[Paper Review] Integrated BIM and Machine Learning System for Circularity Prediction of Construction Demolition Waste

Abdullahi B. Saka, Ridwan Taiwo+4 more|arXiv (Cornell University)|Jul 20, 2024

Recycled Aggregate Concrete Performance4 citations

TL;DR

This study proposes an integrated Building Information Modeling (BIM) and machine learning (ML) system to predict the circularity of construction demolition waste at the project level. Using XGBoost on a dataset of 2,280 demolition projects, the model achieved an R² of 0.9977 and a mean absolute error of 5.0910, enabling precise forecasting of recyclable and landfill waste, with SHAP analysis revealing key material and design predictors for circularity.

ABSTRACT

Effective management of construction and demolition waste (C&DW) is crucial for sustainable development, as the industry accounts for 40% of the waste generated globally. The effectiveness of the C&DW management relies on the proper quantification of C&DW to be generated. Despite demolition activities having larger contributions to C&DW generation, extant studies have focused on construction waste. The few extant studies on demolition are often from the regional level perspective and provide no circularity insights. Thus, this study advances demolition quantification via Variable Modelling (VM) with Machine Learning (ML). The demolition dataset of 2280 projects were leveraged for the ML modelling, with XGBoost model emerging as the best (based on the Copeland algorithm), achieving R2 of 0.9977 and a Mean Absolute Error of 5.0910 on the testing dataset. Through the integration of the ML model with Building Information Modelling (BIM), the study developed a system for predicting quantities of recyclable and landfill materials from building demolitions. This provides detailed insights into the circularity of demolition waste and facilitates better planning and management. The SHapley Additive exPlanations (SHAP) method highlighted the implications of the features for demolition waste circularity. The study contributes to empirical studies on pre-demolition auditing at the project level and provides practical tools for implementation. Its findings would benefit stakeholders in driving a circular economy in the industry.

Motivation & Objective

• Address the lack of project-level pre-demolition auditing tools for construction and demolition waste (C&DW) circularity.
• Overcome the limitation of existing studies that focus on construction waste or regional-level data without circularity insights.
• Develop a data-driven system to predict recyclable and landfill waste quantities from building demolitions using BIM and ML.
• Enable stakeholders to make informed decisions for advancing a circular economy in the construction sector through quantifiable waste forecasting.
• Provide empirical, actionable tools for pre-demolition planning based on feature importance and model interpretability via SHAP.

Proposed method

• Collected and processed a dataset of 2,280 demolition projects to extract features related to building geometry, materials, and structural systems.
• Applied Variable Modelling (VM) to generate synthetic features that enhance model generalization and predictive power.
• Trained multiple machine learning models, including XGBoost, Random Forest, and LightGBM, for regression of recyclable and landfill waste quantities.
• Selected the optimal model using the Copeland algorithm based on performance across R², MAE, and RMSE metrics.
• Integrated the trained XGBoost model within a BIM-based system to enable real-time prediction of waste circularity during design and planning stages.
• Utilized SHapley Additive exPlanations (SHAP) to interpret feature contributions and identify key drivers of circularity in demolition waste.

Experimental results

Research questions

• RQ1 How accurately can machine learning predict the quantities of recyclable and landfill waste from building demolitions using BIM-integrated data?
• RQ2 Which machine learning model performs best in predicting demolition waste circularity across multiple evaluation metrics?
• RQ3 What are the most influential design and material features affecting the circularity of demolition waste, as revealed by model interpretability?
• RQ4 To what extent does integrating BIM with ML improve the feasibility and precision of pre-demolition waste auditing at the project level?
• RQ5 Can the proposed system support practical implementation of circular economy principles in construction demolition planning?

Key findings

• The XGBoost model achieved the highest predictive performance with an R² of 0.9977 and a mean absolute error (MAE) of 5.0910 on the test set.
• The Copeland algorithm confirmed XGBoost as the optimal model across multiple metrics, outperforming alternatives like Random Forest and LightGBM.
• SHAP analysis revealed that material type, structural system, and building volume were the most influential features in determining waste circularity.
• The integration of BIM with ML enabled real-time, project-level prediction of recyclable and non-recyclable waste components.
• The system provides actionable insights for stakeholders to prioritize design strategies that enhance material recovery and reduce landfill disposal.
• The study establishes a robust, empirically grounded framework for pre-demolition auditing that supports circular economy transitions in construction.