Deep Neural Network (DNN) Model to Predict Close-In Blast Load

In recent years, the use of machine learning (ML) has been expanded to several engineering fields with applications in structural engineering. Deep neural network (DNN) models have been implemented to predict structural response of systems under conventional loading. Some of those DNN models are based on data sets of images, test data, and/or finite element models built for a specific environment; the accuracy of these models relies on the size of the data set that can vary from hundreds to thousands of data points. Since DNN models rely on data, their use in blast analysis and/or design is limited to the availability of data, which most of the time is scarce or restricted. This paper introduces the implementation of an ML/DNN model for the prediction of close-in blast loads based on experimental data from multiple blast test programs managed and executed by Stone Security Engineering, P.C. The blast tests were conducted at the Stone-OBL Open-Air Blast Test Facility (outside of Bend, Oregon) with multiple threat sizes and standoffs from the Stone-OBL rigid reaction structure. The current intent of this ML/DNN model is to provide a prediction tool to blast engineers to easily estimate loads from close-in range scenarios in upcoming blast test programs, thus reducing cost and time and optimizing the level of calibration tests that might be needed. This paper shows the accuracy of the model based on test data and compared with computational fluid dynamics (CFD) runs. Example cases are presented to illustrate ML/DNN model output.