Evaluation of dynamic damage indicators on real-life civil engineering structures: measurement uncertainty and environmenal influences considered

The thesis examines the vibration analysis used for condition control on civil engineering structures. Experimental data of laboratory and in situ structures are presented. Force excitation tests as well as a monitoring system with an automated analysis were conducted to collect the experimental data. The investigated structures were precast, reinforced and prestressed, slab elements under laboratory conditions, as well as a two span composite bridge and a two span prestressed box girder bridge in situ. The precast slab elements and the prestressed box girder bridge were subjected to artificial damage with crack formation until ultimate load. These structures were measured stepwise to have different damage states in order to quantify the possible use on damage assessment using vibration analysis. The dynamic analysis consisted in yielding modal parameters, especially the eigenfrequencies and the mode shapes, in order to show the changes according to the damage states. A particular focus is given to the flexibility matrices which clearly show the changes caused by gaining damage in contrast to the intact state. However, the damage must be over a certain empiric threshold limit, as environmental impacts and measurement uncertainty also leads to changes of modal parameters, shown by repeated force excitation tests on the investigated structures. Moreover the influence of the environmental impacts, especially the temperature, on the damage indicators are investigated by repeated measurements using force excitation tests in winter and summer, as well as on data captured by a monitoring system on the two span composite bridge in Useldange. In terms of the eigenfrequencies, variations up to approximately 5% for the measurement uncertainty and up to 1% per Kelvin for environmental impacts are noticed without any damage, which is relatively high in comparison to effects of damage. Considering the most significant damage indicator, the flexibility matrix, changes of 10% for repeated measurements and approximately 2.5% per Kelvin for environmental influences are recognised. Contrarily, bigger real damage can cause over 10% changes on the eigenfrequencies and even over 20% changes in the flexibility matrices, which are therefore obviously detectable. Another part is the use of model updating techniques in order to show the damage locations on the investigated structures by finite element models. The main conclusion is that care should be taken when measuring and analysing the results as small variations cannot only be assigned to damage, but could also be caused by the above mentioned environmental and practical influences. To this end, recommendations minimising these effects are proposed.