Approximate Life-Cycle Assessment of Product Concepts Using Learning Systems

Authors

  • Inês Sousa,

    1. Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
    Search for more papers by this author
  • David Wallace,

    Corresponding author
    1. Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
      David Wallace MIT Department of Mechanical Engineering 77 Massachusetts Avenue, 3-455b Cambridge, MA, 02139 USA drwallac@mit.edu, http:cadlab.mit.edu
    Search for more papers by this author
  • Julie L. Eisenhard

    1. Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
    Search for more papers by this author

David Wallace MIT Department of Mechanical Engineering 77 Massachusetts Avenue, 3-455b Cambridge, MA, 02139 USA drwallac@mit.edu, http:cadlab.mit.edu

Summary

Parametric life-cycle assessment (LCA) models have been integrated with traditional design tools and used to demonstrate the rapid elucidation of holistic, analytical trade-offs among detailed design variations. A different approach is needed, however, if analytical environmental assessment is to be incorporated in very early design stages. During early stages, there may be competing product concepts with dramatic differences. Detailed information is scarce, and decisions must be made quickly.

This article explores an approximate method for providing preliminary LCAs. In this method, learning algorithms trained using the known characteristics of existing products might allow environmental aspects of new product concepts to be approximated quickly during conceptual design without defining new models. Artificial neural networks are trained to generalize on product attributes, which are characteristics of product concepts, and environmental inventory data from pre-existing LCAs. The product design team then queries the trained artificial model with new high-level attributes to quickly obtain an impact assessment for a new product concept. Foundations for the learning system approach are established, and then an application within the distributed object-based modeling environment (DOME) is provided. Tests have shown that it is possible to predict life-cycle energy consumption, and that the method could be used to predict solid waste, greenhouse effect, ozone depletion, acidification, eutrophication, winter and summer smog.

Ancillary