 |
Web Hosting by Netfirms | Free Domain Names by Netfirms |
|
|
|
Critical Success Factors To
Realizing Maximum Impact From Your RTD Implementation
The
semiconductor industry has enthusiastically embraced the Theory of Constraints
as explained by Eliyahu Goldratt in his book ‘The Goal’. The Theory of
Constraints (TOC) seeks to maximize system throughput and maintain a minimum
level of inventory by effectively managing the constraint resource.
Semiconductor fabs with their complex and reentrant processes, face several
challenges in successfully implementing this theory. Managing a constraint or
bottleneck resource in such processes is resultantly complex.
However
with the introduction of AutoSimulations Real Time Dispatcher (RTD), innovative
new approaches are possible. This paper will describe the characteristics of a
successful approach for managing the constraint resource, particularly the case
of a recurring constraint resource, and will explain why the creation of a
global dispatching policy provides a critical link in the value chain of an
enterprise.
Beyond
the creation of an effective dispatching algorithm, several other factors
critical for a successful implementation will be discussed. These factors
include cultural transition, organizational infrastructure, manufacturing
performance metrics, education, communication, training, business processes,
implementation strategy, and continuous improvement.
The
idea of global optimization is not unique to the manufacturing process, but
should be applied to all of the processes associated with running a business. A
business is, after all, a group of interrelated processes, and the goal of a
business is to maximize the profit derived from those processes. A core discipline
of managing a business for profitability has become known as Supply Chain
Management. Supply Chain refers to the flow of materials and information
required for the manufacture of products and services. The more inclusive term
Value Chain refers to the flow of market information (such as demand) and cash
into the system as well as the flow of products and services out of the system.
Value Chain Management involves accelerating the velocity of business
transacted along the chain, thus reducing the amount of money tied up in
interim process steps and improving customer service. The links in this chain
that contain the largest amount of inventory and consume the most operational
expense (these terms will be defined in a moment) provide the largest opportunity
for increased efficiency. For the semiconductor industry a major portion of the
Value Chain resides within the walls of the wafer fab. This portion of the
Value Chain consists of factory planning and factory floor dispatching. Factory
planning is the process of analyzing the capacity of the fab based on the
forecasted demand and product mix to ensure that the demand can be met. This
analysis is used to establish a capacitated starts plan that will determine at
what rate and in what order lots are introduced into the fab. Once material is
in the manufacturing facility it is the responsibility of the dispatching
system to move WIP in the most efficient manner. Of these two processes, the
factory floor dispatching system generally has the greatest potential to
improve economic profit.
We
know that it is desirable to improve the factory level processes of the Value
Chain, so now we must explore how to do this. The body of knowledge that
characterizes the behavior of this portion of the value chain is called Factory
Physics 1. Two of the most basic building blocks for Factory Physics
are Little’s Law 2 and the Theory of Constraints 3.
Little’s Law establishes a relationship between queue length, arrival rates,
and service times and was originally applied to a single queue or point of
service. In factory level manufacturing terms the Law can be stated as:
WIP = Cycle Time * Start Rate
The Theory of Constraints (TOC) is a broader
concept that has application beyond just manufacturing. In it’s most basic
form, TOC says that the goal of business is to “Increase
throughput while simultaneously reducing both inventory and operational
expense”. Goldratt further supplies the following definitions of terms. The throughput
of a system “is the rate at which the system generates money through sales”. In
other words, value is not realized until the revenue is received. Inventory
is defined as “all money the system has invested in purchasing things that it
intends to sell”. And “operational expense is the money the system
spends turning inventory into throughput” 3. The correlation to
Little’s Law is obvious, by reducing cycle time we can either increase the
start rate or reduce inventory, and both options increase throughput.
Additional benefits of reducing cycle time are the ability to satisfy customer
demand more quickly and more rapid yield ramps by decreasing the feedback time
for process development. The principles of Factory Physics clearly address our
challenge of improving the efficiency of the Value Chain.
So how do we implement the theory? Goldratt summarizes
the principles for implementation of his philosophy as follows:
·
The capacity of a system is dictated by the most
constrained resource, called a bottleneck
For semiconductor manufacturing we must expand this
approach to address the additional issues of recurring bottlenecks. For
example, what occurrence of the bottleneck do you run first? Does it matter?
What is a practical approach for driving inventory to the bottleneck based on
the bottleneck's consumption rate? As far back as 1988, Glassey and Lozinski 4
discussed techniques to detect starvation of the bottleneck. Through the years
Dr. Glassey has described various methods to accomplish this goal. The
solutions have ranged from graphical assistance
for operators to queue predictions based on simulation experiments. Dr. Glassey
has also attacked the problem of regulating the flow of material into the
process flow based on a linear control rule called descending
control 5. These approaches have been influenced by the
unavailability of real time data.
Since
the throughput of the system is dictated by the bottleneck, let us take a
moment to look at the factors that affect it.
·
Process speed – governed by the design of the tool and the process.
Of
these factors, setups and idle due to no WIP can be affected by dispatching
policy. Therefore, the solution addressed by RTD should focus on reducing or
eliminating these detractors. Avoiding
idle time on the constraint tool involves all tool sets in the process while
setup optimization is localized to the constraint tool itself.
Because the semiconductor process is reentrant, the
constraint tools are encountered many times in the process at differing process
rates. Our goal must be to feed the constraint tool and maintain a linear
inventory profile. We do this by recognizing that we must not just feed the constraint tool,
but that we must feed all occurrences of
the constraint tool, and we must do it in a consistent and equitable manner. If
we fail to recognize this, we could be faced with the circumstance where there
are large amounts of inventory in front of the constraint tool, but it is all
bound for a single occurrence of the tool with all other occurrences left dry.
In this case we have created a bubble of inventory in one location and a hole
in the inventory profile in other locations. A reentrant constraint must ensure
that it feeds itself. We can also add to this the concept of maintaining a
minimum buffer in front of each occurrence of the constraint to protect it from
the inevitable disruptions and fluctuations that could impede a constant flow
to the constraint 3.
The
solution presented here to the problem of how to feed the constraint is similar
to a well known dispatching policy called critical ratio. Critical ratio is a
method to drive dispatching decisions based strictly on customer due date, and
is simply the time the lot is expected to take to complete divided by the time
until it is needed to be complete. This ratio of expected/needed time gives a
higher priority to product that is farther behind schedule. This same type of
concept can be used to ensure that material arrives at the bottleneck tools on
time.
If
we consider each occurrence of the constraint an end point, then we can drive
work in process (WIP) to the constraint based on criteria relevant to the
constraint tool. If our goal is to keep each occurrence of the tool from
starving, then we can use the relationship between expected time of arrival and
needed time. The expected time is the time for a lot to get to the constraint
and the needed time is the time that the lot is needed at the constraint in
order to keep it from starving (or depleting a minimum buffer). The more
material that is in the constraint’s buffer or is likely to arrive before the
current lot, the lower the priority of the current lot will be (see figure 1
below). We now have a priority factor based on manufacturing efficiency needs
that can be combined with more traditional drivers based on customer delivery
dates. The combination provides a balanced and synergistic approach to
scheduling.
Figure 1We
have discussed an interesting concept, but without a tool that can accomplish
it, the discussion is academic. Fortunately, Real Time Dispatch represents a
significant new capability in managing the manufacturing process. It provides a
user- friendly interface that gives a nonprogrammer such as an Industrial
Engineer or Manufacturing Engineer the ability to design, evaluate, and
implement the exact functionality needed. The ability to access manufacturing
data on a real time basis allows solutions not possible before. Rather than
attempting to analyze recent conditions to predict future states in support of
a manual decision process, rules can instantly assess the current dynamics of
the manufacturing system and apply high function algorithms to them. This new
level of capability is quickly nullified if used to simply automating existing
rules of thumb. Rules of thumb are by definition only adequate solutions based
on applying simple logic to limited information. Solutions that drive a
competitive advantage should be based on Factory Physics concepts that drive
global optimization and maximize the profitability of the business.
Practical experience has shown that simply designing and implementing a good dispatching rule, even one aimed at global optimization, does not yield successful results. Several additional factors have proven to be equally important in realizing the full benefits of a Real Time Dispatching solution.
The
implementation of a new philosophy for operating the manufacturing facility
will never get off the ground without the commitment and confidence of the
management team. The executive management level must provide visible and
consistent support for the effort.
An
RTD installation should be classified as a mission critical application. This
is a software tool that requires dedicated resources for support of the
hardware and repository as well as for the development and maintenance of the
dispatching rules.
It is crucial that a formal statement be published that describes the management team’s approach to manufacturing. This will ensure that all rules implemented will conform to a common goal. RTD should be viewed as a tool to implement the management team’s vision for efficient manufacturing.
A
consistent set of metrics should be established by which any new rules or rule
modifications should be judged. This will ensure that rules are providing
beneficial results to the fab. The metrics must be aligned with the manufacturing
philosophy. Since the implementation of a rule will directly affect the
productivity of the fab, rules should be tested via simulation to insure that
the desired results are achieved. ASAP provides this capability.
Actual
manufacturing performance should also be measured against metrics that indicate
adherence to the manufacturing philosophy. These metrics should measure
contribution to bottom line manufacturing effectiveness. If these do not
align, there will be no compliance to the dispatch policy!
Managers
have a strong tendency to manage according to the information that is available
to them. Therefore it is very important to provide management reports that
gauge the same metrics that the dispatching policy is designed to influence.
Incentive and performance plans cannot be in conflict with the manufacturing
philosophy.
Even
a very advanced dispatching system cannot be effective if it is consistently
over committed by an unachievable starts plan. The two systems must work in
concert. The dispatching system must be integrated with the factory planning
system in such a way that avoids conflict or overlap. For instance, the factory
planning system should not attempt to schedule product movement within the
manufacturing facility, but rather assess and predict inventory movement in
order to provide an effective starts plan. Conversely, the dispatching system
should not attempt to develop starts plans or predict plant capacity.
Since
this is a mission critical application, a documented procedure should be in
place to maintain good standards and to keep the process effective and under
control. Procedures should define the rule submission, assessment, and
development process. They should define standards for rule robustness,
maintainability, and functionality. Rules must have defined criteria for
approval to ensure global optimization, performance and screen design. A formal
test plan for new rules will ensure quality code, and proper training and
rollout plans will encourage user buy-in and support. Finally, there must be an
open process for feedback on the rules to facilitate continuous improvement.
This
paper has discussed the role of Real Time Dispatcher (RTD) as a powerful
inventory management tool as well as a critical link in the overall value chain
of an enterprise. It also described the
characteristics of a successful approach using RTD to avoid starvation of recurring
bottleneck resources. Other critical success factors were highlighted that
focused on the need for cultural transition, support structures, communication,
education, and continuous improvement. These concepts have been proven in
production environments and are already providing substantial economic returns
for the companies who have implemented them.
Clay Rippenhagen is a Sr. Manager of the semiconductor consulting practice at Avicon, where he is responsible for managing all aspects of defining and implementing semiconductor eValue Chain solutions as well as leading the development of Avicon's Austin, Texas office. Prior to joining Avicon, he was the Industrial Engineering Supervisor at Advanced Micro Devices, responsible for the modeling, simulation and dispatching efforts in Fab 25. Mr. Rippenhagen previously spent 10 years with IBM as an Account Systems Engineer and Industrial Engineer. He graduated with a Bachelor of Science in Industrial Engineering from Texas A&M University, and has spoken at a number of semiconductor industry focused organizations including the Winter Simulation Conference in Washington DC.
1.
Wallace J. Hopp and Mark L. Spearman, “Factory Physics, Foundations of
Manufacturing Management Irwin/McGraw-Hill 1996.
5.
C. R. Glassey and
Jeyaveerasingam George Shanthikumar, and Sridhar Seshadri, “Linear Control
Rules for Production Control of Semiconductor Fabs” IEEE Transaction on
Semiconductor Manufacturing. Vol. 9, No. 4. November 1989.