Structural damage identification under non-linear EOV effects using genetic programming

The methods to perform long-term condition monitoring of structures can be generally divided into two main categories, i.e. output-only and input-output methods. The former seeks to detect any anomaly in the structural modal data caused by damage without knowing the cause. Input-output normalisation methods have been used for removing Environmental and Operational Variations (EOV) effects, some examples of which include multiple linear regression [1], artificial neural networks [4] and support vector regression [2].