Pasko R. Intellectual Support System Forensic Building-Technical Expertise

The dissertation is devoted to the solution of an important scientific and practical problem of creation of computerized Support System Forensic BuildingTechnical Expertise on definition of the reasons of destruction and deformations of buildings and constructions that are under construction and operated on subsidence loess soils. The conceptual model of Specialized Intelligent System with integrated neurofuzzy inference system is proposed. The work of the system is based on the use of generalized expert experience and an artificial neural network of the Cascade ARTMAP category. The architecture and the process of forming the knowledge base of the system are described. This system is part of a software package designed to support forensic building-technical expertise and expert research on compliance with building codes, rules and phenomena and is designed to assess the impact of certain environmental factors on the technical condition of forensic construction. It has proposed to use the Fuzzy Logic Toolbox package of the MATLAB environment to form an a priori knowledge base of the artificial neuro-fuzzy Cascade ARTMAP model. Systematized and presented in a form acceptable for processing by computerized systems, the main input data, based on which fuzzy output has implemented. When processing the input variables that characterize the signs of deterioration of the technical condition of objects, preference has given to such reductions that take into account the unity of presentation and simplify the interpretation of data in information exchange between experts and forensic support system through the formation of associations. The expediency of application of associative rules at formalization of heuristic activity of experts has substantiated. To formalize the expert's opinion, it has proposed to use a max-min composition. When formalizing the process of forensic construction and technical examinations, the main attention has focused on taking into account environmental factors that can lead to wetting of the forest soil under the foundation of the object at the stage of operation. The implementation of the system will provide an opportunity to automate the process of supporting forensic building-technical expertise are characterized by nonmonotonic logic, as well as reduce the risk of making wrong decisions by using the system as an intelligent database. Keywords: artificial neural network, associative rule, expert opinion, fuzzy inference, knowledgebase, non-monotonic logic, technical condition, subsidence loess soil