This paper attempts to solve the mode recognition problem using the CNN (convolutional neural network) algorithm. A program is designed by Pytorch and PCL (Patran command language) to recognize both the natural frequency of ship hull and grillage. After building a training set with pictures of ship vibration mode, the VGG model is built to acquire features of the model. Meanwhile, automatic screenshots of the modal analysis results in Patran are captured and sent to the recognition program. Derived from multiple sets of test cases, the results show that mode recognition accuracy and efficiency are satisfactory. The practice of this paper has provided a new method for the fast evaluation of ship vibration.


To control design direction at a primitive stage, it is repeatedly required to calculate the natural frequency of ship hull or grillage during the structural design process. Normally, mode analysis is conducted in FEM (finite element method) software with results spotted by visual recognition and it takes a great deal of time and human resources to obtain proper vibration mode. Designers pick out the required vibration modes and their eigenvalues are the natural frequency.

The first and the second order of vertical and horizontal vibration frequency of ship hull should be obtained and compared with the primary vibration source which is the main engine, to avoid resonance. General vibration modes are more obviously acquired than local grillage modes. As for grillage, the range we are concerned with is the stiffened plates between two bulkheads. The natural frequency of grillage is a vital design parameter that instructs the designer to mismatch the incentive source in the vicinity. Meanwhile, a lot of ship equipment has requirements for grillage stiffness, so knowing the natural frequency of grillage is significant for equipment's function and the vibration and noise control of hull structure.

This content is only available via PDF.
You can access this article if you purchase or spend a download.