A novel natural frequency-based technique to detect structural changes using computational intelligence

Structural changes are usually associated to damage occurrence, which can be caused by design flaws, constructive problems, unexpected loading, natural events or even natural aging. The structural degrading process affects the dynamic behavior, leading to modifications in modal characteristics. In general, natural frequencies are sensitive indicators of structural integrity and tend to become slightly smaller in the presence of damage. Despite this, it is very difficult to state the relationship between decreasing values of natural frequencies and structural damage, since the dynamic properties are also influenced by uncertainty on experimental data and temperature variation. In order to contribute to improving the quality of natural frequency-based methods used for damage identification, this paper presents a simple and efficient strategy to detect structural changes in a set of experimental tests from a real structure using a computational intelligence method. For a full time monitored structure, the evolution of natural frequencies and temperature are used as input data for a Support Vector Machine (SVM) algorithm. The technique consists on detecting structural changes and when they occur based on the structural dynamic behavior. The results obtained on a historic tower show the capacity of the proposed methodology for damage identification and structural health monitoring.