Modeling process and material alternatives in life cycle assessments |
| |
Authors: | Joyce Cooper Christina Godwin Edie Sonne Hall |
| |
Institution: | (1) Department of Mechanical Engineering, University of Washington, P.O. Box 352600, Seattle, Washington 98195-2600, USA;(2) Weyerhaeuser Company, 33663 Weyerhaeuser Way South, Federal Way, Washington zip code, USA |
| |
Abstract: | Background, Aims and Scope Although LCA is frequently used in product comparison, many practitioners are interested in identifying and assessing improvements
within a life cycle. Thus, the goals of this work are to provide guidelines for scenario formulation for process and material alternatives
within a life cycle inventory and to evaluate the usefulness of decision tree and matrix computational structures in the assessment
of material and process alternatives. We assume that if the analysis goal is to guide the selection among alternatives towards
reduced life cycle environmental impacts, then the analysis should estimate the inventory results in a manner that: (1) reveals
the optimal set of processes with respect to minimization of each impact of interest, and (2) minimizes and organizes computational
and data collection needs.
Methods A sample industrial system is used to reveal the complexities of scenario formulation for process and material alternatives
in an LCI. The system includes 4 processes, each executable in 2 different ways, as well as 1 process able to use 2 different
materials interchangeably. We formulate and evaluate scenarios for this system using three different methods and find advantages
and disadvantages with each. First, the single branch decision tree method stays true to the typical construction of decision
trees such that each branch of the tree represents a single scenario. Next, the process flow decision tree method strays from
the typical construction of decision trees by following the process flow of the product system, such that multiple branches
are needed to represent a single scenario. In the final method, disaggregating the demand vector, each scenario is represented
by separate vectors which are combined into a matrix to allow the simultaneous solution of the inventory problem for all scenarios.
Results For both decision tree and matrix methods, scenario formulation, data collection, and scenario analysis are facilitated in
two ways. First, process alternatives that cannot actually be chosen should be modeled as sub-inventories (or as a complete
LCI within an LCI). Second, material alternatives (e.g., a choice between structural materials) must be maintained within
the analysis to avoid the creation of artificial multi-functional processes. Further, in the same manner that decision trees
can be used to estimate ‘expected value’ (the sum of the probability of each scenario multiplied by its ‘value’), we find
that expected inventory and impact results can be defined for both decision tree and matrix methods.
Discussion For scenario formulation, naming scenarios in a way that differentiate them from other scenarios is complex and important
in the continuing development of LCI data for use in databases or LCA software. In the formulation and assessment of scenarios,
decision tree methods offer some level of visual appeal and the potential for using commercially available software/ traditional
decision tree solution constructs for estimating expected values (for relatively small or highly aggregated product systems).
However, solving decision tree systems requires the use of sequential process scaling which is difficult to formalize with
mathematical notation. In contrast, preparation of a demand matrix does not require use of the sequential method to solve
the inventory problem but requires careful scenario tracking efforts.
Conclusions Here, we recognize that improvements can be made within a product system. This recognition supports the greater use of LCA
in supply chain formation and product research, development, and design. We further conclude that although both decision tree
and matrix methods are formulated herein to reveal optimal life cycle scenarios, the use of demand matrices is preferred in
the preparation of a formal mathematical construct. Further, for both methods, data collection and assessment are facilitated
by the use of sub-inventories (or as a complete LCI within an LCI) for process alternatives and the full consideration of
material alternatives to avoid the creation of artificial multi-functional processes.
Recommendations and Perspectives The methods described here are used in the assessment of forest management alternatives and are being further developed to
form national commodity models considering technology alternatives, national production mixes and imports, and point-to-point
transportation models.
ESS-Submission Editor: Thomas Gloria, PhD (t.gloria@fivewinds.com) |
| |
Keywords: | Decision trees expected value inventory analysis material choice process choice |
本文献已被 SpringerLink 等数据库收录! |
|