Description
Sandwich panels in civil engineering have the advantage of a high flexural strength and thermal insulation while providing a very low density. They are made of two steel sheets and a core of polyurethane or polyisocyanurate foam. Due to the manufacturing process, the panels occasionally have defects, that are not visible from the outside. These may cause visible damages after installation in building facades. This paper presents a workflow for the detection and localization of the defects exploiting their influence on the response to a dynamic excitation. It uses an experimentally validated finite element model, feature engineering, and a neural network. The workflow initiates with the development and validation of a mechanical finite element model using experimental data. The model serves as the foundation for a simulation campaign, examining both defect-free and models with defects. The simulation process incorporates variations in material parameters based on experimentally obtained standard deviations, ensuring a robust representation of real-world uncertainties. The resultant simulation dataset is subjected to a feature engineering process encompassing time and frequency domain features. This multi-domain feature extraction enhances the dataset's informativeness. The feature dataset is employed to train a neural network for two primary use cases: Firstly, to differentiate between faulty and fault-free models, providing an accurate fault detection mechanism. And secondly, to localize faults within the model. The effectiveness of this approach is demonstrated using various testing datasets. The proposed methodology represents a combination of a synthetically created database from a validated finite element model, multi-domain feature engineering, and neural network capabilities. It offers a methodology for fault detection and localization in sandwich panels. Using a dataset of several hundred samples, a detection rate of 94.8 % and a localization rate of 81.3 % are achieved in this work. In practical use, this approach can lead to significant reduction of customer complaints, and savings in replacement costs and hence carbon footprint.
