Machine learning (ML) has been expanded to several fields in recent years, with promising advances in structural engineering applications. In particular, Deep Neural Network (DNN) models have been created and used to predict the structural response of systems under traditional loading. However, there remains an opportunity to expand this capability to non-conventional loading. This study presents the use of a DNN model to predict the response of reinforced concrete (RC) walls to contact and near-contact explosions of bared charges.
When evaluating blast loading on RC walls or slabs, the focus is often on assessing flexural and shear behaviour, for which performance guidelines are readily available. However, when blast loads are caused by contact or near-contact detonations, the response is controlled by local failure modes rather than far-field range mechanisms. This research uses a DNN model to predict damage categories (i.e. breach, spall, or no damage) associated with a given explosion scenario. The model is trained using experimental data from multiple test programs available in open-source literature. It considers several parameters related to the explosive charge and RC target. The current model is able to accurately predict 81% of the total breached specimens, 66% of the total spalled specimens, and 71% of the full set of non-damaged specimens, with an overall accuracy of 72%. The DNN Model is shown to be a significantly better predictor of the damage category than the semi-empirical approach outlined in UFC 3-340-02, making it a promising tool that can be improved with the inclusion of more experimental data.
The Terrorism Risk Assessment, Modelling and Mitigation Seminar Series (TRAMMSS) is a virtual seminar series focused on technical topics related to terrorism risk assessment, and modelling, including blast modelling and response; IEDs; vehicles as weapons; CBRN; big data for risk assessment, security and screening; and associated mitigation measures.
Performance prediction of RC wall under contact/near-contact explosives using Deep Neural Networks (David Holgado)
David Holgado graduated in 1985 as Civil Engineer in the national University of Engineering (Lima-Peru), Post-graduate of Earthquake Engineering from the International Institute of Earthquake Engineering and Seismology from Tsukuba (Japan) and with a Master’s Degree in Civil Engineering from the University of Texas at San Antonio and has over 30 years professional experience in design and analysis of a wide range of structures with a general background in structural, earthquake engineering and in-depth expertise in the area of structural dynamics.
Since 2004, Mr. Holgado specializes in the areas of structural analysis and design of structures subjected to extreme loads such as blast, ballistic, high velocity impact, vapor cloud explosion, Artificial Neural Network and Computer Vision in blast applications. Also, Mr. Holgado has been part of teams performing several investigations of accidental explosions in the US and abroad. Mr. Holgado has participated in vulnerability, threat, and risk assessments for more than 50 facilities including blast vulnerability assessments of the US General Services Administration’s (GSA) Federal Buildings and Courthouses, as well as analysis and design of new federal buildings and courthouses. Mr. Holgado has contributed several outside-the-box recommendations for Antiterrorism and Force Protection (ATFP) government building assessments, retrofit and new design.
