Original Research
Multigene genetic programming for sediment transport modeling in sewers for conditions of non-deposition with a bed deposit

https://doi.org/10.1016/j.ijsrc.2018.04.007Get rights and content

Highlights

  • Conditions of non-deposition with a bed deposit are considered for design of large sewers.

  • A multigene genetic programming (MGGP) technique is used for modeling.

  • Models are established using experimental laboratory data.

  • A new Pareto-optimal MGGP model is developed to estimate particle Froude number.

  • MGGP models outperform all conventional regression models in the literature.

Abstract

It is known that construction of large sewers based on consideration of flow with non-deposition without a bed deposit is not economical. Sewer design based on consideration of flow with non-deposition with a bed deposit reduces channel bed slope and construction cost in which the presence of a small depth of sediment deposition on the bed increases the sediment transport capacity of the flow. This paper suggests a new Pareto-optimal model developed by the multigene genetic programming (MGGP) technique to estimate particle Froude number (Frp) in large sewers with conditions of sediment deposition on the bed. To this end, four data sets including wide ranges of sediment size and concentration, deposit thickness, and pipe size are used. On the basis of different statistical performance indices, the efficiency of the proposed Pareto-optimal MGGP model is compared to those of the best MGGP model developed in the current study as well as the conventional regression models available in the literature. The results indicate the higher efficiency of the MGGP-based models for Frp estimation in the case of no additional deposition onto a bed with a sediment deposit. Inasmuch as the Pareto-optimal MGGP model utilizes a lower number of input parameters to yield comparatively higher performance than the conventional regression models, it can be used as a parsimonious model for self-cleansing design of large sewers in practice.

Introduction

Sewer system design considering sedimentation problems is an essential issue in urban hydrology and hydraulic engineering practice. Deposition of sediment significantly impacts the hydraulic capacity of the sewer systems. In order to prevent such problems, sewers and drainage systems must be designed to be self-cleansing. A self-cleansing channel has capability to convey sediment particles with the flow and remove deposited sediment from the channel bed. There are variety of self-cleansing design criteria such as; incipient motion (Bong et al., 2013, Bong et al., 2016, Novak and Nalluri, 1984, Safari, 2016, Safari et al., 2017a), incipient deposition (Aksoy et al., 2017, Loveless, 1992, Safari et al., 2014, Safari et al., 2015, Safari et al., 2016) and non-deposition (May et al., 1996, Ota and Nalluri, 2003, Ota and Perrusquia, 2013, Safari et al., 2017b). As reported by Ackers et al. (1996), May et al. (1996), and Butler et al. (1996, 2003) the non-deposition concept is the most reliable method for sewer system design. It has to be emphasized that the non-deposition concept can be divided into two criteria of non-deposition without a bed deposit and non-deposition with a bed deposit. In the former, sediment particles are transported with the flow, while in the later, a small depth of deposited sediment exists at the channel bed. The justification of this method comes from the finding of Ab Ghani (1993), May (1993), and Ota and Nalluri (2003) who confirmed that large sewers require higher self-cleansing velocity, and therefore, sewer design based on non-deposition without considering the sediment deposited on the bed considers an uneconomical (steeper) channel bed slope. It is concluded that large sewers must be designed based on non-deposition with a bed deposit criterion.

Self-cleansing models available in the literature mostly are developed using multiple non-linear regression analysis such as: Alvarez (1990), El-Zaemey (1991), Perrusquia (1992), Ab Ghani (1993), and Nalluri et al. (1997). Regression models have a simple structure and are valuable for defining the effective variables in the sediment transport in sewer systems, however, their computational ability is not as high as artificial intelligence (AI) techniques. Recently, application of AI techniques has attracted the interest of many researchers for variety of water resources engineering problems (Chang et al., 2016, Chang and Tsai, 2016, Nourani et al., 2014a, Yaseen et al., 2015). Although, there are a few studies in the literature applying AI techniques to model sediment transport in drainage systems for non-deposition without a bed deposit and incipient deposition conditions (e.g., Ab Ghani & Azamathulla, 2011; Azamathulla et al., 2012; Ebtehaj & Bonakdari, 2014; Mohammadi et al., 2015; Roushangar & Ghasempour, 2016; Safari et al., 2016), to the best of the authors knowledge no study on application of AI techniques has been done for modeling sediment transport in sewers with bed deposits.

Genetic programming (GP) (Koza, 1992) is an AI technique that evolves computer programs to automatically solve problems using Darwinian natural selection. In hydrological applications, GP and its advancements are commonly used to infer the underlying structure of either natural (Al-Juboori and Guven, 2016, Danandeh Mehr et al., 2013, Ghorbani et al., 2010, Kisi et al., 2012, Meshgi et al., 2015, Nourani et al., 2014b, Olyaie et al., 2017) or experimental processes (Khan et al., 2012, Sattar, 2014, Uyumaz et al., 2014, Zahiri et al., 2014, Zahiri et al., 2015). In such applications, GP generates some possible programs (solutions) to identify the process mathematically. When the task is to build an empirical model of data acquired from a set of experiments, GP is known as symbolic regression (Searson, 2015), which is a self-structuring technique without requiring the user to know or specify the form of the solution in advance. It differs from either conventional regression analysis or other AI techniques, in which modelers must specify the structure of the process of the interest.

This study, for the first time, investigates the ability of one of the most recent advancements of the GP technique, namely multigene GP (MGGP, Searson, 2009), for sediment transport modeling in sewers with bed deposit. To this end, four experimental data sets including wide ranges of sediment size and concentration, deposit thickness, and pipe size are used. The new models are compared with conventional regression models available in the literature to evaluate how the MGGP technique is beneficial to formulate sediment transport over bed deposit in sewers.

Section snippets

Sediment transport over a bed deposit in sewers

Self-cleansing design for non-deposition with a bed deposit is recommended for designing large sewers. The primarily studies of sediment transport over a bed deposit were done by Ambrose (1953) and Craven (1953) who found that the bed deposit thickness affects the sediment transport capacity of the pipe channels. After decades, May (1982) and May et al. (1989) started again to investigate the effects of a sediment deposit on sediment transport in sewers. Similar to May et al. (1989), in a

Experimental data

Experimental data for non-deposition conditions with a bed deposit collected by El-Zaemey (1991), Perrusquia, 1992, Perrusquia, 1993, May (1993), and Ab Ghani (1993) are used in this study. All the experiments were done in circular channels with a variety of sizes. El-Zaemey (1991) and Perrusquia, 1992, Perrusquia, 1993 did experiments in pipes having 305 and 225 mm diameters, respectively, while May (1993) and Ab Ghani (1993) did experiments in a larger pipe having a 450 mm diameter. The

Data preparation

The characteristics of flow, fluid, sediment, and channel must be considered for determination of the sediment transport process in sewers. In this study, flow velocity (V), hydraulic radius (R), and gravitational acceleration (g) are the flow characteristics; fluid kinematic viscosity (ν) and density (ρ) are the fluid characteristics; sediment density (ρs), median sediment size (d), and volumetric sediment concentration (Cv) are the sediment characteristics; and channel friction factor (λ) is

Results and discussion

In order to develop the MGGP-based model for sediment transport in sewers for non-deposition with a bed deposit condition, an open-source software platform for symbolic regression in MATLAB®, namely GPTIPS2 (Searson, 2015) is used in the current study. Using the experimental data listed in Table 1 and MGGP setting attributes mentioned earlier, GPTIPS2 was executed. Fig. 2 illustrates the summary of runs that were configured to minimize the error metric (RMSE) over the training data. The upper

Conclusions

The estimation accuracy of an experimental model depends on the data features (type, range) and the regression technique used for model development. Moreover, the complexity of a model is another important criterion since, in general, engineers prefer not only accurate but also simple, i.e. parsimonious, models for practical applications. With respect to these issues, this study proposed a new model, namely Pareto-optimal MGGP, to calculate Frp in non-deposition with a bed deposit condition in

Acknowledgments

The first author would like to thank to International Affairs of Iran's National Elites Foundation of Iran (BMN) for financial support of this study under the contract no. 102/2716, which was done during his stay as Post-doctoral Research Fellow at Urmia University and University of Tabriz, Iran. Very special gratitude goes to Prof. Maghsud Solimanpur, Vice Chancellor for Research and Technology of Urmia University and Dr. Behnam Mohammadi-Ivatloo from Technology Affairs Management (TAM) of

References (62)

  • C. Nalluri et al.

    Sediment transport over fixed deposited beds in sewers-an appraisal of existing models

    Water Science and Technology

    (1997)
  • J.E. Nash et al.

    River flow forecasting through conceptual models, Part 1- A discussion of principles

    Journal of Hydrology

    (1970)
  • V. Nourani et al.

    Applications of hybrid wavelet–Artificial Intelligence models in hydrology: A review

    Journal of Hydrology

    (2014)
  • E. Olyaie et al.

    A comparative analysis among computational intelligence techniques for dissolved oxygen prediction in Delaware River

    Geoscience Frontiers

    (2017)
  • M.J.S. Safari et al.

    Artificial neural network and regression models for flow velocity at sediment incipient deposition

    Journal of Hydrology

    (2016)
  • M.J.S. Safari et al.

    Non-deposition self-cleansing design criteria for drainage systems

    Journal of Hydro-environment Research

    (2017)
  • Z.M. Yaseen et al.

    Artificial intelligence based models for streamflow forecasting: 2000–2015

    Journal of Hydrology

    (2015)
  • A. Ab Ghani

    Sediment transport in sewers (Doctoral dissertation)

    (1993)
  • A. Ab Ghani et al.

    Gene-expression programming for sediment transport in sewer pipe systems

    Journal of Pipeline Systems Engineering and Practice

    (2011)
  • J.C. Ackers et al.

    Design of sewers to control sediment problems

    (1996)
  • Ackers, P. (1984). Sediment transport in sewers and the design implications. In: Proceedings of international...
  • Ackers, P. (1991). Sediment aspects of drainage and outfall design. Proceedings International Symposium on...
  • H. Aksoy et al.

    Velocity-based analysis of sediment incipient deposition in rigid boundary open channels

    Water Science and Technology

    (2017)
  • A.M. Al-Juboori et al.

    A stepwise model to predict monthly streamflow

    Journal of Hydrology

    (2016)
  • E.M. Alvarez

    The influence cohesion on sediment movement in channels of circular cross-section (Doctoral dissertation)

    (1990)
  • H.H. Ambrose

    The transportation of sand in pipes free surface flow

    (1953)
  • C.H.J. Bong et al.

    Verification of equations for incipient motion studies for a rigid rectangular channel

    Water Science and Technology

    (2013)
  • C.H.J. Bong et al.

    Sediment deposit thickness and its effect on critical velocity for incipient motion

    Water Science and Technology

    (2016)
  • D. Butler et al.

    Self-cleansing sewer design based on sediment transport principles

    Journal of Hydraulic Engineering

    (2003)
  • D. Butler et al.

    Sediment transport in sewers. Pt. 1: Background

    Proceedings of the Institution of Civil Engineers-Water Maritime and Energy

    (1996)
  • J.P. Craven

    The transportation of sand in part-full pipe flows

    (1953)
  • Cited by (53)

    • Pipe failure prediction of wastewater network using genetic programming: Proposing three approaches

      2023, Ain Shams Engineering Journal
      Citation Excerpt :

      In 1992, Koza proposed a computer model of GP by simulating the biological evolution of living organisms [27,28]. Due to specific characteristic of GP, nowadays, GP is used for different water engineering fields, including stream flow forecasting [29], rainfall-runoff modeling [30], water quality variability modeling [31], surface and subsurface water level prediction [32], soil properties modeling [33], sediment transport and erosion prediction [34], dams and reservoirs flow prediction [35], pipeline flow prediction [36,37], open channel flow [38], precipitation [39], and drought forecasting [40]. Koza proposed GP as a developed method of genetic algorithm that is expressed as tree structures [41].

    • Sediment transport with soft computing application for tropical rivers

      2022, Handbook of HydroInformatics: Volume III: Water Data Management Best Practices
    • Multigene genetic programming and its various applications

      2022, Handbook of HydroInformatics: Volume I: Classic Soft-Computing Techniques
    • Machine learning in natural and engineered water systems

      2021, Water Research
      Citation Excerpt :

      In addition, the multigene genetic programming technique was adopted to estimate the particle Froude number in large sewers. Both works were found to be conducive to self-cleansing sewer design (Safari and Danandeh Mehr, 2018). Recently, an ensemble procedure comprised of the Network K-function geographically weighted Poisson regression, and the RF algorithm was applied to analyze the factors that affect sewer pipe blockages.

    • Water quality modeling in sewer networks: Review and future research directions

      2021, Water Research
      Citation Excerpt :

      For the regression models of sulfides, the covariates (i.e., inputs) can vary ranging from sewer structure and seasons to wastewater characteristics and chemical dosages, and the model outputs can be H2S emission hotspots (Zuo et al., 2019) or sulfide concentrations (Jiang et al., 2011). Similar observations can be made for the ANN models applied to sediments, with covariates including pipe sizes, slopes, sediment sizes, sediment concentrations and deposit thickness, and outputs including blockage locations, Froude number or critical flow velocity (Safari and Mehr, 2018; Safari, 2019). It is anticipated that the comprehensive details given in Table 1 can provide significant knowledge regarding the similarities and differences of the modeling processes, model inputs, and model outputs of each model type applied to water quality parameters, which is a useful contribution to the literature.

    View all citing articles on Scopus
    View full text