An Analysis Of The Dynamics Of Work In Process

An Analysis on the Dynamics of Work in Process
(WIP) in a High Volume Manufacturing System of a
Semiconductor Company Using System Dynamics
Dennis T. Beng Hui
Department of Industrial Engineering, College of Engineering
De La Salle University-Manila, 2401 Taft Avenue, Malate, Manila
dennis.benghui@dlsu.edu.ph

Abstract – High volume manufacturing (HVM) typically uses a The focus of this paper is on the manufacturing processes
push system to prioritize throughput, but inadvertently promote of a “slider” which is a critical component in the head of a
the growth of WIP. Part of an HVM is the presence of re- storage device. The slider allows the storage device to perform
entrants resulting to the explosion of outputs coming from a the read and write functions. The manufacturing process of a
single input. Such is the case of the Slider Manufacturing Process.
slider involves more than a hundred processes starting from
Keywords— Dynamic Work-in-Process, System Dynamics, the cutting of the wafer into Quad to a Row Bar up to a Slider.
Simulation, High Volume Manufacturing Figure 2 shows the conversion from Quad to Slider.

I. INTRODUCTION Fig. 2 Conversion from Quad to Slider

A High Volume Manufacturing (HVM) set up is defined
in this paper as a manufacturing process engaged in the
production of almost homogenous set of products where the
production network is composed of “nested loops of re-entrant
processes and low yield due to imperfect equipment”[1].

These manufacturing processes are typically used to make When cutting the Quad into row bars, the Quad undergoes
products that are considered fundamental consumer needs and a re-entrant process where each row bar in a Quad undergoes a
thus expected to match the output of a traditional mass lapping and cutting process. Similarly, the Row Bar
manufacturing system. One type of product that an HVM set eventually goes through a re-entrant process where each Slider
up is compatible to is the need to manufacturing and assemble is cut from the row bar. The paper will not discuss the
data storage devices. A few of the known companies in the technical properties and operation in the transformation of the
Philippines involved in manufacture/assembly of data storage Quad to the slider. Suffice it to say, that re-entrant activities
devices are Hitachi GST Philippines (HICAP), TDK-Fujitsu, are built into the Slider Manufacturing Process especially in
and Toshiba Philippines. Figure 1 below shows a typical the slicing/cutting of the Row Bar from the Quad, and the
design of a storage device (www.visualdictionaryo.com). Slider from the Row Bar. The re-entrant processes are shown
in figure 3.

Fig. 3 Conversion from Quad to Slider

III. CAUSAL LOOP DIAGRAM

More than the physical and technical properties of the
Fig. 1 Example of Storage Device Component manufacturing processes are the observed relationships
II. SLIDER MANUFACTURING SET-UP between the following indicators critical in the management of
the slider manufacturing process or any mass manufacturing

system for that matter. These are Work-in-Process (WIP),
Throughput (TP), and Lead Time (LT). The causal Another complicated condition in the Slider
relationship of the following variables can be seen in figure 4 Manufacturing Process is the presence the explosion of
below. components from a single input item resulting from the re-
entrant activities. This adds into the dilemma of reinforcing
+ growth in our WIP.
Release rate of
jobs/materials to In practice, it can be observed that WIP is often less
production
Gap (Release rate - prioritized over throughput. This common perception of
Output rate)
- +
+ prioritizing WIP over TP often leads to the practice of pull
manufacturing, where items are always loaded into the
manufacturing process [4] regardless of the WIP.
Actual Average
Throughput (Output -
Average WIP in system
rate)
(Jobs in the system) IV. STOCK FLOW DIAGRAM
-

+ A system dynamics model is constructed using the basic
Average Leadtime (from
release time to completion + Process Utilization
stock and flow diagram (SFD) introduced by Jay Forrester
time) (% busy) from MIT. The resulting SFD for the high volume
manufacturing with re-entrants and explosion of outputs are
shown in figure 5.
Fig. 4 Causal Relationship of WIP, LT, and TP

The causal relationship above shows the presence of 3
positive feedback loops, which indicates that the value of WIP
will be accumulating. This means that WIP is expected to
build up in any manufacturing process. This concept basically
follows the principle of Little’s Law [2].

L = λW
Little’s Law states that L is average amount of items in
queue, W are the average waiting time per item, and Lambda
is the arrival rate of the units. The same Little’s law has been
used to explain the relationship between the three variables
shown above [3]. The equivalent notation is expressed as Fig. 5 Stock Flow Diagram for 3 Process Level HVM
follows: with re-entrants.

WIP = TPxLT The re-entrants in the SFD were modeled using a delay
command to indicate that the entire output of a single input
thus , will be completed upon the completion of the re-entrants. Re-
entrants are expressed as time based units in order to reflect
WIP the period when the re-entrant processes have been completed
TP = and thus the transfer of outputs as well.
LT
When the re-entrant process is on-going, partial output
Where L = WIP, Lambda = TP, and W = LT. items are already transferred to the next WIP level, thus
promoting a push like system as what is happening in practice.
However, in practice, WIP although accumulates The basic form of managing the WIP is the inclusion of a
eventually goes down. What exists is the presence of flushing activity represented by the regulated release rate
adjustments in releasing work into the manufacturing process. situated at the beginning of the manufacturing process.
The Release rate, in this case the release of Wafer into the
Slider Manufacturing Process, is being regulated based on the Each WIP level represents the buffer area in between
expected slider output against that of the actual output. processes, while the process rates represent the processing
areas where inputs are converted into multiple outputs.
Thus, the unbridle accumulation of WIP is in a way being
regulated.

V. SIMULATION RESULTS B. Asynchronous Re-entrant Processes

The resulting SFD were analyzed by changing the The basic concern right now is “what makes this type of
parameters of the model and running the simulation for a total manufacturing process behave in an oscillating manner?”
of 100 hours.
This can be observed when the presence of delay across
A. Uniform Re-entrant Processes the three processing is asynchronous. This means that the
highest delay values are at least 3 times larger than that of the
When the re-entrant processes are small and relatively lowest delay values.
similarly in ratio across the three processing rates, the system
tends to achieve equilibrium over time because of the presence This means that delays are interpreted to not only
of a regulated release rate which was not captured in the represent re-entrant activities, but delay in the handling of
causal loop diagram. This is shown in the figures below. information from the processing area to the release rate. The
following figure below shows the oscillating pattern when one
of the processing rates experience 3 times delay over the rest
of the processing rates.

Fig. 6 Pattern of WIP from three buffer areas

Fig. 8 Oscillating Pattern of WIP from three buffer areas

Fig. 7 Pattern of Release and Processing Rates

The results show that changing the processing time,
WIP levels, and other operational parameters results to a
leveling of performances across the WIP and Throughput.

Lead Time was expressed as processing time and that
the processing time is interpreted as the total processing
time of a given input, which includes the processing of
multiple outputs. Re-entrant time is represented as delay Fig. 9 Oscillating Pattern of Processing Rates from three
and is also the same as the multiple outputs for each buffer areas
process rates. It can be noticed that the release rate has
lower value compared to the processing rates since the It can be seen that the oscillations are more prominent and
input value is actually factored by a multiplier that would that equilibrium will take longer to achieve even when the
represent the multiple outputs in each of the process rates. expected output per period remains the same.

VI. CONCLUSION

High Volume Manufacturing is thought of as a simple
and straightforward environment for managing WIP. However,
it seems that from the system dynamics simulation that even
by isolating the other parameters of the manufacturing
environment, except for the presence of re-entrants and
information delays to the release rate (which acts as a
regulator of the process) do make a significant impact on
whether the process will oscillate or not.

The significance of the dynamic nature of WIP in a high
volume manufacturing set-up is largely influenced by the
presence of unbalanced re-entrant process and untimely
production and WIP information which affirms the problems
that are currently being experienced by these semiconductor
companies.

References:

[1] Y. Nonaka, Y. Suginishi, A. Lengyel, M. Ono, and K.
Sugimoto, "TSUNAMI Effect Prediction Methodology for
Critical Resource Analysis," in Manufacturing Systems and
Technologies for the New Frontier, 2008, pp. 337-340.
[2] M. F. Ramalhoto, J. A. Amaral, and M. T. Cochito,
"A Survey of J. Little's Formula," International Statistical
Review / Revue Internationale de Statistique, vol. 51, pp. 255-
278, 1983.
[3] W. J. Hopp and M. L. Spearman, Factory Physics:
McGraw Hill, 1995.
[4] M. L. Spearman and M. A. Zazanis, "Push and Pull
Production Systems: Issues and Comparisons," Operations
Research, vol. 40, pp. 521-532, 1992.

An Analysis Of The Dynamics Of Work In Process

More Related Content

Viewers also liked (6)

Similar to An Analysis Of The Dynamics Of Work In Process (20)

More from De La Salle University-Manila (20)

Recently uploaded (20)

An Analysis Of The Dynamics Of Work In Process