Data Driven Tools for Dynamic Analysis of Building Frames

Abstract

The determination of the fundamental period of vibration of a structure is essential for earthquake design. Current BIS code equations (IS 1893 (Part1): 2002) provide formulae for the approximate period of earthquake-resistant building systems, which are dependent only on the height and base dimensions of the building. Such a formulation may be overly conservative and unable to account for structures with geometric irregularities. This study estimatesthe fundamental period of three different types of reinforced concrete (RCC) earthquake-resistant building structures: symmetric special moment resisting frames (MRF), shear wall dominant symmetric RCC buildings and shear wall dominant unsymmetrical RCC buildings.The data driven techniques: Artificial Neural Networks (ANN), Genetic Programming (GP) and Wavelet Neural Network (WNN) are explored to estimate the period of vibration, base shear force and top floor displacement of buildings. Data-driven modeling can be considered as an approach to modeling that focuses on using the Machine Learning methods in building models that would complement or replace the “knowledge-driven” models describing physical behavior (Solomatine and Ostfeld, 2008). These techniques require the data in the form of input and output parameters. ANN and WNN techniques give the results in the form of values of parameters of interest whereas the GP technique yields equation as a result. In order to check the adequacy of the employed techniques, these techniques are applied for performing the dynamic analysis of 206 symmetric MRF buildings. The statistical measures used to assess the accuracy of the developed models are Root Mean Squared Error (RMSE), Correlation Coefficient (r) and Coefficient of Efficiency (CE).Once the applicability of the techniques is ascertained for MRF buildings, these are applied to other types of buildings. Total 52moment resisting frames, 70 shear wall dominant symmetric buildings and 55 shear wall dominant unsymmetrical buildings are analyzed with ETABS v.9.7.2. The fundamental period based on vibration theory for each case is compared with estimated value by data driven tools including current code equations as well as equations proposed in recent literature. Based on the results obtained from the ANN, GP and WNN techniquesempirical equations for the approximate fundamental periods are suggested for aforementioned buildings. The proposed equations of period of vibration are validated with measured period of vibrations cited in the literature. The design base shear force obtained from the response spectrum recommended by IS 1893 (Part1):2002 is found to be much higher than the actual base shear force in case of shear wall dominant unsymmetrical buildings. The safety margin so presumed is reduced in some cases by proposed modification in the response spectrum recommended by IS 1893 (part1):2002. The same is also validated using the response spectrum suggested for Mumbai city by Desai and Chaudhary, 2014. A linear relationship between the top floor displacement of the building and shorter period of vibration. This relation is applicable for peak ground acceleration (PGA) values between 0.10g and 0.16g. It can be said that the data driven tools of ANN, WNN and GP can be applied for dynamic analysis of buildings. Additionally the period equation for the buildings proposed in the present study by these tools can be used in practice as evident from closeness of the results with measured period values.