Projects / Programmes
An intelligent system for condition monitoring of rotating machinery
| Code |
Science |
Field |
Subfield |
| 2.06.00 |
Engineering sciences and technologies |
Systems and cybernetics |
|
| Code |
Science |
Field |
| T121 |
Technological sciences |
Signal processing |
| T125 |
Technological sciences |
Automation, robotics, control engineering |
| T150 |
Technological sciences |
Material technology |
| P175 |
Natural sciences and mathematics |
Informatics, systems theory |
monitoring, fault diagnosis, signal processing, approximate reasoning,
Researchers (9)
Organisations (2)
Abstract
In the majority of manufacturing and process industries there is a growing need to replace periodical checks of the physical condition of rotating machinery with reliable on-line condition monitoring systems able to reveal tentative causes for malfunction as early as in their incipient stage. As worsening of machine condition typically progresses gradually, the maintenance actions can be planed well before a fault evolves into a failure provoking system break-down. In the underlying subproject an intelligent system for condition monitoring of rotating machinery will be developed. The main objectives of the subproject are:development of the feature extraction modules in time and frequency domain,construction of the knowledge base, prescription of the rules for maintenance action on the base of diagnosed state of machine components,development of reasoning mechanism to be used in on-line or off line conditions,design of the integrated experimental environment,design of a test rig and test of the development intelligent system.One of the key ideas of the underlying concept is to fully exploit the diversity of information sources i.e. standard on-line measurement instrumentation complemented with off-line special purpose laboratory checks of wear particles and lubricant samples. Additional power of the system reflects in accommodation of feature extraction mechanisms based on up to date signal processing techniques. These will include spectral analysis techniques such as Fast Fourier Transform, Time Fourier Transform, wavelet analysis, parameter spectrum analysis and envelope analysis – to mention the most relevant. The configuration will depend on the operating conditions and required sensitivity of the system with respect to the anticipated faults. The next important part of the system is a hybrid reasoning mechanism specially adopted to take full advantage of available domain knowledge and data driven classification methods. The innovative open world paradigm behind reasoning approach will encounter means to properly associate the beliefs for suspected faults as will be able to deal with unanticipated faults. The system will be equipped with self-learning capabilities owing to employed data driven machine learningtechniques.
