Abstract
This paper proposes a methodology for the health monitoring of structures capable of identifying the damage at the earliest possible stage using the acceleration time response data obtained from piezoelectric accelerometers. In this, a unique combination of databased models to extract the damage sensitive features and Shewhart control charts to monitor the variations of the selected damage sensitive features are presented. The applicability of the proposed method is tested with the welded structure model by fixing it to the multi axes electro dynamic vibration shaker. In this, damage will be defined as a crack through the thickness of the weld and damage was introduced into the structure by cutting a slot in the weld using electric discharge machining. The acceleration time response data from the test structure are measured for five damage levels. Damage sensitive features are extracted by fitting a time prediction databased model called an Auto Regressive model to the measured acceleration time response data obtained from the undamaged structure. Then, the residual errors are calculated at each time step. To monitor the variation of mean and standard deviation of the extracted damage sensitive features, X-bar and S control charts are used. It is found that X-bar and S charts are capable of identifying the presence of damage for different damage levels considered in this work.