Data-Driven Quality Assurance: Defect Prediction to Reduce Carbon Footprint in Manufacturing Logistics

In modern manufacturing environments, minimizing carbon emissions while maintaining product quality has become a strategic imperative. One critical yet often overlooked contributor to elevated CO₂ emissions is the occurrence of unexpected quality defects, particularly those caused by human error for which no permanent poka-yoke solution exists. These defects frequently disrupt production flows, delay customer deliveries, and necessitate urgent corrective shipments leading to increased use of premium, carbon-intensive transportation. Traditional production planning systems typically neglect the variability introduced by non-conforming parts, assuming uniform quality across all outputs. To address this and to build on Industry 4.0 and Quality 4.0 frameworks, the research compares statistical and machine learning models for defect prediction using six months of real-world production data. An automatic training and testing procedure is applied to evaluate model performance, enabling an automatic selection of the optimal method per workstation. This allows for precise estimation of defect rates across the production line, supporting dynamic and environmentally-conscious planning decisions. By incorporating predicted quality deviations into daily planning routines, the model not only enhances on-time delivery performance but also reduces the reliance on expedited transport, contributing to significant carbon footprint reduction and more sustainable operations.