Quantification of uncertainty in life cycle analysis

Amaninder Singh Gill

The aim of this thesis is to accumulate uncertainty in Life Cycle Analysis (LCA). LCA is a technique to identify environmental impacts of processes and thereby product's life cycle from material extraction to product disposal. An important aspect of LCA is that it is not always feasible to measure the environmental impact of individual processes occurring during a product life cycle. Therefore, data from similar processes from different time frame, different geographic location, different process technology, etc., is utilized. Such approximate data introduces uncertainties in the results of LCA. The uncertainty in LCA is usually a mix of two types of uncertainty; aleatory, arising from natural process variability, and epistemic, arising due to lack of information regarding the process and related environmental impact. Aleatory v uncertainty has been applied and implemented in LCA software using Monte-Carlo Simulations. To the best of our knowledge, epistemic uncertainty or mixed uncertainty has not been applied in a LCA In order to apply epistemic or mixed uncertainty in LCA, a model is developed based on evidence theory and random sets. A specific variation of evidence theory, called Dempster-Shafer theory is utilized in creating the model. Random sets help in quantifying uncertainties from multiple data sources regarding the same process. These sets may have statistical distribution and will have belief and plausibility functions associated with them. In the model the random sets with distributions, belief and plausibility functions from multiple processes in the product life cycle are accumulated to provide statistical distribution and mean for the resultant environmental impact. To apply this theory, a test case of TV Remote Control was taken. As a first step, the environmental impacts obtained from product's LCA (environmental impacts) of individual life cycle stage were modeled with epistemic uncertainty as random sets. The uncertainties were then discretized into corresponding belief and plausibility functions. Then, a cumulative distribution was created from the random sets to accumulate the results. Results in terms of statistical variation and mean can then be obtained from the accumulated cumulative distribution functions.

Quantification of uncertainty in life cycle analysis

Files and links (1)

Abstract

Metrics

Details