THESIS ABSTRACT

Assessment of Time-Cost Relation for Predicting Construction Time of Trail Bridge Using Bromilow Time Cost Model and General Regression Neural Network

Assessment of Time-Cost Relation for Predicting Construction Time of Trail Bridge Using Bromilow Time Cost Model and General Regression Neural Network

Student: Roshan Kumar Das

Supervisor: Er. Sujan Nepal/Er. Sudeep Pokhrel

Submitted Date: September, 2024

Abstract

The success of construction companies relies on completing projects within the agreed cost and time limits, with construction time playing a crucial role in achieving this success. Accurate forecasting of the final construction duration early in the project is vital, as errors in prediction can result in significant losses. Considering the several aspects that influence construction time, precise duration estimates are essential for ensuring project success. This research investigates the relationship between the final cost and construction time of trail bridges in Nepal, utilizing both the Bromilow Time Cost (BTC) Model and General Regression Neural Network (GRNN). The primary objectives were to analyze the time- cost relationship using these models and to evaluate their predictive accuracy for estimating construction durations. In this study, KII was conducted to understand the existing methods for calculating contract duration and the viewpoints on the effectiveness of the approach employed. Secondary data for the study was collected from a total of 254 trail bridge projects categorized into D, N, MN, and combined types. The data was refined using MS Excel and analyzed with DTREG software for GRNN, while regression analysis was performed for the BTC model. The findings reveal that neither the BTC nor the GRNN models were statistically significant in explaining the variability in construction time based on project costs, with both models displaying low R-squared values across all bridge types. However, the GRNN model demonstrated slightly better predictive accuracy than the BTC model. The GRNN model outperformed the BTC model in predictive accuracy for some bridge types but still exhibited low overall R2 values, indicating that neither model effectively predicts construction time. The study concludes that both the BTC and GRNN models are insufficient for accurately predicting construction time based on final cost for trail bridges. It recommends further research incorporating additional variables, such as project location, labor availability, weather conditions, and material supply, to enhance predictive accuracy. Future studies should also focus on alternative modeling approaches and the collection of more detailed project data to improve forecasting precision.

Keywords

Bromilow’s time cost model (BTC); General Regression Neural Network (GRNN); Time Cost Model; Duration Prediction; Trail Bridge