Decision making under uncertainties for supply chain optimization

Thumbnail Image
Hu, Zhengyang
Major Professor
Guiping . Hu
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Organizational Unit
Industrial and Manufacturing Systems Engineering
The Department of Industrial and Manufacturing Systems Engineering teaches the design, analysis, and improvement of the systems and processes in manufacturing, consulting, and service industries by application of the principles of engineering. The Department of General Engineering was formed in 1929. In 1956 its name changed to Department of Industrial Engineering. In 1989 its name changed to the Department of Industrial and Manufacturing Systems Engineering.
Journal Issue
Is Version Of
Industrial and Manufacturing Systems Engineering

Efficient supply chain design and operation are essential for manufacturing production. The main stakeholders in a supply chain network include upstream suppliers, downstream customers and competitors. Manufacturing plants acquire raw materials from upstream suppliers and convert them to final products, which are shipped to downstream facilities such as distribution centers and other manufacturing plants. At the strategic level, supply chain management involves designing the configuration of the network, i.e., where to build manufacturing plants, warehouses and distribution centers. At the tactical level, supply chain management involves purchasing raw materials from upstream suppliers, production planning, distribution of products to downstream customers. At the operational level, supply chain management involves demand fulfillment, inventory control and transportation. In this dissertation, mathematical models have been formulated to study various manufacturing supply chain problems, with a focus on decision making under uncertainties for network design, production planning and closed-loop supply chain.

In the first paper, I proposed a novel two-stage stochastic programming model for relay network infrastructure to improve work-life balance of truck drivers. Various valid inequalities have been generated to enhance computational performance and results indicated that up to 30% computational time can be saved. The robustness of the model has been tested by generating scenarios and checking the feasibility of the deterministic model. In addition, we identified the bottlenecks in the system and provided insights on how to improve current network configuration.

In the second paper, I studied a lot-sizing and scheduling problem, which is at the tactical level of supply chain management. Decisions include determining batch sizes and production sequences. A multi-stage stochastic programming model has been developed. Scenario generation and reduction have been used to generate scenarios and identify the most representative subset. A case study based on a manufacturing firm has been conducted to illustrate and verify the model. Results show that by using the multi-stage stochastic programming model, the objective values reduced by 10% - 13% compared to the two-stage stochastic programming model.

In the third paper, I proposed a hybrid stochastic and robust optimization model for the lot- sizing and scheduling problems. Different types of uncertainties (demand and overtime processing cost) have been studied, simultaneously. I assumed there was not enough historical data for demand and hence robust optimization was adopted to handle demand uncertainty. On the other hand, I assumed there was sufficient historical data for overtime processing cost, therefore, stochastic programming was used to handle overtime processing cost uncertainty. Various sensitivity analyses have been conducted and results shown that considering uncertainties are very crucial since the hybrid model outperformed the deterministic model in the objective values.

In the last paper, I studied a closed-loop supply chain problem which integrates network design and production optimization. A fuzzy multi-objective mixed integer linear programming model has been proposed. The two objective functions are minimization of overall system costs and minimization of negative environmental impact. Several uncertain parameters are studied such as demand, return, scrap rate, manufacturing cost and negative environmental factors. The original model with uncertain parameters is firstly converted to a crisp model and then an aggregation function is applied to combine the objective functions. Sensitivity analyses on various parameters have been examined.

In order to improve the data utilization and interpretation of outcomes, various statistical methods such as Monte-Carlo simulation, moment matching method for scenario generation, and Fast Forward Selection for scenario reduction are applied. The main goal of this dissertation is to quantify the uncertainties in the supply chain design and operational planning processes. Insights have been provided for decision makers in network design and production planning. The results derived from this dissertation have the potential to contribute to the decision making processes under uncertainties by providing analytic solutions for designing robust and efficient supply chain networks.

Subject Categories
Sun Dec 01 00:00:00 UTC 2019