Crack identification driven by the fusion of mechanism and data for the Variable-Cross-Section cantilever beam

Abstract

Crack identification for variable-cross-section cantilever structures is a main means of avoiding sudden failure caused by cracking and realizing the prognostics and health management (PHM) of such structures. This paper proposes an inverse crack identification method driven by mechanism and data for variable-cross-section cantilever beams. First, the Rayleigh method is revised for the fast, accurate calculation of the natural frequencies of a healthy cantilever beam, which is regarded as a basic criterion. Subsequently, for the rapid, accurate expression of the relationship between crack information and natural-frequency variations, a data-driven method based on the radial basis function is proposed to calculate the response variations accurately according to the crack information. Then, the obtained acceleration of the cantilever beam is decomposed to determine its natural frequencies, and the variations in these natural frequencies are obtained and compared with the results of the mechanism model. Finally, crack positions, lengths, and widths are identified based on the forward model and decomposition results. Particle swarm optimization is employed to minimize the discrepancy in the changes between the monitored and calculated natural frequencies. The proposed forward model is compared with other methods, and the superiority of the crack identification method is verified experimentally. Results indicate that the proposed method accurately identifies cracks on variable-cross-section cantilever beams and provides an effective PHM approach for beam-type structures.