APPLICATION OF NEURAL NETWOKS IN DESIGN OF RCC COLUMNS

Abstract

In this project work application of artificial neural network in rcc column design have been implemented. As a structural engineer we are looking forward for an optimal solution for problems encountered so here we design rcc column by using an application of artificial neural networks which provide solution with decreased computational time and increased efficiency. Recent advancement in the field of artificial neural networking has paved way for optimization of complex and tedious design processin the field of structural engineering. The design of column cross-section for given axial load and biaxial moments is done, firstly by preassuming the dimensions of column section,steel reinforcement, grade of steel and grade of steel and then check its adequacy which is a complex trial and error process we tryed to map this complicated design procedure using artificial neural networking. Artificial neural networks are algorithms for cognitive tasks such as optimization and learning.These algorithms are developed with capability to learn and generalize from training example data set presented to them with no knowledge of rules. Artificial neural networks are group of numerical learning technique. This whole computational model is made up of many interconnected non-linear calculation units called neurons. A multilayer perceptron and levenberg-marquardt algorithm for training the neural network have been used. The example data set for training of neural network is generated using programming in Microsoft excel. rcc column design is done based on design procedure explained in sp 16 as per IS 456:2000. 5271 column design are done to generate data for training the network. Matlab software is used to develop the neural network architecture used for column design. The effect of performance of various parameters of network on network output was studied and necessary modification was made to get desired target output. Network performance is checked by performance function vs epoch curve and by compairing network output with the result obtained through conventional design. Thus best topology of network architecture is achieved for the best performance of network in function fitting application of column design.