| Popis: |
In addition to the dynamic characteristics; namely the natural frequencies and mode shapes, the structural and accidental eccentricities are important parameters for the seismic evaluation of existing buildings. Hence the precise distance between the mass and rigidity centers including the accidental eccentricity is required for classification and analysis. This paper presents a method which utilizes dynamic responses of at least two points on a building floor to identify the corresponding mass and rigidity centers using an artificial neural network. For this purpose, a model of neural network has been elaborated and trained using the responses of two distant points on the floor of a multistory structure. The sensitivity of the model has been investigated for the different influencing parameters such as the type and the length of the input signals as well as the structure characteristics. The conditions where the best predictions can be obtained were identified and interpreted. The merits and the limits of the proposed model were exposed together with illustrative examples. More importantly is the potential to use this method for experimental data from instrumented buildings or ambient vibration records. |
