Session: DFMLC-03-01-Design for Supply Chain and End of Life Recovery

Paper Number: 97923

97923 - Expert Elicitation and Data Noise Learning for Supply Chain Material Flow Analysis Using Bayesian Inference 

Bayesian inference allows the transparent communication of uncertainty in supply chain material flow analyses (MFAs) with the uncertainties reduced as newly collected data are included. However, the method is undermined by the difficultly of defining proper priors for the MFA parameters and quantifying the noise in the collected data. We start to address these issues by first deriving an expert elicitation framework suitable for generating MFA parameter priors. Second, we propose to learn the data noise concurrent with the MFA parametric uncertainty. These methods are demonstrated using a case study on the 2012 U.S. steel flow. Eight experts were interviewed to elicit distributions on steel flow uncertainty from raw materials to intermediate goods. For each parameter, the experts' distributions were combined and weighted depending on the expertise demonstrated in response to seeding variables with available observations. These aggregated distributions form our Bayesian model priors. MFA data were then collected from the United States Geographical Survey (USGS) which are published without uncertainties. A sensible, weak-informative prior was used to describe data noise. Bayesian inference was then used to update the parametric and data noise uncertainty. The results show the reduction in MFA parametric uncertainty when incorporating USGS data. Only a modest reduction in data noise uncertainty was observed; however, greater reductions were achieved when identifying time-invariant parameters across time and incorporating data from multiple years. These methods generate transparent MFA and data noise uncertainties learned from collected data rather than based on input uncertainty assumptions, providing a more robust basis for decision-making that will affect the system.

Presenting Author: Daniel Cooper University of Michigan

Presenting Author Biography: Dan Cooper is an Assistant Professor in the Mechanical Engineering department at the University of<br/>Michigan. He heads the Resourceful Manufacturing and Design (ReMaDe) group, which is dedicated to<br/>pursuing environmental sustainability through process innovations in resource efficiency and optimized<br/>manufacturing and recycling supply chains. Dan’s work is at the nexus between Industrial Ecology (IE)<br/>and Mechanical Engineering (ME): he uses and develops IE methodologies such as material flow analysis<br/>and life cycle assessment to identify opportunities and quantify impacts at the process, factory, and<br/>supply chain scale, and then pursues an experimental and mechanistic modeling approach to generate<br/>the scientific knowledge underlying those opportunities. Dan received all his degrees in Mechanical<br/>Engineering from the University of Cambridge before completing a post‐doc at MIT. He is the recipient<br/>of the ASME Ben C. Sparks Education Medal and the SME Outstanding Young Manufacturing Engineer<br/>Award.

Authors:

Daniel Cooper University of Michigan