Experimental validation of a soft identification algorithm for a MDOF frame structure

A neural networks-based structural identification method using absolute acceleration without mode shapes and frequency extraction is proposed and validated with vibration absolute acceleration measurements from shaking table test of a two-storey frame structure. An acceleration-based neural network modeling for acceleration forecasting and a parametric evaluation neural network for parametric identification are constructed to facilitate the whole identification process. Based on the two neural networks and by the direct use of absolute acceleration measurement time histories of the object frame structure under base excitation, the inter-storey stiffness and damping coefficients of the frame structure are identified. The identified results by the proposed methodology are compared with them by solving eigenvalues equation. Results show that the structural stiffness and damping coefficients identification accuracy is acceptable and the proposed strategy can be a practical tool for model updating and damage detection of engineering structures.