Abstract
Monitoring the vibrations of a gas turbine during its development tests requires manual inspection and interpretation of thousands of diagrams displaying measurements made by hundreds of strain gauges and other transducers. This paper describes a prototype knowledge-based computer system to assist in this task. The features on each diagram are located by image processing methods and interpreted by stored knowledge which represents the experience of vibration engineers. The objective of the system is to distinguish the relatively few diagrams which have unusual features from the majority which have only expected features, leaving the interpretation of the unusual diagrams to real experts.
The paper gives an overall description of the problem of vibration in gas turbines and some details of the image processing and knowledge processing methods used, and shows some examples of typical diagrams handled by the system.
Get full access to this article
View all access options for this article.