The Integrated Computational Environment for Optimization of Complex Systems

Abstract

The concept, general structure and main components of the integrated computing environment (ICE) for optimization of complex systems based on the methods of conditional minimization of the objective functional characterizing the initial formulation of the problem of mathematical modeling on supercomputers are considered. The inverse problem is solved based on a sequence of direct initial-boundary value problems, each of which can be a resource-intensive interdisciplinary problem described by a system of multidimensional differential and/or integral (and/or discrete) equations in multi-scale computational domains with multi-connected piecewise smooth geometric configurations of boundaries and contrast material properties of multiphase media. This includes, in particular, the actual problems of surrogate optimization. The functional core of the ICE is focused on a high-performance support of all major stages of a large-scale computational experiment, including geometric and semantic modeling, generation of computational grids, approximation of initial equations, solution of algebraic problems, optimization algorithms for solving inverse problems, post-processing and visualization of results, decision-making on the results of the study. Technical requirements for the ICE involve a flexible expansion of the models and methods used, adaptation to the evolution of computer platforms, effective reuse of external software products, coordinated participation of different teams of developers and focus on a broad demand of the end users from different industries. These qualities are designed to provide a long life cycle and a qualitative increase in productivity in the development and application of the new generation software.