thoughts on designing a new plant
DESCRIPTION
designTRANSCRIPT
Thoughts on designing a new plant
Being with Dr. Goldratt and a managements team from Dr. Reddy’s
By Eli Schragenheim.
When I showed up at Eli’s home in Holland, he asked what is the purpose of me
coming to the meeting. He did not expect anything new about designing plants would
come out from the meeting. I said there is nothing in writing about the topic. He
agreed to that.
From my perspective the topic of the TOC contribution to designing new plants is still
to be verbalized. I think that the new development in the TOC thinking, especially
the global strategic scope, is making a certain difference. While the meeting had been
focused on the specific needs of Dr. Reddy’s more generic and important insights
have been raised, at least for me. I like to share them with you.
Dr. Reddy’s is a manufacturer of both generic drugs and APIs, the active materials in
drugs that are produced for both internal use and also for other manufacturers.
The meeting itself took two days, but Eli has used much of the time to stress more
generic topics. The article “Standing on the shoulders of giants” was a key
introduction to the issue at hand. We deal here about designing a plant and the focus
on the FLOW has to be very clear.
When we began dealing with the specific subject: designing a plant for generic drugs,
it started with the usual way where the manager in charge of the design put on the
table a pretty detailed plan of the plant.
Personal remark: this is always a problematic obstacle for me. As a TOC consultant I
always strive to see the forest, not the trees. However, most of the people we talk to
are buried in the details too deep to be able to present a clear simple picture. This is
why I detest doing any “plant tours” and prefer asking very generic questions in an
office setting, where people can draw a simple picture on a paper or a board and then
verbalize the real problem. I was keen to see how Eli handles the situation.
Eli did it very smoothly. He did not move to see better the plan with all its tiny
details. He sat in the armchair and explained that he first wanted to make sure of
some basic facts.
What type of products is the new plant for? Dr. Reddy’s makes both drugs and APIs
and these are quite different production environments. The team answered that it is for
the drugs, “the tablets” as the team from Dr. Reddy’s described. This means that the
overall strategy is make-to-availability. This is a key basic fact, because make-to-
availability requires substantial protective capacity even on the CCR.
Now came a discussion on where should the CCR lie?
Personal opinion: I never liked the question of “where the constraint should be?”
because I think that the location of the CCR within the shop floor is only one of the
parameters and not even the main one. Well, when we speak about dedicated
production-line (which is relevant for the specific case) the location of the CCR is
more important than in the other case.
Eli raised several important parameters. He explained that in a line situation the space
between every adjacent machine is limited and thus when a machine is down both the
downstream operations and also the upstream operations have to stop pretty quickly
(the upstream because there is no space to put the output). This situation causes the
need for more protective capacity from an operation the further it is from the CCR. In
other words the required capacity from the CCR towards the start and finish looks like
a funnel.
This means that assuming that the cost of capacity is about the same through the line
the best location for a CCR is in the middle – minimizing the total cost of capacity for
the same level of output.
As said the above refers to a production-dedicated-line where the space between
operations is limited. What the drawing does not show is the two space buffers
required around the CCR in such a line. The buffer before the CCR (not required
when the CCR is at the very start of the line) ensures the CCR won’t stop due to very
short downtime upstream. The space buffer after the CCR ensures the CCR can
continue to work even when a downstream operation is down.
An important insight concerning production lines: Only when the CCR stops to
produce the line itself stops to produce. As long as the CCR continues to work – the
line continues to work, even when one or more of the other stations along the line
stops. By the way, this is certainty true in a regular shop floor where the space
between work centers is not limiting the flow. In production lines it is less obvious,
and some lines are made fully automatic to the extent that whenever one of the
stations stops all the line stops. This is a huge mistake in the design.
The placement of the CCR within the production line (or the whole shop floor) is
hardly the only parameter. What every planner of new plants must realize is that the
uncertainty regarding the future product mix is very significant. Certainly this is the
case with Dr. Reddy’s. First, the demand per product might change (the second
instability defined in Standing on the Shoulders of Giants). Second, new products
enter Dr. Reddy’s production at a rate of about 20% of the current SKUs every year,
and this rate is increasing due to the CCPM being implemented at the R&D of Dr.
Reddy’s. An interesting point is that the processing times cannot be nicely predicted
ahead of time. Even at the proto-type production the processing times are still not
really indicative of what would be the processing times once mass production starts.
All in all it seems clear that when product mix changes it is not certain what resource
would emerge as the CCR. If it is possible to find a resource with a very narrow band
of processing times of different SKUs, it’d nice to make this resource the CCR
(providing the other parameters are not against choosing this resource). However, we
learned that in Dr. Reddy’s this is not the case.
This means that the planning has to ensure very high level flexibility, so any change
in product mix will be handled by the plant, still ensuring very smooth and fast flow.
The management team of Dr. Reddy’s, having already good knowledge of TOC, was
ready to present the main dilemma in the planning – how many lines to design
according to “product – lines” and how much capacity to have as “process design”.
Generally speaking the chemical processes are divided into three main processes (the
first could be further divided into three sub-processes, and in this case we’ll have five
main processes, but we did not pursue this in the discussion). We can make automatic
lines where the output of the first process would flow to the next process. Or it is
possible to have the output of every process be poured into a tank, which can be
stored, then moved to the next process.
Let’s see the pros and cons of these designs.
Product line construct means the fastest flow. All the line is working on one SKU at a
time. It can be set to any SKU and any quantity, but as the whole line is working
according to the pace of the slower process within the line, there are certain SKUs that
fit better the specific line. Also, the construct of the line fits much better the fast
runners. The regular plan is to dedicate the line for certain SKUs. For instance, one
line could be dedicated to two or three SKUs that each of them would run
continuously for two weeks.
The process structures means several machines, in different sizes (capacity wise), for
the particular process are grouped together. This seems to fit better the smaller
quantities as this design gives a lot of freedom in choosing on which particular
machine to put the order on. At the end of the process the output is put into a tank
that is moved ahead to the area where the machines of the next process are located.
Again the freedom of choice helps to move the order ahead while achieving good
capacity utilization.
A clear advantage of the product line design is that it is free from possible
contamination, which is a real concern in the process design. Every time material is
moved it is prone to contamination. Moves of people and tanks are definitely
problematic due to the risk of being contaminated.
So, how do you balance between the two possible designs?
It is interesting to watch how certain patterns of thoughts are kept unchallenged for
quite some time. A simple pattern is to believe that a product line is much better
tuned to mass production. Well, the line probably costs more to build than just
purchasing the required machines and organize them within the process design, but
why cannot the product line be used for short runs?
More, when the line is dedicated to more than just a single SKU then we get relatively
long replenishment time equal to the cycle time, the time until the same SKU will be
produce again. The TOC approach does not like long replenishment times, especially
those that are long without any real need. Production batches should be much shorter
than the 2-week batch, which is assumed as a typical batch for using a line layout, to
allow for the maintenance of very low stocks.
Here is an important insight to bear in mind. We need to think about the system as a
whole, not just the cost of capacity of the plant. It makes no sense to save some cost
in the build of the plant and pay so much more in the need to hold large inventories of
products, for which the demand is far from being stable for long time.
Realizing that even the lines would produce relatively small batches opens the mind
to re-assess the situation. The huge advantage of the line is that it provides the fastest
flow. What seems to be a disadvantage is that it could waste quite a lot capacity due
to the variety of processing times. Suppose that for a certain SKU the pace of the first
process is 2 tons per hour, the second process can do 10 tons in an hour and the third
process can do 4 tons per hour. If the batch size were 20 tons, processing it as a line
would take 10 hours, as dictated by the first process, which is the slowest for this
SKU. Process 2 would then “waste” 8 hours as it could process the whole batch in
two hours. Another SKU could pose different timings like: 10 tons per hour in the
first process, while the second and the third can do only 5 tons per hour.
For a specific product mix it could be the case that the “waste” for one of the three
processes is too much and that process turned out to be a bottleneck.
The regular approach would be to have several lines and then a process area
consisting of the sub-areas, each dedicated to machines for a specific process, with
appropriate storage area in between. Using this layout would minimize the “waste of
capacity” as each machine would be able to process at its full speed.
Eli suggested to go all the way with the “flow-preferred” layout of lines, but to add
the technological option of feeding a tank, rather then the next process, being able to
seal the tank and move it either to a storage place or to the next process at another
line. In order to keep the risk of contamination very low, the tanks should not be
moved through the lines and not through space that is manned by people. The
solution suggested by Eli is to have a system of conveyers higher up the production
floor. So, when a process finishes to fill up a tank, the tank is sealed and goes up.
Then it is moved above where it should be and only then it should move down.
In the discussion that follows another optional solution was raised: to build the lines
vertically (!) – process 1 should be at the top and the output goes down to process 2
and then to process 3 at the bottom. Then the move of the tanks can be mainly
horizontally.
The purpose of this scheme is to capitalize as much as possible on the production line
to flow the material as fast as possible. Eli assessed that in 90% of the cases there
will be an available line to do the batch, and only in about 10% of the cases the batch
would start at one line and then be completed in another line (theoretically it could
have that one batch would pass through three lines – each for just one process).
Planning the shop floor as consisting of many product lines, each containing the main
chemical processes, requires more space then if we would group the machines of each
process. The estimate is that it would require 25% additional space for the same
amount of capacity. Still, the value of the flexibility it gives is very high.
Eli mentioned that to operate this kind of shop floor three different software systems
must be in place:
a. Making recommendation on which line should an order be processed.
b. Handling the move of the tanks through the shop floor.
c. Supporting the POOGI. Not just the Pareto List based on the reasons for
entering the red zone. A wider approach to note problems (quality and safety
as well as obstacles to the flow) and preparing the inputs for analysis.
Another question was raised by the team. Between any subsequent processes a
certain quality check is required at the beginning of the processing. From touch-time
point-of-view the check of a batch is a matter of several minutes. But, temporary
peaks on the checking might cause delays of several hours and considerable waste of
capacity, causing long wait time of orders for the resources that were delayed due to
the checks and now face a temporary peak of load.
Personal observation: An operation like a quality check, even when, like in this case,
it is done by a dedicated resource, usually does not appear in the routing. Same thing
happens to the use of forklifts that are required to move the parts between work
centers. Why these operations do not appear in the routings? It not because of
technical difficulties to include them in the routing. Is it because they seem not to add
any value to the product itself? If this is the reason then it is grossly flawed, because
from the perspective of the flow they are quite critical. Lack of capacity of these
resources causes quite a lot of chaos, which is difficult to analyze as this sort of
capacity is not even registered in the operational database.
Eli Goldratt said very clearly that it is imperative to have A LOT of spare of these
quality check people. This is relatively CHEAP resource and it is stupid to tamper the
flow because of lack of enough people at the quality check. According to Eli the
average load on these people should be 33% - meaning about 2/3 of protective
capacity to maintain smooth flow.
So, after all these thoughts and guidelines, what should be the CCR? As reality could
change the product mix very fast and quite substantially, and as a change in the mix
could impact the required capacity of each of the main processes, it is futile to decide
upon one of the three main processes.
There is one additional operation that comes after the three chemical ones: the
packaging. Every tablet produced through the three main processes might be
packaged in several ways (including different languages printed on the package).
This is the last operation and it is a divergence point where higher number of end-
products are generated. For the case at hand it is convenient to have the packaging
line to be the weakest link.
We already mentioned that producing to availability necessitates that the weakest link
would not have load in access of 90%. The load control should warn when the total
load on that resource is above 80%, so the appropriate actions either to add capacity
or restrain the rising demand would be taken.
The three processes should have higher level of capacity. It is impossible to put a
clear line here, as we do not even have a reference product mix to base the capacity
calculations. Then, we need even more capacity from the production of the materials
for the drugs – meaning the API plants.
Goldratt and Dr. Reddy’s team have dedicated time also to analyze the flow at the
API plants. In the end it was decided that dealing now with guidelines to planning a
new plant for the API is premature. As the implementation of SDBR in the API
plants did not truly started, and the current choke of the release is based on the
capacity of the first operation, which is definitely NOT the CCR (and not what they
want to be the CCR, which is the Clean Room where the finished quality checking
and testing is done) and thus there is a lot of room to gain additional capacity just by
implementing the SDBR in the current plants of API.
Let me summarize my own understanding of the insights for planning a new plant:
1) An underlining critical concept is the emphasis on providing fast and smooth
flow. 2) Evaluating the uncertainty of the future product mix. The factors to evaluate
are:
a) The relative fluctuations of process times between SKUs in resources that
might become CCRs. Comment: While theoretically every resource might be
the capacity constraint we assume that only very few are reasonable
candidates and only those are considered. b) If there are considerable fluctuations between SKUs then the next relevant
factor is the level of fluctuations of demand between the SKUs. In other
words, what are the changes in the product mix for the existing SKUs? c) How many new products are expected and their impact on the product mix?
3) The need to provide wide flexibility in the future. This is a conclusion from
realizing the overall uncertainty in the product mix and its impact on the identity
of the CCR. 4) Is there really a need to decide what should be the CCR? In the early discussion
about the parameters for deciding it was understood as a critical planning
decision. Certainly in a production-line where it is imperative to provide space
buffers around the CCR one must plan what should be the CCR. However, in the
specific case the convenient decision was to have the real CCR outside of the
production-line. In this particular case the line is short enough and designing
space around the central process is definitely good enough. In the more generic
case, especially when the uncertainty regarding the future mix is substantial, my
personal opinion is that we should recognize that we do not really know apriori.
More, the S-DBR procedures are not so strict regarding the exploitation of the
internal CCR (notably in make-to-availability). Certainly the load control need to
watch the right CCR and base the safe-dates calculation on it, but accommodating
to a shift of the CCR does not require radical changes in the daily procedures. 5) What is definitely important is to recognize what should NOT be the CCR!
Support activities like setups, conveyers and quality checks should have high
excess capacity thus not interfering with the flow.