Real-Time Life Cycle Assessment in the Production of Composite Structures using Visual Recognition

Kurzfassung

Increasing use of carbon fiber reinforced polymer (CFRP) structures in the aerospace industry demands efficient manufacturing methods. Therefore, life-cycle assessment (LCA) and life-cycle cost analysis (LCCA) are valuable tools to examine associated carbon footprint and direct cost. However, due to the dominating proportion of manual work in manufacturing of aerospace CFRP structures, acquisition of relevant manufacturing data is a tedious process. In this work a new system within the concept of smart-work-station (SWS) is introduced, which uses sensor fusion to automatically track process steps during preforming of CFRP structures. Therefore, the work surface of a workstation is equipped with digital scales while cameras are installed to provide a top view. Elementary flows measurement is initiated by a change of weight at the scales. Combined with visual recognition these activities are recognized and recorded as events. The gathered data is fed to the ecoefficiency assessment model (EEAM) to allow process analysis in real-time. Items used are either identified via quick response (QR)-codes or an object detection system. Therefore, a convolutional neural network (CNN) was trained to recognize tools and ancillaries common for preforming of CFRP structures using transfer learning. The system has a modular design that allows customizable workstation configurations within the SWS concept. Several workstations can be connected to a central database. This approach enables deployment at the whole shop-floor level of a CFRP manufacturing facility. It is concluded that real-time LCA is an advantageous tool to improve CFRP manufacturing processes. Broadcasting LCA results to the shop-floor helps to raise awareness for a more efficient use of resources. Ecological and economic benefits as well as insights for different processes support decision makers towards sustainable direct applications