PCA-BASED NOVELTY DETECTION USING MODAL FLEXIBILITY AND NATURAL FREQUENCY UNDER TEMPERATURE VARIATIONS

In SHM, the vibration features are not only affected by damage in structure, but also environmental conditions. The subtle change caused by delicate damage in structure may be masked because of such effects. To make structural health monitoring more reliable, it is highly necessary to select proper modal parameters sensitive to structure damage and eliminate environmental effects on modal parameters. In this paper, modal flexibility and natural frequency are chosen as damage features. Via numerical model of simply supported beam, temperature effects on modal flexibility and natural frequency are analyzed. PCA (principal component analysis) and novelty detection method are applied to these features under environmental variations. Baseline states of structure are established on the residual errors of the PCA model under healthy condition. Damage is diagnosed by examining whether or not residual errors of the PCA model deviate significantly from the baseline. Comparative analyses of PCA-based novelty detection using modal flexibility and natural frequency are conducted. It's demonstrated that the PCA-based damage detection method using modal flexibility gives better performance than the method using natural frequencies, taking into account the effects of temperature variations.