Ivanov D. Diagnosis of digital devices based on an evolutionary approach

Object: the process of diagnosis, control and verification of digital devices defined on a structural. Methods: the methods and models of technical diagnostics and switching circuits, methods of mathematical simulation, evolutionary computation, and the theory of Boolean functions, finite automata, theory of sets. Objective: to increase the effectiveness of computer-aided diagnosis of large digital device. Theoretical and practical results: it is proposed a solution to actual scientific and practical problem of developing of unified methodology for constructing evolutionary models, methods and algorithms, including parallel, which allow to solve a wide class of problems in diagnosis of large-scale digital devices; in practice, the proposed approach allows us to build modern software simulation and diagnosis systems with required functionality for parallel computing systems of different classes. Scientific novelty: first proposed unified methodology for constructing evolutionary models, methods and algorithms which allows to develop new methods of generating identifying sequences, including parallel, capable of handling large-scale digital devices; further developed evolutionary methods of construction of identifying sequences of digital devices basing on the problem-oriented component of these methods, that enables them to increase the efficiency and to develop new methods; first time introduced the concept of the functions of behavior, which enables to formalize information about the behavior of devices at the structural level and to improve the convergence of evolutionary methods; further developed methods for design of energy-efficient digital devices, which allows to build tests with less heat dissipation and provides the accurate estimates of peak heat dissipation; further developed methods of construction of parallel genetic algorithms on a scheme "master-worker" and "islands", allowing them to be adapted to the modern parallel computing systems of different classes, as well as significantly improve the speed and the quality of search; on the basis of scheme of fault list partitioning it is further developed parallel fault simulation methods of digital devices, which allow significantly improve the performance of such methods; it is first developed a parallel fault simulation method of digital devices for multi-core computers, which perform parallel simulation of groups of faults for each input pattern; further developed an evolutionary approach to the identification of digital devices by applying the simulation annealing, which gives the opportunity to develop new effective methods of construction of identifying sequences. The degree of implementation: scientific and practical results are implemented at JSC "SKTB RTU" and NPMP "Amplitude". Applications: electronic digital industry.