DYNAMIC RESPONSE ANALYSIS OF STRUCTURES THROUGH ARTIFICIAL NEURAL NETWORKS

Abstract

Engineers and technocrats face • many problems in their respective fields of interest, which are highly nonlinear, difficult to formulate and if formulated take very long time. Recently Artificial Neural Networks have emerged as alternative methodology and attempts have been made to solve many problems that are difficult to formulate, with the help of these Artificial Neural Networks. The encouraging results obtained by those attempts have prompted this dissertation work. The Artificial Neural Networks may be defined as parallel, distributed information processing structures consisting of processing elements, which can possess a local memory and can carry out localized information processing operations, interconnected via unidirectional signal channels called connections. Through this dissertation work, a neural network model using feed forward, error back propagation network with 10 input nodes, 8 hidden nodes and two output nodes, for the dynamic response analysis of a prefabricated, jointed, multipanel shear wall under dynamic loading (under various earthquakes) has been developed. This network has been trained with the acceleration spectrogram data for Elcentro x 1.0 (both Transverse and Vertical compo'nents) and the corresponding displacement response (both horizontal and vertical displacements) at the right hand top corner of a 12 storeyed, 36m high, 9m wide shear wall. For this training a learning rate equal to 0.002 and a momentum factor equal to 0.7 have been used. The momentum factor was gradually reduced to 0.