Maciej Glowacki

On Efficiency Increase of Evolutionary Algorithms for Large Non-linear Constrained Optimization Problems with Applications to Mechanics

Advances in development of highly efficient dedicated Evolutionary Algorithms (EA) for a wide class of large non-linear constrained optimization problems are considered in this paper. The first objective of this general research is development and application of the improved EA to residual stress analysis in railroad rails and vehicle wheels. However, the standard EA are not sufficiently efficient for solving such large optimization problems. Therefore, our current research is mostly focused on development of various new very efficient acceleration techniques proposed, including smoothing and balancing, adaptive step-by-step mesh refinement, as well as a’posteriori error analysis and related techniques. This paper presents an efficiency analysis of chosen speed-up techniques using several simple but demanding benchmark problems, including residual stress analysis in elastic-plastic bodies under cyclic loadings. Preliminary results obtained for numerical tests are encouraging and show a clear possibility of practical application of the improved EA to large optimization problems.

On dedicated evolutionary algorithms for large non-linear constrained optimization problems

This paper considers advances in development of dedicated Evolutionary Algorithms (EA) for efficiently solving large, non-linear, constrained optimization problems. The EA are precisely understood here as decimal-coded Genetic Algorithms consisting of three operators: selection, crossover and mutation, followed by several newly developed calculation speed-up techniques based on simple concepts. These techniques include: solution smoothing and balancing, a--posteriori solution error analysis and related techniques, non-standard use of distributed and parallel calculations, and adaptive step-by-step mesh refinement. Efficiency of the techniques proposed here has been evaluated using several benchmark problems e.g. residual stresses analysis in chosen elastic-plastic bodies under cyclic loadings. These preliminary tests indicate significant acceleration of the large optimization processes involved. The final objective of our research is development of an algorithm efficient enough for solving real, large engineering problems.

On Development of a New Approach for EA Acceleration in Chosen Large Optimization Problems of Mechanics

In this paper we briefly discuss new advances in development of an efficient approach based on Evolutionary Algorithms (EA) for solving a wide class of large, non-linear, constrained optimization problems. Two important applications to engineering mechanics are intended, namely residual stress analysis in railroad rails, and vehicle wheels, as well as a wide class of problems resulting from the Physically Based Approximation of experimental data. However, the primary objective of our long-term research is to obtain significant acceleration of the EA applied to large optimization problems, and to provide ability to solve problems when the standard EA fail. The efficiency of new speed-up techniques proposed was examined using several simple but demanding benchmark problems of computational mechanics. Results obtained so far indicate possibility of practical application of the new approach to real large engineering problems.