Time-Cost Analysis of Trail Bridge Construction in Nepal Using Bromilow Time-Cost Model and General Regression Neural Network

Abstract

Timely completion of construction projects is essential for ensuring cost efficiency and project success, with accurate forecasting of construction duration playing a crucial role. This study examines the relationship between final cost and construction time in trail bridge projects across Nepal, using two modelling approaches: the Bromilow Time-Cost (BTC) Model and the General Regression Neural Network (GRNN). The analysis covers 254 completed trail bridge contracts implemented by the Suspension Bridge Division (SBD) between fiscal years 2018/19 and 2021/22 across the mountain, hill, and Terai regions. Bridge types include D-type (dismantled), N-type (new), and MN-type (modified new). Data are primarily sourced from official Suspension Bridge Division records, supplemented with interviews with engineers and project managers for validation and contextual understanding The BTC model was analysed using Microsoft Excel, and the GRNN model using DTREG software. Results show that neither model establishes a statistically significant relationship between cost and duration, with R² values below 0.10 across all bridge types. Although GRNN slightly outperformed BTC, predictive accuracy remained limited. Interview findings identified contextual factors undermining model performance, including fixed-price turnkey contracts, limited qualified contractors, subcontracting, delayed budget releases, frequent regulatory changes, labor shortages, remote material delivery challenges, and environmental and social disruptions. The study concludes that cost alone is insufficient for predicting construction duration. It recommends incorporating broader explanatory variables and applying advanced or hybrid modelling approaches informed by expert insights to improve timeline forecasting in such settings.