Non-linear modeling parameters for new construction RC columns

Modeling parameters (MP) of reinforced concrete columns are a critical component of performance-based seismic assessment methodologies because in these approaches damage is estimated based on element deformations calculated using non-linear models. To ensure model fidelity and consistency of assessment results, performance-based seismic assessment methods in ASCE 41, ACI 369.1, and ACI 374.3R prescribe modeling parameters calibrated using experimental data. This paper introduces a new set of equations to calculate reinforced concrete column non-linear modeling parameters optimized for design verification of new buildings using response history analysis. Unlike modeling parameters provided in ACI 369.1 and ASCE 41, intended for columns of older non-ductile buildings, the equations for modeling parameters a_nl and b_nl presented in this study were calibrated to simulate the load-deformation envelope of reinforced concrete columns that meet the detailing requirements of modern seismic design codes. Specifically, the proposed equations are intended for use with provisions in ACI 374.3R, Chapter 18 and Appendix A of ACI 318-19 and Chapter 16 of ASCE/SEI 7-16. The proposed equations were calibrated using the ACI Committee 369 column database, which includes column configuration parameters, material properties, and deformation capacity modeling parameters inferred from the measured response of columns under load reversals. Dimension reduction techniques were applied to visualize different clusters of data in 2D space using the negative log-likelihood score. This technique allowed decreasing the non-linearity of the problem by identifying a subset of columns with load-deformation behavior representative of new construction conforming to current codes requirements. A Neural Network model (NN) was calibrated and used to perform parametric variations to identify the most relevant input parameters and characterize their effect on modeling parameters, and to stablish the degree of non-linearity between each input variable and the model output. Developing equations for modeling parameters applicable to a wide range of columns is challenging, so this research considered subsets of the database representative of new construction columns to calibrate simple practical equations. Linear regression models including the most relevant features from the parametric study were calibrated for rectangular and circular columns. The proposed linear regression equations were found to provide better estimates of new construction column modeling parameters than the available tables in ACI 374.3R and ASCE 41-13, and the equations ASCE 41-17.