Modal analysis and calibration of finite element model of a three-story steel frame using machine learning and physics-based techniques

Purpose This paper aims to focus on the state-of-the art methods for modal parameters extraction from modal testing. Design/methodology/approach The finite element (FE) model is updated using hybrid method (machine learning-based) and physics-based approaches. A three-story bookshelf frame has been used for the experimental study and a free vibration test has been conducted. The bookshelf frame, made of galvanized steel, has the following dimension: 60 cm width, 27 cm depth and 133 cm height. The frame has been instrumented with tri-axial wireless sensors. Three accelerometers have been installed on each floor of the frame. The frequency domain decomposition (FDD) and modified complex Morlet wavelet methods have been used to extract the modal properties from dynamic response. Findings The extracted results from both methods have been compared, and they are found to be close to each other. The MATLAB-based compiler called M-FEM is used to create FE models. The initial FE model is updated using different approaches. Originality/value The updated FE model output shows the efficiency of hybrid technique in updating the FE model, and the results are well correlated with the physics-based approach.