255 / 2021-04-15 11:30:46

Multiple Damage Identification Analysis of Frame Structure by Artificial Neural Network

Multiple Damage Identification; Framed structure; Artificial neural network; Inverse problem; Natural frequencies

Structural Health Monitoring

In recent years, structural health monitoring techniques that can accurately determine the location and degree of damage caused by aging, earthquakes, etc. have been used to ensure the safety of structures. For example, visual inspection and ultrasonic testing are used as non-destructive inspection methods that have little influence on the structure. Although these methods are effective for relatively small structures, they require a great deal of time and effort to ascertain the health of large structures. Therefore, it is an important issue to develop a highly accurate and efficient damage identification analysis method. In this paper, a damage identification analysis method using neural networks (ANN) is proposed for the multiple damage identification problem of frame structures. Most of the conventional damage identification analysis methods use both natural frequencies and eigenmodes as measurement data.  However, it is not easy to measure the eigenmodes, and even if it is possible, it requires a lot of time and cost to process a large amount of measurement data. In order to apply the method to real problems, it is desirable that the amount of measurement data required for identification is small. Therefore, in this paper, we propose a damage identification analysis method using only natural frequencies as measurement data. In addition, the proposed method is applied to a relatively difficult multiple damage identification problem and its effectiveness is verified by numerical calculations. In particular, it is examined how accurate damage identification can be achieved from a small amount of observation data.