Sulimov

Modern methods for optimization investigation of complex systems are based on development and updating the systems’ mathematical models in connection with solving the corresponding inverse problems. The optimization approach is one of the main approaches to solving the inverse problems. In the main case it is necessary to find a global extremum of not everywhere differentiable criterion function. When the number of variables is large they use the stochastic global optimization algorithms. As stochastic algorithms yield too expensive solutions, so this drawback restricts their applications. Developing hybrid algorithms that combine a stochastic algorithm for scanning the variable space with deterministic local search method is a promising way. A new hybrid algorithm that integrates a multiple Metropolis algorithm and the Hooke–Jeeves method for the local search is proposed. Some results on solving the global optimization benchmark are presented.

This article deals with the global optimization issues on the computational diagnostics of hydromechanical systems. The criterion functions are assumed to be continuous, Lipschitzian, multiextremal, and not always differentiable. The article proposes two novel hybrid algorithms with scanning a search space by the stochastic Multi-Particle Collision Algorithm that uses the analogy to the absorption and scattering processes for nuclear particles. The local search is implemented using the hyperbolic smoothing function method for the first algorithm, and the linearization method with two-parametric smoothing approximations of criteria for the second one. There are some results on the model solution of computational diagnostics of the coolant phase composition in the reactor primary circuit.

This article deals with problems of global optimization, model updating and diagnostics of hydro-mechanical systems. It was assumed that criterion functions were continuous, Lipschitzian, multiextremal and not always differentiable. Two novel hybrid algorithms were proposed; these algorithms use the modern stochastic Particle Collision Algorithm based on an analogy with absorption and scattering processes for nuclear particles, for scanning a search space. Local search was implemented with the use of a linearization method with smoothing approximations of criteria for the first algorithm, and a convergent variant of the Nelder-Mead simplex method was used for the second algorithm. Some results on solving model problems on computational diagnostics of the coolant phase constitution in the reactor primary circuit were presented.

This article deals with problems of global optimization and computational diagnostics of hydro-mechanical systems. It was assumed that criterion functions were continuous, Lipschitzian, multiextremal and incompletely differentiable. Novel hybrid algorithms were proposed; these algorithms use the modern stochastic Particle Collision Algorithm based on an analogy with absorption and scattering processes for nuclear particles, for scanning a search space. Local search was implemented with the use of two derivative-free methods: a space filling curve method and a convergent variant of the Nelder-Mead simplex method. Results of solving the model problem in identification of the coolant phase constitution anomalies in the reactor circuit were presented.

Novel hybrid multiobjective optimization algorithms are presented for solving problems of computational diagnostics. A vectorial variant of the linearization method is implemented. Global solutions for individual criteria are determined by use of hybrid algorithms that combine the Metropolis algorithm scanning the space of variables and deterministic methods for local search. The vector optimization algorithms generate a set of non-dominated solutions to approximate a Pareto-optimal front. Simulation results for one of the benchmarks and corresponding valuations of the algorithm computational efficiency are received. The proposed hybrid algorithms can be applicable for computational diagnostics systems, problems of intellectual models teaching, complex dynamic systems control, and other intellectual technologies.