Damage Identification of Welded Structures Using Time Series Models and Exponentially Weighted Moving Average Control Charts

The main aim of this paper is to demonstrate a new approach for the health monitoring of structures to identify the damage at earliest possible stage using the acceleration-time data obtained from the piezoelectric accelerometers. This paper presents a unique combination of time series models to extract the damage sensitive features and exponentially weighted moving average (EWMA) control charts to monitor the variations of the selected features. First, the damage sensitive features are extracted by fitting a time series prediction model called an auto-regressive (AR) model to the acceleration-time data obtained from the undamaged structure. Then the residual errors are calculated which quantify the difference between the actual acceleration-time data and the prediction from the AR model at each time interval is defined as the damage sensitive feature. The variation of these features is monitored using EWMA control charts. The applicability of the proposed damage identification approach is tested with the welded structure like cantilever plate. The damage is introduced to the test structure by cutting a slot in the weld using electrical discharge machining. Three damage levels are considered and named damage level zero, damage level one and damage level two. As the outliers are statistically significant in number and are increasing as the damage level increases, it is concluded from the EWMA control charts that this approach not only identifies the presence of damage but also sensitive to the severity of the damage.