Abstract:
To address the early-stage uncertainties in quantity estimation of prestressed concrete (PSC)
bridges due to limited information, this research introduces a novel approach using multi expression programming (MEP) to develop predictive models for estimating the steel weight
and concrete volume. The study gathered 172 datasets, encompassing eleven input variables
compiled from a range of international research publications i.e., average height of pier Hp,
number of piles Np, average depth of piles Dp, average span length SL, total length of bridge
Lb, width of deck Wd, average height of abutment Ha, total number of spans Ns, total number
of girders Ng, height of girders Hg and number of lanes NL, as well as two output variables: the
weight of reinforcing steel Bs and the volume of concrete Vc. The MEP models produced simple
mathematical expressions to calculate the steel weight and concrete volume for PSC bridges.
The accuracy of these models was evaluated using various performance metrics, including
mean absolute error (MAE), root mean square error (RMSE), Nash-Sutcliffe efficiency (NSE),
R-squared (R2
), Relative squared error (RSE), and Relative Root Mean Square Error (RRMSE).
Additionally, sensitivity and parametric analyses were conducted on the MEP models. The
results demonstrate that MEP-based methods deliver highly accurate predictions,
outperforming the multilinear regression model Vc (R2
avg=0.784) and Bs (R2
avg=0.787) used in
this study. Specifically, the MEP model predicting steel weight Bs (R2
avg= 0.965) showed better
results than the MEP model for concrete volume Vc (R2
avg=0.910) prediction. The study
suggests that the proposed MEP-based formulae offer a quick, reliable, and rational method for
estimating materials in PSC bridges, aiding decision-makers in reducing the time required for
quantity estimations, evaluating alternatives and minimizing the risks associated with over and
underestimating material requirements in preliminary estimates.
