Carbon footprint estimation and carbon data validation in the manufacturing phase of electromechanical products

The credibility of product carbon footprint estimates is closely related to the validity of carbon data, both of which are essential for sustainable industrial manufacturing. To address the challenges of simultaneously estimating the carbon footprints of multi-products under a multi-variety, variable-batch (MVVB) production mode and validating carbon data shared via blockchain, this study proposes a multi-product carbon footprint estimation method based on a Progressive Layered Extraction (PLE) multi-task learning (MTL) model. This method also supports carbon data validity verification. First, the influence mechanism of carbon data on carbon footprint results is analyzed. The Sobol sampling method is employed to ensure the uniformity and representativeness of multidimensional and fluctuating carbon data samples, thereby constructing a high-quality dataset for carbon footprint estimation. Second, the PLE-based MTL model is employed to capture the coupling relationships between fluctuating carbon data and product carbon footprints. A high-voltage switchgear is used as a practical case study. Based on the fluctuating carbon data from real-world production, a corresponding carbon footprint dataset is constructed, and the accuracy and effectiveness of the proposed estimation model are validated. Finally, this paper proposes a carbon data validity verification approach based on the PLE estimation model and discusses it in detail. The aim is to enhance the credibility and practical utility of product carbon footprint results. Overall, this study provides theoretical support for manufacturing organizations to optimize product carbon footprints based on business activities, thereby promoting carbon trading and sustainable development within the discrete manufacturing industry.