Evaluating Quality Management Policies

Project Contact: Karl Kempf, Intel Corp.
Proposal Contact: Karl Kempf, Intel Corp.

Go to LFM Research Group 5 Mission Statement page


Background: On any manufacturing floor, there must be policies in place to support the cost effective utilization of capacity and movement of work n progress. This is true whether the factory is building to stock or to order, whether the process is fabrication or assembly (or both), whether execution is primarily manual or highly automated. The problem for any manufacturing organization is to select and implement the most appropriate quantity management policy to meet its business goals, realizing that this policy will have to evolve over time given the progress of manufacturing technology (both equipment and process), the continuous flow of new products, and the dynamics of the global marketplace.

A source of confusion for manufacturing personnel is the variety of quantity management policies available in the technical literature. The "kanban" approach controls small buffers at each station in a manufacturing line as a way to control the overall factory performance (ref. 1). The "hedging" approach exercises global factory control by controlling production rates at each individual station (ref. 2). The "constraint" approach uses one buffer and one rate to control the performance of the entire factory (ref. 3). This list could go on.

Although in principle, given an appropriate model of the factory under consideration, each approach could be simulated or subjected to analysis, such comparison is a practical nightmare. Commercial simulation packages do not support such high level strategies. Although they do allow users to exit into their implementation languages to construct custom features, building these high level quantity management policies from scratch is a serious undertaking. Furthermore, it is not the case that analysis methods exist that are powerful enough to directly compare such sophisticated policies. In any case, it is not often that manufacturing organizations have personnel with the skill set or the time to build these custom features for simulators or to develop new analysis methods.

Proposal Description: The research required would include three tasks:

Note that an obvious follow-on project is to The current proposal is to design QuaN. The follow-on project is implement QuaN.

Schedule: The schedule would be broken into three sections, one for each task .

References: Appendix: Kodak, Polaroid, and Intel are willing to supply concrete examples for Task 1 as briefly described here :

This industrial example from KODAK will focus on a manufacturing "job shop" operation where many different products are produced for several different internal marketing organizations. Collectively these marketing organizations have a diverse customer base representing a wide range of customer wants. One manufacturing line, involving many sequenced shop floor operations, is used as a shared resource for all products. Quantity and cost strategies differ between the internal marketing organizations that rely on this line for products. Other factors that impact optimization of this manufacturing facility include: cycle times, quality out, reliability of manufacturing steps, alignment with upstream and downstream manufacturing operations, etc. This example will provide a realistic view of a complex manufacturing operation using a shared resource to meet corporate goals.

POLAROID currently has a situation in which one of its processing equipment arrays is a focal point and bottleneck for two product platforms and variants of those platforms. This complex processing equipment array provides a key sub-component for other subsequent operations. The factors that effect the output of this equipment are varied: Platform and variant dependent process reliability, number of re-entry steps, amount of rework, amount of scrap, the output rate, the set-up/tear-down times of equipment and supporting equipment, the amount of time and materials to obtain steady state, and amount of support resources. In addition to these internal factors, the external factors of the various customer requirements for product mix create a situation that illustrate the requirements for a quantity model, and in the long term could be an excellent pilot for the entire quantity/quality/cost optimization model. The current situation is an industrial example of the critical need for a model that will enable local management to optimize quantity policy selection in complex, rapidly changing environment in manufacturing.

INTEL operates a number of different types of factories which require different types of quantity policies. In our fabrication facilities, the basic problem is highly re-entrant flow on a linear process over unreliable machines. Multiple products, long throughput times, and an equipment set that includes machines that batch, require setups, and pipeline make the manufacturing problem more difficult. In our assembly facilities, the basic problem is the complexity of the mapping between incoming raw die and outgoing packaged products. Multiple products, strong pressure to deliver to the (external) customer on time, unreliable machines, and an equipment set that includes machines that batch and require setups make the manufacturing problem more difficult. Manufacturing supervision spends a significant time on continuous improvement of our quantity policies, and will be willing and able to support Task 1 of this project with concrete examples from current factories.



This site maintained by Joseph Nemec ( nemecj@hierarchy.mit.edu)
and Maria Carrascosa ( maria@hierarchy.mit.edu
) Last updated: 3/20/96