Preliminary design of double layer grids using ANN

Abstract

With the industrialization and development of the world there is a demand for efficient and adaptable long-span structures. The Double Layer Grid (DLG) is a valuable tool for the architect or engineer in the search for new forms, due to their wide variety and flexibility. DLG is the truss spanning in three directions to cover large spans without intermediate support. Except the planner beams and trusses all other structures are 3D in sense as they have length, depth and thickness. Beams and trusses are predominantly 2D in their actions as they effectively resist loads applied only in one direction between their supports. But from stability point of view third dimension is also important. Hence 3D action is considered. Groups of DLG were made according to the span range from 25 meter to 100 meter with the two different sets of aspect ratio Height of DLGSpan Length (Height of DLG) / (Span length) = H/L Viz- 0.035 to 0.065 and 0.05 to 0.08 with interval of 0.01. The initial analysis and design work of DLG is carried out on STAAD Pro as per IS 800:1984. No of Neural Networks are trained in MATLAB and one networks which gives best results among all is chosen. The results of Artificial Neural Network (ANN) and STAAD for same conditions were analyzed.