Learning to Repair Through AI-Driven Geometry Reconstruction for Sustainable Manufacturing

Remanufacturing plays a vital role in advancing circular economy strategies by extending product lifecycles and reducing energy consumption. However, the variability in part condition and lack of predictive tools often hinder efficient repair planning and sustainable decision-making. This paper presents the EcoRemanufacturing Architect, an AI-driven solution designed to reconstruct degraded geometries and support informed decision-making in additive remanufacturing processes. The tool integrates edge-based contour extraction and Fourier descriptor encoding to capture degraded component geometries, which are then predicted from process parameters using a pair of Random Forest Regressors. To enhance generalization across diverse repair scenarios, the training dataset is augmented using a Variational Autoencoder. The reconstructed curves are evaluated using both descriptor-based and spatial metrics, confirming the model’s ability to capture both global shape and local detail with high accuracy. Beyond geometric feasibility, the system estimates key environmental indicators—such as energy consumption, material use, and carbon footprint—enabling multi-criteria evaluation of repair strategies. The proposed pipeline demonstrates how digital intelligence can empower more sustainable and cost-effective remanufacturing workflows by enabling accurate part assessment and resource-aware repair planning.