The Research Of Computing Anomaly In Evolutionary Structure Optimization

As an independent discipline, optimal design is playing an important role in engineeringfield. However, traditional optimization technology no longer has enough effect because it hasno capability of changing topology class of structure. This can be achieved by topologyoptimization perfectly but some anomalous phenomenon of computing still exists such asmesh-dependent, checkerboard pattern and local optimal optimization. After principleintroduction of evolutionary structural optimization, some methodologies are put forward tosuppress computing anomaly and applied into dynamic problems including frequenciesoptimization and harmonic response optimization.Sensitivity numbers are derived after principles of evolutionary structural optimizationare introduced under consideration of design dependent load. An improved evolutionary flowusing uniform ordering for sensitivity number is applied in bi-directional evolutionaryoptimization after simple introduction of three common ways of evolutionary topology.Perimeter constraint method utilizing combination sensitivity number is proposed basedon optimal condition after interpretation to reason causing mesh dependent. Filter computingis improved to suppression checkerboard in a more general form after its nature is discussed.Complete analysis is used because of principle defects of evolutionary structuraloptimization according to reasons for local optima. Adaptive growing method is introduced togenerate initial structure as well as decrease evolutionary time. Continuum evolution waysincluding methods of continuum perimeter method and improved continuum filter method areproposed to resolve non-convex topology optimization and get global optima.Sensitivity numbers for damped frequencies and amplitude of harmonic response isderived to resolve dynamic problems including damped frequencies optimization andamplitude optimization of harmonic response using improved evolutionary structureoptimization. In addition, a continuum objective function is introduced with its sensitivitynumber derived.Some numerical examples are implemented using methods proposed in this paper, whichshows that methodology proposed here can improve structural performance, modifycomputing abnormalities and be implemented conveniently.