Advancing structural integrity prediction with optimized neural network and vibration analysis

Abstract

Detecting and locating damage is a crucial endeavor within the field of structural integrity. While Artificial Neural Networks (ANNs) have shown promise in this regard, they have certain limitations that can be overcome through modifications in terms of their structural design and training methodologies. In this study, we propose a new optimization approach, specifically leveraging the Grasshopper Optimization Algorithm (GOA), to enhance the performance of ANNs for predicting multiple damages represented by holes in the aluminum plate. Input parameters are derived from natural frequencies, while hole locations serve as outputs. We utilize a Finite Element Model (FEM) to generate data through simulation, varying hole locations for comprehensive analysis. To authenticate our method, we gather experimental data from vibration analyses of damaged plates spanning various hole locations. A comparative analysis is conducted of proposed algorithm by evaluating its performance against two established metaheuristic algorithms: the Genetic Algorithm (GA) and Ant Colony Optimization (ACO). This comparison was performed to assess the relative effectiveness of our approach. Our novel approach demonstrates superior performance in damage forecasting, offering promising prospects for structural integrity applications.