Shear strength assessment of self-compacting concrete beams using lasso regression technique

Existing code provisions for shear strength prediction of self-compacting concrete (SCC) beams have often fallen short of its degree of predictability in relation to experimental responses. The research study seeks to develop a model that better predicts the shear capacity of self-compacted concrete beams without shear reinforcement. In addition, the critical parameters that influence the shear strength of an SCC beam was also investigated by using varying regression techniques (Linear, Stepwise, Lasso, Ridge and Elastic Net regressions). A pooled database having a total of 179 SCC beams without shear reinforcement was compiled for the analysis. The Lasso regression was the most effective from statistical analysis having the least relative and mean squared errors. In comparison with existing codes: ACI 318-08, AASHTOLRFD Bridge Design Specification-2007, Eurocode 2 and BS8110, the Lasso model performed better with least mean percentage error (12.23%), least average safety factor(1.1012) and the least coefficient of variation(0.159). The Lasso model also showed that compressive strength, height, breadth, depth of beam, shear span to depth ratio, longitudinal reinforcement ratio, maximum aggregate size and fine to coarse aggregate ratio were all relevant parameters in shear strength prediction of SCC beams without stirrups.