Prediction of Strength Characteristics of Fiber-Reinforced Concrete Based on an Intelligent Analysis and Visualization of Multidimensional Data

The problem of predicting the reliability and durability of fiber-reinforced concrete was examined. A complex approach based on neural networks, intelligent analysis, and visualization of multidimensional data that allowed the influence of the characteristics and their interaction with the strength parameters of a concrete composite to be considered was described. The dimensionality of the data was reduced and dependences were found using cluster analysis, weakly bound Kohonen neural networks, and self- organizing maps. The prognostic capabilities of machine-learning models were comparatively analyzed.