Innovative approach to estimate structural damage using linear regression and K-nearest neighbors machine learning algorithms

Abstract

Conventional structural design methodologies often utilize elastic analysis techniques, such as the equivalent static force method and the response spectrum method. While these methods are known for their simplicity and computational efficiency, they prove inadequate in capturing the extent of structural damage caused by seismic forces. Additionally, employing nonlinear dynamic analysis to estimate structural damage represents a challenging and intricate task, posing difficulties for many structural designers. Consequently, the objective of this paper is to present an innovative methodology for evaluating seismic structural damage of moment-resisting frame structures. This involves the utilization of machine learning algorithms, which have been trained and tested on a large data set generated using a newly developed and numerically efficient simulation procedure. The machine learning algorithms employ both linear regression and K-nearest neighbors approaches to accurately replicate the Park-Ang structural damage index.

Keywords

• Machine learning algorithms
• Nonlinear dynamic analysis
• Seismic structural damage
• Time efficient seismic simulations