---|
| Tipo de recurso: | capítulo de libro |
| Fecha de publicación: | 2017 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/118926 |
| Acceso en línea: | https://hdl.handle.net/2117/118926 https://dx.doi.org/10.5772/63169 |
| Access Level: | acceso abierto |
| Palabra clave: | Electromechanical decives Fault-tolerant computing condition monitoring electromechanical systems failure diagnosis feature reduction industrial monitoring applications novelty detection open set recognition Electromecànica Tolerància als errors (Informàtica) Àrees temàtiques de la UPC::Informàtica::Automàtica i control |
| Sumario: | Dealing with industrial applications, the implementation of condition monitoring schemes must overcome a critical limitation, that is, the lack of a priori information about fault patterns of the system under analysis. Indeed, classical diagnosis schemes, in general, outdo the membership probability of a measure in regard to predefined operating scenarios. However, dealing with noncharacterized systems, the knowledge about faulty operating scenarios is limited and, consequently, the diagnosis performance is insufficient. In this context, the novelty detection framework plays an essential role for monitoring systems in which the information about different operating scenarios is initially unavailable or restricted. The novelty detection approach begins with the assumption that only data corresponding to the healthy operation of the system under analysis is available. Thus, the challenge is to detect and learn additional scenarios during the operation of the system in order to complement the information obtained by the diagnosis scheme. This work has two main objectives: first, the presentation of novelty detection as the current trend toward the new paradigm of industrial condition monitoring and, second, the introduction to its applicability by means of analyses of different novelty detection strategies over a real industrial system based on rotatory machinery. |
|---|