Screening methodologies for life cycles inventory models (Data quality)

Two screening methodologies are presented that provide Life Cycle Assessment (LCA) practitioners with a tool and framework for streamlining the life cycle inventory-modeling phase. The two methodologies screen both deterministic and stochastic forms of inventory models. Their development and application resolves the problem of needing operational methods to tell LCA practitioners where to invest in future data quality research with high priority. The first screening methodology ranks each input data element in the deterministic inventory model. This ranking is based upon the amount each input data element contributes toward the final output. The application is proven to be effective at improving and streamlining the inventory modeling process during the conversion stage to its stochastic modeling form. For those inventory models already in a stochastic form, a second screening methodology is presented that allows LCA practitioners to identify and determine the level of quality the input data elements should receive given any constraining requirements. This second methodology utilizes the stochastic nature within the inventory models to solve the problem by combining Monte Carlo simulation and a genetic algorithm. Both methodologies were validated by application to real-world beverage delivery system LCA inventory models. The results from the application show that by screening and improving the quality of the input data elements, reductions in the inventory models output variance are obtainable, thus improving the discriminating ability when comparing alternative system designs. To complete the screening framework, variance reduction techniques are applied to the Monte Carlo based genetic algorithm to improve the efficiency in the required simulation time for evaluating the large number of potential solutions. A factorial design is used to determine which type of variance reduction technique is applicable and to approximate the required number of replications. Lastly, future research ideas are presented to enhance and improve upon the developments obtained within this dissertation.