The article provides a review and systematisation of works devoted to the application of machine learning for solving problems of research, calculation and design of reinforced concrete structures. It considers the aspects, which are relevant today, related to calculation, design, as well as assessment of the technical condition of objects with the help of various approaches that implement machine learning schemes, including deep learning, ensemble algorithms. It is shown that nowadays in the world construction science this area is rapidly developing and improving. Thus machine learning algorithms solve problems of prediction of design parameters, problems of identification of these or those parameters, defects, damages on the basis of classification algorithms and others. The materials presented in the article will allow specialists to choose the subject area of research more precisely and determine the directions of adaptation and improvement of their own developments in the field of machine learning.

Keywords: machine learning, reinforced concrete structures, regression equations, identification, approximation, artificial intelligence

One of the actual problems in the field of analysing loads and impacts on bearing structures is their identification. It means the point of application, the type of action and its intensity in cases where there is an impact result, but the parameters that caused this result are not determined. For example, it is an accident action, as a result of which the structure is deformed and collapsed. The solution of such problems arises when analysing accidents on load-bearing structures in construction, as well as when monitoring the deformed state of structures in time. The paper proposes to use the principles of neural network modelling to solve the problem of identifying the impact in the form of a concentrated force on the example of beam systems. The values of linear and angular nodal displacements at some action are considered as input data to neurons. As an example, the linearly deformable beam of constant stiffness is considered, the material of which is a continuous isotropic medium.

Keywords: neural network, deflections, load-bearing structure, displacement, deformation, identification

Alternative approaches to estimating the stiffness of corrosion-damaged beams taking into account localization of corrosion focus and corrosion damage development of concrete based on the model of V.M. Bondarenko are considered. The following methods of determining displacements of corroded reinforced concrete beams of rectangular cross-section are considered. In the first one, the stiffness of the beam in determining the deflection is considered to be constant. In this case, the corrosion-damaged deflections values may be underestimated due to failure to take into account the actual work of reinforced concrete. In the second, the deflection is calculated considering the height of the concrete compression zone that varies along the length of the beam. For these approaches, stiffness reduction in the beam sections in the presence of corrosion damage to the concrete of the compressed zone is modeled. It is shown that in the presence of corrosion damage, the deflections of the structure can significantly increase, which requires mandatory consideration in the life cycle of load-bearing structures of structures when assessing their mechanical safety.

Keywords: reinforced concrete beams, strength of reinforced concrete elements, bending stiffness, deformability, corrosion damage of concrete, height of compressed zone, deflections.