The effects of variability on damage identification with inductive learning
This work discusses the effects of inherent variabilities on the damage identification problem. The goal of damage identification is to detect structural damage before it reaches a level which will detrimentally affect the structure’s performance. Inductive learning is one tool which has been proposed as an effective method to perform damage identification.
There are many variabilities which are inherent in damage identification and can cause problems when attempting to detect damage. Temperature fluctuation and manufacturing variability are specifically addressed. Temperature is shown to be a cause-effect variability which has a measurable effect on the damage identification problem. The inductive learning method is altered to accommodate temperature and shown experimentally to be effective in identifying added mass damage at several locations on an aluminum plate.
Manufacturing variability is shown to be a non-quantifiable variability. The inductive learning method is shown to be able to accommodate this variability through careful examination of statistical significances in dynamic response data. The method is experimentally shown to be effective in detecting hole damage in randomly selected aluminum plates from a manufactured batch.