Planning, Coordination, and Control of Supply Chains

Funding Agency: National Science Foundation, Directorate for Engineering, Division Division of Design, Manufacture and Industrial Innovation (DMII), Manufacturing Enterprise Systems (MES) Program.

Award Number: DMI-0300359.

Principal Investigators: Yannis Paschalidis and Michael Caramanis, Boston University.

Project Summary

In an era of time-based competition, the management and control of supply chains has emerged as a critical component of manufacturing and distribution enterprises. Customers have become more demanding and require customized products delivered in a consistently timely manner. Production speed along all stages of the supply chain and high service rates are key performance measures. Inventory cost reduction and Quality of Service (QoS) provisioning are in the foreground of practitioners' radar screens. Despite successful decision support tool development to date, significant gaps exist. The primary research objective of this proposal is to develop methodology that enables a practical and effective framework for decentralized, yet coordinated, management and control of supply chains. To that end, it is proposed to address a host of key challenges:

  1. Model lead-times dynamically. State of the art industry practice is to model lead times at various stages of the supply chain as constant delays. In a highly dynamic environment, this often leads to a vicious cycle of excessive inventory followed by excessive backlog. Instead, the proposed work will rely on performance analysis results for stochastic queueing networks to accurately express lead times as functions of capacity utilization, production mix, and lot sizes conditional upon dispatch policies, and distributions of stochastic disturbances such as failure and repair times.
  2. Provide Quality-of-Service (QoS) guarantees. This requires moving from traditional performance metrics such as average backlog costs or average lead times to constraints that bound the probability of backlog or the probability of long lead times within a desirable range. Such constraints better reflect customer satisfaction considerations and closely follow industry practice.
  3. Model uncertainty accurately. This requires moving from traditional models that either ignore variability or adopt simple models of uncertainty involving renewal stochastic processes to models that involve autocorrelated stochastic processes and capture strong temporal dependencies in the demand and processing capacities that drive the supply chain. Such dependencies are key in modeling realistic demand scenarios and failure-prone processing.
  4. Resource Allocation. A supply chain may be viewed as a collection of resources whose management must deliver the required performance. In a traditional ``just in case'' environment, resources are relatively easy to manage, since congestion is not a concern. In an agile "just in time'' system, however, congestion management is the operational centerpiece. Resources have to be allocated to the right task at the right time in the right place.

To meet these objectives the proposed work will draw upon the PIs expertise and prior experience with modeling and analysis of stochastic systems, simulation as a design and analysis tool, dynamic optimization, and optimal control. To deal with computational complexities it is proposed to (i) exploit the time scale decomposition among decisions with different scope and functionality, (ii) employ analytical approximations to estimate complex quantities of interest, and (iii) use efficient simulation methodologies. The broader impact of this research stems from its interdisciplinary science base contributions to information and production systems engineering, and their potential to enable significant productivity growth in the manufacturing and distribution industries. On the educational front, plans consist of (i) integrating the proposed research into courses, (ii) training graduate students by exposing them to a balanced mix of relevant theory and systems engineering practice, (iii) involving undergraduate students in the work, (iv) creating interactive, java-based, educational software and demos, and (v) working through the NSF science ambassador award to Boston University (BU) -- one of the ambassador fellows is our doctoral advisee -- to reach out to high school students. Finally, and in addition to the usual means of disseminating the outcomes of the proposed work (publications, presentations at conferences, invited lectures, etc.), the PIs plan to: (i) leverage their association with the recently established BU Center for Information and Systems Engineering (CISE) to work with affiliated companies including Foresight Systems, Brooks-PRI Automation, Genuity, Sycamore, Nokia, Hewlett-Packard, Solectron, and others; (ii) present findings on the Web using java-based interactive examples; and (iii) leverage their recent NSF IGERT award on Advanced Computing in Engineering and Science to sponsor doctoral student internships at participating companies and support a post doctoral fellow who will focus on applied work in industry.