Optimizing the yellow grease supply chain: Inbound inventory routing at a biofuel company.

摘要

The purpose of this thesis is to develop a simulation framework to evaluate a machine learning approach to solve a long term stochastic Inventory Routing problem.;Rocky Mountain Sustainable Enterprises (RMSE), a bio-fuel company has difficulty in constructing an efficient supply chain for her raw material, yellow grease, collection. Their service includes placing and periodically recycling standard containers on the customer sites. RMSE wants a solution which can minimize the long term operational cost and provide an optimal service policy.;The operation of RMSE is modeled as an Inventory Routing problem. Because the customer set is very large, the yellow grease production rates are difficult to predict, and the dataset provided by RMSE is inaccurate, this case is more complicated than previous studies. An algorithm that can automatically update its own parameters is proposed and implemented in C++. Simulations are run on toy data that are based on real scenarios.;Computational experiments show that the algorithm is robust to a certain extent. The parameter updating works but more sophisticated fine-tune strategies are needed. Experiment results also identify some patterns that are not documented in literature. But the current experiment has limitation in identifying the impact of cost structure change on pickup patterns. This limitation is to be addressed in future extensions.