XAI–LCA random forest models and datasets for PLA material extrusion (MEX)

This Zenodo record provides the code, trained random forest models, processed datasets, and STL geometries accompanying the study: “Towards Sustainable Material-Extrusion Additive Manufacturing: Explainable AI Analysis of Printing Parameter Effects on Life Cycle Assessment”Yanlong Zhu et al. (manuscript under review). The materials are organised into four components: 01_model_building 134-sample training datasets for the Direct-RFM, EC-RFM, and MC-RFM models (Excel). Trained random forest models (Direct-RFM.pkl, EC-RFM.pkl, MC-RFM.pkl). A 434-sample dataset (predictions_434_samples.xlsx) combining 134 semi-real slicer-based cases and 300 virtual LHS samples, with predictions for four ReCiPe 2016 Midpoint (H) indicators (GWP, FRS, OFHH, TA). Python scripts (ecmc_model.py, predict_lca.py) implementing the EC/MC decomposition model and a command-line interface for inference. 02_figure_generation Processed data tables (CSV/XLSX) and Python scripts to reproduce Figures 6–12 in the manuscript, including: predicted-vs-actual plots, Direct-RFM vs. EC/MC-RFM comparison, SHAP importance, response surfaces/PDP-style curves, Bayesian optimisation convergence, optima scatter plots, perturbation analysis, ratio-to-optimum heatmaps, and cost/material-transfer charts. Scripts rely on standard Python packages (e.g., pandas, numpy, matplotlib, scipy, scikit-learn, joblib; seaborn is only used for the heatmap). 03_benchy_application An end-to-end application demonstration on the widely used 3DBenchy benchmark object (PrusaSlicer → G-code → automatic LCA Excel report). Includes example G-code files for PLA/PETG and infill variations, factor tables (electricity per minute; material per kg), automatically generated Excel reports, and the post-processing script (post_lca_report.py). geometries (STL files) rect_specimen_180x15x2p4mm.stl: rectangular tensile specimen used to generate the main dataset. 3DBenchy_calibration_boat.stl: 3DBenchy geometry used for the application example (see https://www.3dbenchy.com). This dataset is intended to support reproducibility of the proposed XAI–LCA workflow and facilitate reuse of the models and figure-generation pipeline in related research on sustainable material-extrusion additive manufacturing.