Supply Chain Servo Control: Models and Algorithms
Open Access
- Author:
- Lee, Seok Gi
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 20, 2013
- Committee Members:
- Vittaldas V Prabhu, Committee Chair/Co-Chair
Paul Griffin, Committee Member
Tao Yao, Committee Member
Douglas J Thomas, Committee Member - Keywords:
- supply chain management
capacity
manufacturing
maintenance
inventory
delivery - Abstract:
- In this research, we develop rigorous mathematical models and algorithms for servo control of five important functions in the manufacturing supply chain: (i) capacity, (ii) maintenance, (iii) production, (iv) inventory, and (v) delivery, that are continuously controlled in a unified manner to adapt to the frequently changing needs of supply chain conditions in real time. We treat the five functions together, to provide near-optimal solutions, considering overall supply chain performance. All of these algorithms use distributed feedback control of continuous variables and therefore can be modeled using differential equations, making them amenable to control theoretic techniques. Performing a variety of computational experiments, we show how operational performance of supply chains can be enhanced through unified controls that pool flexibilities and redundancies across these functions. The contribution of this research can be summarized as follows. We develop analytical models of five supply chain functions to characterize real-time dynamics that span from sensors that monitor the health of manufacturing machinery to distribution of finished goods to customers. Dynamic models for five supply chain functions are developed based on theories of nonlinear and discontinuous differential equations, in order to characterize real-time dynamics in a unified manner. Furthermore, we develop multi-function servo control algorithms which iteratively adjust continuous control variables based on feedback obtained from simulation. The dynamic models serve as a unified mathematical framework, consisting of a system of differential equations for design, analysis, and optimization of these five supply chain servo control algorithms. Such a unified approach allows pooling of flexibility and redundancy across supply chain functions, and thus offers potential for significant improvements in operational performance. Applying unified control to the Assemble-to-Order (ATO) production system, average 8.94% improvement in terms of due date deviations of assembly jobs and 86.31% improvement in inventory discrepancy can be obtained, compared to the conventional production system approach. Computational experiments on the unified control of machine capacity, maintenance scheduling, and production scheduling indicate that approximately 10 ~ 20% performance difference is observed compared to the optimal approach, but a much faster rate of convergence, below 0.2 CPU time for almost all cases, is guaranteed by the unified control approach. Such computational tractability of the unified control approach facilitates real-time simultaneous control of various supply chain functions. We apply the unified control approach for solving the capacitated open vehicle routing problem with time windows (COVRPTW), considering just-in-time delivery and greenhouse gas emissions as important performance metrics, and achieve in 1 ~ 9% discrepancies for both metrics compared to optimal solutions, while keeping extremely short computational time for large scale problems. We also explore potential use of the unified control approach for the home care crew scheduling problem (HCCSP), which has more complex time and resource constraints than COVRPTW but has many system analogies in terms of time constraints, and we show how such a complex scheduling problem can be solved effectively.