Work number - M 4 FILED
Presented Lviv Polytechnic National University
Authors: Ph.D. Izonin I.V., Ph.D. Skorokhoda O.V., Ph.D. Teslyuk T.V., Lenko V.S.
The aim of scientific work is to increase the efficiency of methods of synthesis of intellectual components and neural network data processing, create models of dynamic analysis for the processes of “smart” buildings and enterprises components functioning.
Methods of construction the tools for data collection and processing, seamless data exchange within smart home and enterprise information systems have been developed. They facilitate the formalization of synthesis processes and the development of intellectual components that meet high technical and economic indicators. Models of dynamic analysis for the processes of “smart” buildings and enterprises components functioning have been developed, which facilitate the verification of the correspondence between components behaviour and their specification, thus reducing the time and cost of development.
A number of neural networks methods for image pre-processing (change image resolution, segmentation, and reconstruction) that operate in computer vision sub-systems of smart home and enterprise have been developed and patented, along with the hardware implementations of one- and two-dimensional median filtration devices that minimize computational resources, facilitate on-line data processing and expand the available functionality comparing to the domestic and foreign analogues.
New models of a formal neuron and a method of parallel-vertical data processing in neural networks have been developed. The method provides an orientation of models for VLSI implementation and adaptation of an artificial neuron to the requirements of a specific application. As a result, the VLIS-oriented neuro-elements of a parallel-vertical type that ensure high efficiency of hardware utilization are synthesized.
Several tools for smart home and enterprise systems development have been researched and synthesized, and their effectiveness has been confirmed experimentally. An intellectual component that is based on the artificial neural networks and implemented in the form of a neuro controller enables processing of fuzzy and unstructured inputs within the information systems of smart home and enterprise. Models of components for smart home and enterprise systems have been constructed based on the theory of Petri nets, which allows studying the dynamics of the processes they are involved in, determine the attainable states of the synthesized component, the presence or absence of deadlocks – important components of system design.
The results of the work are used at the department of methods and systems of remote sensing in Karpenko Physico-Mechanical Institute of the National Academy of Sciences of Ukraine, LLC “Clinic of modern diagnostics methods and innovative medical technologies “Aquarius”, LLC “ELSI” and Lviv Polytechnic National University.
Number of publications: 168, incl. 3 monographs and 3 sections of collective monographs, 1 manual, 72 articles (14 articles in periodicals indexed by scientific databases Scopus, Web of Science, 20 - in foreign editions), 70 conference papers. According to the Scopus database, the total number of references to authors’ publications is 389, h-index (based on work) = 11; according to the Google Scholar database, the total number of links is 519, h-index (based on work) = 13; According to the Web of Science database, the total number of links is 67, h-index (at work) = 5. The novelty and competitiveness of the developed technical solutions are protected by 5 patents for inventions of Ukraine, 7 patents for utility models and 7 copyright certificates.