COST-EFFECTIVE RETROFIT PLANNING FOR AGEING FLUID PIPELINES: A SIMULATION-DRIVEN STOCHASTIC PROGRAMMING MODEL

Ageing Heating, Ventilation, and Air Conditioning (HVAC) fluid pipeline systems pose significant operational and economic challenges due to degradation, energy inefficiency, and high retrofit costs. This study aims to develop a simulation-driven, machine learning-based framework to accurately predict retrofit costs and support cost-effective decision-making for ageing pipeline networks. A quantitative research methodology was adopted, where synthetic data were generated via stochastic simulation using Monte Carlo techniques in Python (NumPy and Pandas). Three supervised machine learning models, Random Forest, eXtreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN), were implemented and evaluated using RMSE, R², and MAPE as performance metrics. The results showed that XGBoost achieved the best performance with an RMSE of 221.19, R² of 0.929, and MAPE of 5.90%, followed closely by Random Forest, while ANN underperformed with an RMSE of 335.03. XGBoost and Random Forest are closely aligned with actual retrofit costs, indicating strong predictive accuracy. The study concludes that ensemble models trained on simulation-derived data offer a robust solution for proactive retrofit planning. The findings have significant implications for enhancing infrastructure resilience and optimising maintenance investment. However, the study is limited by the use of synthetic data and recommends future work with real-world datasets and expanded modelling techniques for multi-objective predictions.