Experimental research on frequency based damage identification of beams with free boundary condition

It is essential for decision makers to obtain real-time information on damage to a ship after it has been damaged, but there is no effective method currently. A method for identifying ship damage based on frequency is proposed, which is a data-driven approach using neural networks. The frequency database under different damage con-ditions is obtained by migration matrix method, optimized by normalization method, and trained by Probabi-listic Neural Network (PNN) to form the agent model. Considering the limitations in the deployment of measurement points in practical applications, an optimization method based on Modal Assurance Criterion (MAC) that takes into account the offset of the measurement points is introduced and studied. Methods presented are experimentally validated by a simplified beam model which simulates free boundary conditions of ship. Results show that the agent model constructed can accurately identify the damage location. The damage quantification error decreases with the increase of the preset damage extent, which indicates that the accuracy is higher for larger damage. The test scheme optimization method can effectively obtain the best measurement point placement scheme for efficient measurement of modal parameters. The results of this paper can provide technical and methodological support for the real-time identification of ship damage.

Automated summary

This paper proposes a frequency-based method for ship damage identification using neural networks. Experimental results show that the proposed method can accurately identify the damage location, with the quantification error decreasing as the damage extent increases.