Research on damage identification of hull girder based on Probabilistic Neural Network (PNN)

Abstract Real-time localization and quantitative assessment of hull girder damage are indispensable for subsequent decisions. To deal with the difficulties that ship damages are hard for real-time assessment, this paper proposes an indirect damage identification method based on Probabilistic Neural Network (PNN), utilizing the natural frequency changes to indirectly identify the damage. The natural frequency database of the damaged hull girder is created by the transfer matrix method, and the normalization method is applied to optimize the database characteristics. Based on the Particle Swarm Optimization Algorithm (PSO) and Genetic Algorithm (GA), the influence of different parameters of the optimization algorithms on the smooth factor is discussed. On this basis, the best smoothing factor of the PNN is searched to improve the identification accuracy and efficiency. Results show that the optimal results of the two optimization algorithms are close. And the GA owns higher iterative efficiency. The PNN optimized by the optimization algorithms has higher damage identification accuracy. The method proposed in this paper provides a novel approach for the real-time identification of hull girder damage and contributes reference significance for engineering application.