Structural system identification and strain measurement based model updating with closed-loop data

Abstract

This research deals with both structural system identification and strain mesaurement-based model updating using the closed-loop data that are the input and output data sets experimentally obtained from the closed-loop system. As the first research, we explain how to enhance structural system identification performance with the closed-loop data. Structural system identification generally utilizes the open-loop data except the case that the open-loop system is inherently unstable. The number and placement of actuators and sensors can change the open-loop data, thus, affecting the mathematical model of the identified system. If we happen to place actuators or sensors near the nodal points of the specific mode and collect the input and output data sets with noise, we are likely to omit that mode, thereby making a mistake of determining the system order. The barely visible modes due to placing actuators or sensors near the nodal points when there are measurement errors are called the almost hidden modes in this work. We need to optimize placement of actuators and sensors before collecting the experimental data to include the almost hidden modes. We firstly propose the method to find the almost hidden modes and determine the system order and modal data more accurately, thus enhancing structural system identification, instead of finding the optimal placement of actuators and sensors. As the second research, we deal with the method of improving model updating with closed-loop data. Structural system identification reduces to model updating problem when the structure and order of the mathematical model is known. The amount of modal data, such as natural frequencies and mode shapes, can determine the performance of model updating. The identified natural frequencies are generally much less sensitive to noise than the identified mode shapes. If we choose the only natural frequencies for model updating because of their accuracy, we are likely to be difficult to secure t...