Using Simulation to Determine the Batch Size for I/O Drawer Test Process in a High-End Server Manufacturing Environment
Using Simulation to Determine the Batch Size for I/O Drawer Test
Process in a High-End Server Manufacturing Environment
Lawrence Al-Fandi1* and Faisal Aqlan2
* Corresponding author: Tel.: +965-2225-1400 Ext. 1288; Fax: (+965) 2654-8484; E-mail: lawrence.alfandi@aum.edu.kw
1Industrial Engineering Department
American University of the Middle East
Egaila, Kuwait 54100
2Industrial and System Engineering
University of New Haven
West Haven, CT 06516 USA
ABSTRACT
In this research, discrete-event simulation is used to study the I/O drawer test process in a high-end server manufacturing
environment in order to identify the optimal batch size of the I/O drawers to be tested per driver. High-end server manufacturing
environment is characterized by fast lead time and lengthy build process. Therefore, main components of the server, such as I/O
drawers, are tested ahead of time (fabrication test) and stored in inventory as tested parts to be ready for prompt fulfilment of the
customer orders. In this research, a simulation model is developed for the I/O drawer test process with a focus on batching the I/O
drawers on testing. Different scenarios for the batch size are considered and a statistical comparison is performed against the
current scenario used by the I/O drawer Fab test operators. Unlike the “one-piece flow” lean concept which encourages small
batch size (even one), the results show significant savings in cycle time and energy consumption when the batch size is increased.
This is attributed to the lengthy setup time of the I/O drawer testing process as using small batch sizes requires very short set-up
time. The optimal batch size scenario results in cycle time savings by 20% which is equivalent to 8116 hours per year. Other
savings include: electrical energy and less consumption of chilled water for the cooling units.
1. INTRODUCTION
High-end server environments are characterized by the lengthy testing processes of servers and server components,
short lead time, and stochastic nature of production processes and order arrivals. Every customer order for a high-end
server can be unique. Moreover, before an order is shipped, the customer may alter the server configuration or even
cancel the order. The lead time requested by the customer is much less than the time required for building a server.
Hence, there is a need to adopt the build-to-plan strategy (push system) in order to be ready for any customer orders.
When a customer order is received, tested parts are pulled from the inventory and assembled to the customer order (pull
system). One of the main components of the high-end servers is the I/O drawer which is tested before being assembled
into the final product to guarantee the quality of the finished product, decrease the number of defects that occur during
the assembly processes, and reduce the subsequent quality control tests. The I/O drawer assembly and test consist of
three major processes: (1) I/O drawer assembly, (2) Fab test for I/O drawers, and (3) dekiting of tested I/O drawer.
I/O drawers are assembled then batched in a frame for the fabrication test. The number of I/O drawers in the batch
varies from 2 to 8 depending on the availability of the I/O drawers and the personal decision of the assembly operator.
Usually, the operators assemble six or less I/O drawers per frame since batching more than six requires arm extension
for the 7th and the 8th I/O drawer’s positions at the top of the driver. Currently, there is no recommended batch size for
the operators to follow. Therefore, the operators batch different number of I/O drawers per frame depending on their
availability.
Fab test process is characterized by high energy consumption due to the prolonged test durations. Furthermore, the
test process requires lengthy set-up time even for one I/O drawer. The test time consists of a set-up time of 24 hours and
about 7% extra hours for any further I/O drawers. Figure 1 represents a high level flow chart of the Fab test process,
which shows the main operations of the Fab test and the number of operators and test cells for each operation. Fab test
of I/O drawers is performed in batches in order to save the time of testing per unit I/O drawer. In this study, simulation
modeling is used to analyze the I/O drawer assembly and test process with the focus on batching the I/O drawers on