Machine learning for impact detection on composite structures. (2021)

• Record Type: Journal Article
• Authors: Cuomo, Stefano; De Simone, Mario Emanule; Andreades, Christos; Ciampa, Francesco; Meo, Michele
• Is Part Of: Materials today. Volume 34(2021)Supplement Part 1
• Page Start: 93
• Page End: 98
• Publication Date: 2021
• Subjects: Low velocity impact -- BVID -- Machine learning -- Cross correlation -- Impact localization; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2020.01.295
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: In order to overcome the current limitations of the impact localisation process in composite materials, such as the a-priori knowledge of the mechanical properties and the direction dependency of the wave speed, a novel method is here proposed based on the machine learning approach. The algorithm is formed by two steps: the first is the training process, in which a baseline consisting of the structural responses due to impact tests is acquired; the second one evaluates the impact location exploiting the highest cross-correlation coefficient, obtained after the interpolation of the impact response baseline using the Radial Basis Function (RBF) method. Numerous experimental tests are performed on a simple carbon fibre reinforced polymer (CFRP) plate fitted with three piezo-sensors at three different drop heights to validate the training process. The results showed high accuracy in both the reconstruction and the impact localisation, with an error less than 10 mm.