IDENTIFICATION AND LOCALIZATION OF IMPACT LOAD ON CANTILEVER PLATE BY DEEP NEURAL NETWORK SYSTEM

Accurately obtaining information dynamic load information acting on a structure is crucial for ensuring its dynamic strength. However, these loads are often unmeasurable and difficult to precisely identify. Although deep neural networks-based dynamic load identification methods have the advantage of not relying on precise forward models and matrix inverse operations, their generalization performance is difficult to guarantee. Additionally, research on using deep neural networks for dynamic load localization is still relatively limited. This paper proposes a method for localizing impact loads using deep neural networks. The method involves normalizing the structural vibration signal, extracting its feature signal, enhancing signal information, and using it as the training dataset for deep neural networks. The effectiveness of this method was verified through simulation of a cantilever plate structure.