Page 91 - AJWEP-22-6

ML-based C for side trapezoidal labyrinth weirs
d
A B

Figure 6. Performance assessment of the machine learning models’ outcome using the DDR index. (A) Training
phase. (B) Testing phase.
Abbreviations: GEP: Gene expression programming; MARS: Multivariate adaptive regression splines;
MLP: Multilayer perception; SVM: Support vector machine.

overdesigns or underdesigns and save construction four distinct models – SVM, GEP, MLP, and MARS
costs. They also allow proper overtopping from dams – to anticipate the C of TALW. The assessment of
d
and prevent or mitigate dam failures in extreme weather performance evaluation metrics, while affirming the
conditions. Climate change has significantly enhanced capabilities of the quartet mentioned above, revealed
the chances of extreme weather conditions; therefore, the preeminence of the MLP model in contrast to its
some of the hydraulic structures made for usual weather three counterparts. This ascendancy was consistently
conditions might not withstand properly. Therefore, evident across both the training and testing phases. The
having an accurately predicted labyrinth weir in such outcomes underscored that the SVM, GEP, and MARS
conditions would be beneficial. models secured successive positions as commendable
Ecosystem protection is another attribute of the simulators following the MLP model.
properly designed labyrinth weir. Downstream soil While this study provides robust MLMs for predicting
erosion and excessive sediment transportation can be C in TALW under controlled laboratory conditions,
d
mitigated through the accurate prediction of C . These translating these findings to field applications faces
d
advantages are aligned with the Sustainable Development inherent limitations. Laboratory models assume optimal
Goals (SDGs) – SDG6: clean water and sanitation, structural integrity, whereas field structures are subjected
SDG9: industry, innovation, and infrastructure, SDG to factors such as concrete erosion, joint displacement,
11: sustainable cities and communities, SDG 13: and corrosion over time, which may alter their hydraulic
climate action, and SDG 15: life on land. Therefore, the performance. Furthermore, reservoir sedimentation
research presented here has a greater practical potential and organic debris accumulation – both absent in lab
in achieving the SDGs. Furthermore, the impact on settings – can reduce effective crest length and amplify
developing countries would be significant. Labyrinth local head losses, systematically biasing C . Field
d
weirs are more cost-effective compared to traditional environments also introduce hydrologic uncertainties,
spillways and offer greater benefits to rural communities such as stage measurement errors (e.g., gauge drift
facing water management challenges. and thermal effects) and unsteady flow regimens not
replicated in steady-state experiments. Critically,
4. Conclusion neglected maintenance (e.g., spillway cracking and
gate misalignment) may induce unexpected flow
Given the pivotal significance of C in the operational behavior, as evidenced by real-world failures like the
d
efficacy of weirs, alongside the advancing application Oroville Dam incident in 2017, where operational
52
of data-driven models for predicting this coefficient, stresses exceeded design assumptions. These real-world
the avenue for forthcoming research in the domain of complexities necessitate field validation of the models
TALW is explored. The present research employed and the integration of probabilistic safety margins in

Volume 22 Issue 6 (2025) 85 doi: 10.36922/AJWEP025120081