Оцінка зміни технічного стану понівечених багатоповерхових будівель за допомогою штучного інтелекту

The article presents the results of the analysis and prospects for applying information technologies to select effective organizational, technological, and technical solutions for eliminating the emergency destruction of multi-story buildings damaged as a result of Russian aggression. Information and mathematical modeling is considered a key tool for assessment and decision-making, especially under conditions of limited access to the research object and a lack of information about its technical condition. The subject of this research is a previously unstudied issue: the development of a method for the urgent stability assessment of damaged multi-story buildings (DMBs) amid large-scale surveys, which will reduce the time for inspection, modeling, and decision-making regarding the reinforcement and reconstruction of a DMB or its dismantling. An important aspect of this method is forecasting the technical condition of DMBs using modern digital elements of artificial intelligence–neural networks. Optimizing the decision-making process under uncertainty is possible with the prior development of standard organizational and technological anti-emergency measures and methods for their application to typical DMB objects. Linking existing, pre-developed solutions using information and mathematical models of typical objects to a specific emergency DMB based on the pattern recognition principle allows for accelerating the selection of an option and ensuring the conduct of emergency rescue operations. In turn, this will help rescue potential victims, prevent accidents, and become part of the emergency response plan. In the post-war period, the use of the presented methodology will allow for the rapid assessment and forecasting of the technical condition of DMBs and the selection of an optimal strategy for their stabilization and reconstruction, including the frequency of monitoring needs and repair timelines. The application of neural networks, particularly hybrid models (CNNs, LSTMs, autoencoders), opens fundamentally new opportunities for shifting from a reactive to a proactive approach in assessing the technical condition of protective engineering structures (PES). The implementation of such technologies will enable automation of damage analysis, forecasting of damage progression, and the generation of well-grounded recommendations for repair, reinforcement, or demolition of structures. This significantly enhances the efficiency of decision-making and reduces risks for rescuers and engineers.