Fault detection using neuro-fuzzy networks
Generally three methodologies to develop and test fault detection (FD) algorithms can be distingguished: software benches, hardware benches and industrial data. The current approach uses a hardware bench that consists of process under supervision (two interconnected stations), supervision unit, faul...
| Autor principal: | |
|---|---|
| Outros Autores: | , , |
| Formato: | article |
| Idioma: | eng |
| Publicado em: |
2019
|
| Assuntos: | |
| Texto completo: | http://hdl.handle.net/10400.21/10789 |
| País: | Portugal |
| Oai: | oai:repositorio.ipl.pt:10400.21/10789 |
| Resumo: | Generally three methodologies to develop and test fault detection (FD) algorithms can be distingguished: software benches, hardware benches and industrial data. The current approach uses a hardware bench that consists of process under supervision (two interconnected stations), supervision unit, fault diagnosis unit and fault simulation unit. All elements of the bench are connected to a PROFIBUS network that acts as the communication system exchaging information between automation system and distributed field devices. A realistic and fexible environment for developing and testing FD systems has been constructed using elements commonly used in industry. During the current studies actuator faults, sensor faults and leakages have been considered as incipiente and abrupt faults. The proposed FD algorithm bases on neuro-fuzzy models that are responsible for residual generation. |
|---|