Reduction of temperature effects for bridge health monitoring

Structural health monitoring of concrete bridges can be achieved by tracking static load-testing results or dynamic properties as for example eigenfrequencies. Deviations from a healthy reference state can be used as damage indicators and even more, help to localize zones of stiffness reduction, i.e. cracking. However, outdoor temperature effects also lead to changes of monitored physical characteristics in the same order of magnitude as damage. Hence, temperature effects need to be removed prior to any condition analysis. The present paper presents a new two-step approach by applying physical compensation first, before using a statistical method based of Principal Component Analysis (PCA) or more exactly on principal vectors and singular values. This technique is here applied to eigenfrequencies, first of a new bridge without damage, but with extreme temperature variation due to thick asphalt layer and special bearing constraints, thus showing strong sensitivity along seasonal temperatures in the intact state. The second object is the Z24 Bridge in Switzerland, which is well documented in literature and where artificial damage was applied prior to demolition. The proposed techniques allow removing noise and temperature effects in a coherent and efficient way. The corrected measurement data can then be used in subsequent steps for its definite purpose, i.e. detection and localization of damage for instance by updating a numerical finite element model which allows assessing a stiffness loss.