TY - GEN
T1 - Data-Driven Approaches for Estimation of Particle Froude Number in a Sewer System
AU - Shakya, Deepti
AU - Agarwal, Mayank
AU - Deshpande, Vishal
AU - Kumar, Bimlesh
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2023/1/1
Y1 - 2023/1/1
N2 - The deposition of sediments has a major impact on the hydraulic capacity of a channel. The deposition of sediment over time may result in a diminished ability of the sewers to carry waste and other materials. For designing sewers and drainage systems in urban areas, the self-cleansing mechanism is often deployed. In this study, the accurate prediction of the particle Froude number (Fr) plays a significant role. This study investigates the insights of the performance of the data-driven approaches to determine the particle Fr with regard to non-deposition with a deposited bed. The dataset obtained comprises variety of studies published in literature having a range of values for volumetric sediment concentration (Cv), the dimensionless grain size of particles (Dgr), the median size of sediment (d), hydraulic radius (R), and the friction factor of the pipe (λ). Three data-driven approaches, namely Random Forest (RF), M5Prime (M5P), and Reduced Error Pruning Tree (REPT), have been utilized in this study for modeling purposes. Results show that the RF approach is superior in comparison with M5P and REPT approaches. The best performing method in this study is RF (CC = 0.966, NSE = 0.932, RMSE = 0.64, and R2 = 0.933) followed by M5P, REPT.
AB - The deposition of sediments has a major impact on the hydraulic capacity of a channel. The deposition of sediment over time may result in a diminished ability of the sewers to carry waste and other materials. For designing sewers and drainage systems in urban areas, the self-cleansing mechanism is often deployed. In this study, the accurate prediction of the particle Froude number (Fr) plays a significant role. This study investigates the insights of the performance of the data-driven approaches to determine the particle Fr with regard to non-deposition with a deposited bed. The dataset obtained comprises variety of studies published in literature having a range of values for volumetric sediment concentration (Cv), the dimensionless grain size of particles (Dgr), the median size of sediment (d), hydraulic radius (R), and the friction factor of the pipe (λ). Three data-driven approaches, namely Random Forest (RF), M5Prime (M5P), and Reduced Error Pruning Tree (REPT), have been utilized in this study for modeling purposes. Results show that the RF approach is superior in comparison with M5P and REPT approaches. The best performing method in this study is RF (CC = 0.966, NSE = 0.932, RMSE = 0.64, and R2 = 0.933) followed by M5P, REPT.
KW - Data-driven
KW - Froude number
KW - Machine learning
KW - Sedimentation
KW - Sewer
UR - https://www.scopus.com/pages/publications/85172262081
U2 - 10.1007/978-981-99-1901-7_47
DO - 10.1007/978-981-99-1901-7_47
M3 - Conference contribution
AN - SCOPUS:85172262081
SN - 9789819919000
T3 - Lecture Notes in Civil Engineering
SP - 583
EP - 593
BT - Geospatial and Soft Computing Techniques - Proceedings of 26th International Conference on Hydraulics, Water Resources and Coastal Engineering HYDRO 2021
A2 - Timbadiya, P.V.
A2 - Patel, P.L.
A2 - Singh, Vijay P.
A2 - Mirajkar, A.B.
PB - Springer Science and Business Media Deutschland GmbH
T2 - Proceedings of the26th International Conference on Hydraulics, Water Resources and Coastal Engineering , HYDRO 2021
Y2 - 23 December 2021 through 25 December 2021
ER -