NEURAL NETWORK APPLICATION FOR SCHEDULING OF BUILDING CONSTRACTION REPAIR

摘要

This article discusses the application of mathematical simulation for planning the timing of repair of building structures according to their degree of physical deterioration. For the current study, 296 apartment buildings were selected from the housing stock in Archangelsk city, using neural network Self-Organizing Maps (SOMs). Optimal clustering of selected housing was achieved using special Deductor platforms to define 16 housing clusters with maximum difference but high internal similarity. For more detailed scheduling of building overhaul and running repairs, selected clusters were divided into 4 model-groups that needed the same repairs in similar time frames. Integration of the particular characteristics of each group created a framework for calculating the capital investments for each model group, with specific payments identified for each cluster.