engleski

Timber Structural Design Based on Neural Networks Application nad FE 3D Modelling

This paper represents the efficiency of trained Neural Networks in producing useful solutions to the structural design assignments and possibility of their application. The experiment in timber structural environment could be also applied on other materials or different typical entities. Trained NN was used in design of 3D glued laminated structural forms with main-girders of various special geometric shapes. The design of these structural forms is based on the same type of analysis and repetitive engineering procedure that gives sense to their parametric modelling, as the most suitable manner to describe them. FE analysis of 3D parametric modelling structures (using COSMOS/M program) gave results relevant for evaluation of states of safety, serviceability and stability of whole structure and its corresponding elements. The most important fact is that inputs and outputs of FE analysis were recognisable for generated Neural Networks that have been trained on them.

NN ; FEA ; 3D-parametric modelling ; timber structures design ; EC5

nije evidentirano