Full length articleImprovement of transportation cost estimation for prefabricated construction using geo-fence-based large-scale GPS data feature extraction and support vector regression
Introduction
Panelized construction is a manufacturing-oriented construction method that provides advantages in project delivery time, construction cost, site safety, and environment protection compared to the site-oriented traditional stick-built construction method [1]. Panelized construction can also be understood as a form of prefabricated construction in which wall, floor, and roof panels are prefabricated in an offsite environment (i.e., factory) to reduce construction site activities. Due to its advantages, the application of the panelized construction method has increased in a variety of construction contexts. Especially in the residential sector, panelized construction has been garnering more attention because of its impact on reducing site operation costs while maintaining flexibility in design [2]. To reduce site operation time and cost, the prefabricated panels are transported to a construction site from an offsite factory, and a mobile crane is generally used to perform final assembly. Due to the intensive utilization of transportation processes, panel delivery to a construction site becomes a critical operation in panelized construction, and transportation cost as a proportion of the total project cost becomes considerable [3], [4], [5], [6], [7]. Despite the importance of this cost in panelized construction, though, estimation of transportation cost has not been well studied due to an under-utilization and lack of transportation operation data, and due to the uniqueness of residential project conditions, which makes it difficult for estimators to identify and collect the cost data associated with transportation and to have a reliable estimation model that reflects the operations data.
In residential construction, various types of residential projects (e.g., single-family dwelling, duplex, townhouse, apartment) exist with varying designs, dimensions, and project locations, which make each project unique. For example, even residential projects of the same size (i.e., floor area) could have different transportation demands, such as the number of trailers required and the delivery cycle times, due to different design requirements. Each house can be customized for exterior features such as a veranda and a balcony with different shapes and sizes, and the roof design can be modified as well to meet customers’ demands. These different combinations of exterior features create different trailer requirements even though the floor areas may be identical. As a result, in practice, project estimators at the case panelized construction company in Edmonton, Canada, under study in the present research, report that inaccurate estimations of transportation costs frequently occur when using a fixed-cost approach that considers the transportation cost as part of the overhead cost. Since this fixed-cost approach over-generalizes the transportation costs, accuracy of the cost estimation has not been reliable and has reduced the case company’s profit margin. In this regard, a more in-depth understanding of transportation operations may allow for identifying any waste of resources including operation time, and also reducing overhead costs through accurate transportation cost estimation and scheduling. In addition, the high overhead caused by transportation equipment such as trailers, trucks, and mobile cranes has been regarded one of the major barriers to adopting and implementing the prefabricated construction method [8].
To estimate the transportation cost associated with panelized construction, this study proposes the utilization of global positioning systems (GPSs) attached to trucks, by which key features (e.g., the number of visits) can be extracted from the GPS data and used to model the relationship between such features and the transportation cost. To efficiently extract key information from a large-scale GPS operation data, the proposed approach specifically investigates (i) a virtual fence (e.g., geo-fence) to observe and quantify transportation demands for existing projects, and (ii) a machine learning model to estimate the demands and predict the resulting transportation cost for new projects using the historical data collected. The use of a GPS device commonly installed for construction equipment and vehicles may help understand various factors (e.g., utilization or idling factor) potentially affecting transportation operation costs by tracking the temporal locations (whereas the utilization of GPS data in the past has been limited primarily to crime prevention purposes [9], [10]). To test and validate the proposed approach, a case study is undertaken of a company prefabricating panel for residential construction projects in Edmonton, Canada, in which the transportation costs are estimated and compared with the actual costs.
Section snippets
Research background
With the development of prefabricated construction in past decades, attention has been focused on the manufacturing aspects to adapt to new ways of performing offsite construction operations [11]. Meanwhile, relatively little research has been carried out with respect to controlling cost in transportation operations compared to panel or module manufacturing processes [12], even though transportation processes (e.g., panel delivery) are considered an essential activity linking the offsite
GPS-Data-Based transportation cost estimation
For the estimation of transportation costs associated with panelized construction, this study presents the spatial data-based filtering and abstracting approaches by using the geo-fence to extract operational events of equipment from the GPS data and predict transportation demand required for a particular project. With GPS tracking measurements (e.g., GPS coordinates) periodically recorded and transferred to database, this paper focuses on how to determine geo-fences (e.g., boundary, radius)
Implementation of the proposed Framework: Case study
The proposed framework is implemented and assessed through a case study of a panelized residential construction company in Edmonton, Canada. The company mainly focuses on local residential construction projects, and on average 2 to 3 projects are executed per day, this rate being constrained by the availability of mobile cranes for on-site assembly operations. To deliver prefabricated panels to sites, the company operates 5 trucks with 50 flatbed trailers. For tracking purposes, each truck has
Results
To determine the accuracy of the proposed approach, outcomes (e.g., truck visit counts and durations) from the algorithm are compared against the actual demands from the truck operation logs that show the truck dispatching history on each residential project. In addition, different geo-fence radiuses were examined to determine the most accurate geo-fence setting. The comparison results are displayed in Fig. 9. The results showed that the smallest (50 m) geo-fence had the highest accuracy with
Discussion
The proposed GPS-data-based transportation cost estimation method is able to provide improved cost estimation over a range of different house sizes, locations of projects, and various traffic conditions. The results of the case study are summarized as follows: (1) the average difference of 0.77 number of trucks and −0.3 h of unloading duration from the fleet GPS data analysis given a 50 m geo-fence radius setting; (2) prediction accuracy of 86% and 88% in transportation demands of the visit and
Conclusions
In prefabricated construction, the associated transportation cost has not been clearly quantified, which leads to a lack of control over cost and to inaccurate project cost estimations. To improve the cost transparency and estimation accuracy, this study proposed a transportation cost estimation framework using the historical fleet GPS data and SVR models. The fleet GPS data analysis algorithm is developed to extract key information from the fleet GPS data, and the outcomes from the algorithm
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research was supported by the Natural Sciences and Engineering Research Council of Canada (Grant No. CRDPJ 516160-17), and by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2018R1A5A1025137). The authors are also grateful for the support from ACQBUILT Inc. Any opinions, findings, or conclusions and recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the Natural Sciences and
References (36)
- et al.
The evolution of prefabricated residential building systems in Hong Kong: A review of the public and the private sector
Autom. Constr.
(2009) The economic implications of subcontracting practice on building prefabrication
Autom. Constr.
(1997)- et al.
Identification of workstations in earthwork operations from vehicle GPS data
Autom. Constr.
(2017) - et al.
Critical evaluation of off-site construction research: A Scientometric analysis
Autom. Constr.
(2018) - et al.
Progressive 3D reconstruction of infrastructure with videogrammetry
Autom. Constr.
(2011) - et al.
Learning and classifying actions of construction workers and equipment using Bag-of-Video-Feature-Words and Bayesian network models
Adv. Eng. Inf.
(2011) - et al.
Automated 2D detection of construction equipment and workers from site video streams using histograms of oriented gradients and colors
Autom. Constr.
(2013) - et al.
Vision-based action recognition of earthmoving equipment using spatio-temporal features and support vector machine classifiers
Adv. Eng. Inf.
(2013) - et al.
Construction equipment activity recognition for simulation input modeling using mobile sensors and machine learning classifiers
Adv. Eng. Inf.
(2015) - et al.
Automatic spatio-temporal analysis of construction site equipment operations using GPS data
Autom. Constr.
(2013)
Application of dynamic time warping to the recognition of mixed equipment activities in cycle time measurement
Autom. Constr.
Support vector regression model for the prediction of loadability margin of a power system
Appl. Soft Comput.
Carbon footprint of panelized construction: an empirical and comparative study
Sustainable construction aspects of using prefabrication in dense urban environment: a Hong Kong case study
Constr. Manage. Econom.
Factors affecting the use of precast concrete systems in the United States
J. Constr. Eng. Manage.
Chapter - 02: Aggregating global products for just-in-time delivery to construction sites
Capital cost optimization for prefabrication: A factor analysis evaluation model
Sustainability
Cited by (29)
Multi-objective optimization for coordinated production and transportation in prefabricated construction with on-site lifting requirements
2024, Computers and Industrial EngineeringAdaboosting graph attention recurrent network: A deep learning framework for traffic speed forecasting in dynamic transportation networks with spatial-temporal dependencies
2024, Engineering Applications of Artificial IntelligenceOperations planning and scheduling in off-site construction supply chain management: Scope definition and future directions
2023, Automation in ConstructionCloud-based information system for automated precast concrete transportation planning
2023, Automation in ConstructionIntegrating off-site and on-site panelized construction schedules using fleet dispatching
2022, Automation in ConstructionCitation Excerpt :The generated dispatching sequences can also be periodically processed to continually optimize the operation schedule based on the current situation. Owing to the increasing popularity of panelized construction within the residential sector, several studies have been conducted to improve its operational efficiency [7–10]. However, these previous studies have focused mainly on the prefabrication operations in factories.
A framework for transportation infrastructure cost prediction: a support vector regression approach
2022, Transportation Letters