Firstly, the finite element model of the large tower crane is established, and the validity of the model is verified by the type test report of the tower crane. From this model, the dynamic displacement signals of the tower crane in the intact state and damaged state are obtained. Based on the time series analysis of the intact state data, the AR model is established, the mean values of the model parameters are obtained. The displacement values of the unknown are used to fit the model, and the residuals and the mean values in each group are used as eigenvalues. An adaptive calculation method of clustering radius in density-based clustering is proposed, and the extracted eigenvalues are further analyzed by density-based, which achieves automatic classification of tower cranes in intact states and in structurally damaged states. The analysis result is further verified by single-limb experimental data. This diagnosis method of density-based clustering needs only a few sensors for automatic diagnosis of the structural state of the tower crane.
