Simplified support reaction profiles of RC beams under low-velocity impact: From experimental observations to data-driven prediction

Abstract

The vulnerability of engineering structures to extreme dynamic loads necessitates reliable understanding and prediction of impact performance. During impact events, support reactions directly indicate internal force states within reinforced concrete (RC) beams, yet accurately quantifying these local responses remains challenging due to the transient complexities of impact mechanics. Consequently, developing a predictive reaction force model is essential for systematic and simplified assessment of RC beam impact behavior. This study examined representative research on the dynamic reaction force distribution of RC beams under impact loading. Drop-weight impact tests were conducted considering four key parameters: impact velocity, drop mass, longitudinal reinforcement ratio, and structural size. By combining new test results with published data, the effects of multiple factors on support reactions were summarized. A simplified reaction force profile model involving nine influential variables was then proposed. To support data-driven modeling, a database of over 300 sets of experimental and simulated data was established. Two approaches were explored: explicit multiple linear regression to derive prediction equations for key inflection points and a machine learning method to evaluate factor importance and predict complete reaction force histories. Predicted profiles agreed well with test results, verifying the feasibility of both methods. The proposed model enables practical evaluation of local impact performance and provides a direct reference for structural design.