Suggestion of a method for predicting different response characteristics including major cracks induced by blast loading in concrete slabs using machine learning

Abstract

The paper suggests a technique for predicting major cracks and other response characteristics of different concrete slabs under blast loadings using the methods of supervised statistical machine learning. To construct the input-output dataset for the learning stage, data from finite number of experiments with different spherical and cylindrical charges have been used. The parameters of the covariance function have been optimized maximizing the log of the marginal likelihood for the experimental results. The obtained parameters are then used in the inference stage when the behaviour of the concrete slab is statistically predicted for new blast cases not considered in the experiments. The results demonstrate that by conducting only a finite number of experiments and by applying the proposed machine learning techniques, one can predict the responses of the concrete slabs for infinite number of combinations of the parameters of the experiment. This can significantly decrease the number of experiments required for constructing reliable mechanical models. The proposed method provides predictions even in the case when there are only a few experimental results. The uncertainty will be high in this case of few experiments; however the uncertainty will be significantly reduced once new experimental results are provided to the proposed Bayesian predictive model.