A comparison between advanced hybrid machine learning algorithms and empirical equations applied to abutment scour depth prediction

Complex vortex flow patterns around bridge piers, especially during floods, cause scour process that can result in the failure of foundations. Abutment is a complex three-dimensional phenomenon difficult to predict with traditional formulas obtained using empirical approaches such as regressions. This paper presents test standalone Kstar model five novel hybrid algorithm bagging (BA-Kstar), dagging (DA-Kstar), random committee (RC-Kstar), subspace (RS-Kstar), and weighted instance handler wrapper (WIHW-Kstar) depth (ds) for clear water condition. The dataset consists 99 data from flume experiments (Dey Barbhuiya, 2005) abutment shapes vertical, semicircular 45° wing. Four dimensionless parameter relative (h/l), excess Froude number (Fe), sediment size (d50/l) submergence (d50/h) were considered prediction (ds/l). A portion was used calibration (70%), remaining validation. Pearson correlation coefficients helped deciding relevance input parameters combination finally four different combinations used. performance models assessed visually quantitative metrics. Overall, best vertical shape Fe, d50/l h/l, while wing Fe most effective combination. Our results show incorporating h/l lead higher involving d50/h reduced power more error. WIHW-Kstar provided highest RC-Kstar outperform other