Risk-Informed Prediction of Dredging Project Duration Using Stochastic Machine Learning

Jui-Sheng Chou,Ji-Wei Lin

doi:10.3390/w12061643

Abstract

Dredging engineering projects are complex because they involve greater uncertainty from the natural environment, social needs, government policy and many stakeholders. Engineering companies submit tenders that draw on similar cases undertaken in recent years. However, weather, earthquakes, typhoons and other disasters often change landforms. Therefore, evaluating the duration of dredging projects with reference to only a few previous cases is inadequate, often leading to an unnecessarily long construction duration if the scope of the project is not clearly defined at the early phase. The goal of this investigation aimed to estimate project duration at the beginning of construction and the probability of risk. Evolutionary machine learning was used to build a deterministic model of dredging project duration. Monte Carlo simulation was then utilized to establish the probabilistic distribution of the project duration based on historical patterns. The analytical outputs are displayed through a graphical user interface that provides project coordinators with a means of assessing the uncertainty of project duration in the initial phase of the project. This study will provide a practical reference for contractors and the Water Resources Agency.

Highlights

The Central Mountain Range in Eastern Taiwan has a high altitude and steep slopes
The Water Resources Agency (WRA) only needs to focus on the working quantities of soils and gravels the contractors dredge in a river and the settled amounts the buyers pick up from the dredged materials that should be paid to the WRA
The primary parameters of this model were optimized by particle swarm optimization (PSO) and its stability was verified using 10-fold cross-validation

Summary

Introduction

The Central Mountain Range in Eastern Taiwan has a high altitude and steep slopes. most rivers in Taiwan are short, torrential, have strong scouring forces and flow in the east–west and west–east directions. The WRA has separated dredging from the selling of soil and gravel and developed a cloud-based dredging management system. This strategy enables the split dredging and selling which works more clearly than conventional bids. The WRA only needs to focus on the working quantities of soils and gravels the contractors dredge in a river and the settled amounts the buyers pick up from the dredged materials that should be paid to the WRA. It features numerous modules that programmers can apply and develop. In this investigation, Python 3.6 was used in conjunction with the Tkinter module to program the interface. The interface was named “Dredging Project Quantitative Risk Assessment 123.” The term “123”

Methods

Results

Conclusion