You are on page 1of 8

CEP March 2011 www.aiche.

org/cep 53
Back to Basics
ix Sigma is a management strategy that combines a
variety of statistical and process tools to provide a
framework for continuous improvement, optimization,
and problem-solving. It is based on data and the premise
that variations in the data, whether they are raw material
properties, product properties, or processing parameters, are
undesirable. It employs simple statistical tools and enables
decisions to be made based on statistical thinking that
is, by understanding the process and the data, and proceed-
ing in a disciplined, focused manner.
The methodology (Figure 1) consists of fve phases that
must be completed in a specifc order for any project
Defne, Measure, Analyze, Improve, and Control (DMAIC).
The problem to be solved or parameter to be optimized can
vary in scope, as can the time required to complete each
phase. Moving from one phase to the next involves passing
a checkpoint (or gate) that serves as confrmation that the
targets of a phase have been met. The key is to complete
each phase diligently, confrming that the data and facts are
clear and verifable before continuing.
Six Sigma is usually implemented at the corporate level
by specially trained and certifed experts referred to as Green
Belts and Black Belts. Most large organizations establish a
company-wide Six Sigma group and train their own prac-
titioners. This permits a more-focused dissemination of the
methodology and provides continuity. Chemical engineers
with no formal Six Sigma training often fnd the tools and
techniques valuable as well.
This article introduces the Six Sigma approach, with a
focus on applying the DMAIC steps to existing chemical
processes and products. The standard protocol has been tai-
lored to include tasks pertinent to most chemical engineering
projects, and the basic toolbox has been consolidated to the
main statistical, graphic, and managerial/process tools most
useful to the chemical engineer. The article illustrates the
concepts and tools discussed through an example involving
a dibutyl phthalate (DBP) process.
Sigma and Six Sigma
The Greek letter sigma (s) has three common uses, the
frst two being integral parts of the third the Six Sigma
s = standard deviation. Standard deviation is a mea-
sure of variation in statistics. For a population of n events or
measurements, where X
represent the actual measurements
Many manufacturing processes are not run as
efciently as they could be. The Six Sigma methodology
provides a foundation for continuous improvement and
for solving quality, operational, and other problems.
George Liebermann
Xerox Research
Centre of Canada
Apply Six Sigma for
Process Improvement
and Problem-Solving
p Figure 1. Each DMAIC phase requires specic actions to be taken.

1. Define
Define the problem or opportunity versus technical
and/or business requirements
Measure the current process or product performance 2. Measure
Search for key factors that have the biggest
impact on performance and determine the
root causes of problems
3. Analyze
Develop and demonstrate solutions to
improve the process or product
4. Improve
Implement and monitor the solution 5. Control
This article is based on an AIChE webinar presented by the author. It
is available in the ChemE on Demand archives at
Copyright 2011 American Institute of Chemical Engineers (AIChE)
54 March 2011 CEP
Back to Basics
and X-bar is the mean of the measurements, the standard
deviation is:
v =
h /
Figure 2 illustrates the concept of standard deviation.
Slightly more than two-thirds of the measurements fall
within one standard deviation of the mean, while more than
99% lie within three standard deviations.
= process capability. The statistical measure of the
ability of a process to meet specifc requirements, process
capability is expressed in terms of sigma quality level
(SQL). SQL is the difference between average performance
and the specifcation limit, divided by the standard devia-
tion. Table 1 translates SQL into a corresponding number
of defects. For example, a process operating at an SQL of 6
could be expected to have 3.4 defects per million opportuni-
ties (DPMO).
Off-specifcation product is a common defect, but
most chemical processes are continuous and their outputs
are characterized by continuous data, such as product
yield. Thus, in the chemical industry, process capability is
expressed as an SQL (or related indexes, such as C
, which
will be discussed later). SQL is measured against a target,
typically the lower and upper specifcation limits (LSL and
USL), as depicted in Figure 3. Table 1 compares number of
defects and process yield for various SQLs.
6s = a management/process methodology. Six Sigma
is a continuous-improvement and problem-solving protocol
that involves measuring success against technical require-
ments established by customer and business targets.
Six Sigma/DMAIC in the chemical industry
Because the Six Sigma DMAIC protocol focuses on
optimization and problem-solving related to current pro-
cesses and products, its main application is in chemical
manufacturing, but it can be applied during process demon-
stration (including at the pilot-plant scale) and pilot-plant
or semi-works interim production as well. It is also useful
in technical support activities, including quality control and
analytical methods, maintenance, and utilities supply, as well
as for work processes that support a chemical process, such
as purchasing, inventory management, and sales/marketing.
Six Sigma is generally used to solve such chemical
process problems as: low yield or too much variation in
yield; long cycle times; product or intermediate quality; and
instrumentation/controls performance. Offine product- or
process-related issues, such as raw material quality, analyti-
cal methods problem-solving and improvement, and better
utilities utilization, have been successfully addressed using
Six Sigma.
The basic Six Sigma/DMAIC approach outlined in the
following sections can be used by almost anyone with a
chemical or technical background.
Phase 1
The frst step in the Defne phase is to devise a Six Sigma
charter (Table 2). This is a detailed plan that summarizes
what needs to be resolved and why. Preparing such a plan
will improve your chances of success.
A charter summarizes key process components. The
information entered into the right column should be concise
and contain as much specifc data as possible. The charter is
an evergreen document updated as work progresses.
+/1 68.27%
6 3 2 1 X +1 +2 +3 +6
p Figure 2. A measurements distribution curve illustrates the mean value
(X-bar) and percent of total measurements at multiple standard deviations.
p Figure 3. Sigma quality level for a continuous process is determined
relative to the upper specication limit (USL) and the lower specication
limit (LSL).

X (Mean)
Sigma Quality Level (SQL) =
Table 1. Sigma quality level is determined by the frequency
of defects in the nal output or the yield of the process.
Sigma Quality
Defects per
million opportunities
2 308,000 69.2%
3 66,800 93.9%
4 6,210 99.4%
5 230 99.977%
6 3.4 99.9997%
Copyright 2011 American Institute of Chemical Engineers (AIChE)
CEP March 2011 55
The second step is to review the current process. Walk
the process to develop an intimate understanding of it. This
will enable you to summarize it using process maps with
various detail levels, which is the next step.
Process fow diagrams (PFDs) and piping and instrumen-
tation diagrams (P&IDs) are widely used in the chemical
industry. The Six Sigma equivalent is a value stream map,
which is often an enhanced PFD or P&ID that identifes all
process steps, material quantities, process parameters (e.g.,
temperatures, pressures, fows), measurement points, sam-
pling points, and cycle times.
More-general process information is contained in a
supplier-input-process-output-customer (SIPOC) map. This
document (Figure 4) provides basic essential information
required for more-detailed project activities, identifes sup-
plier information and key input/output parameters, and helps
to establish elements for a cause-and-effect analysis.
The third step involves identifying the targets for the
process or product that need to be achieved. These key
deliverables (which are available from internal sources) are
often referred to as critical customer requirements. Six Sigma
provides tools to obtain and analyze the voice of the cus-
tomer (VOC) from external or internal customers. The voice
of the business is normally provided by departments dealing
with these matters.
The fourth defne step is to develop a proposed schedule,
including estimates of the amount of time required for each
of the remaining DMAIC phases. The fnal step of this phase
involves identifying and securing the necessary people and
material resources.
While much of this sounds simple, it is surprising how
often these basic steps are not followed. This creates a
project that does not have the right focus, information, or
resources at the start, and as a consequence will face major
Table 2. The DMAIC charter outlines the information needed
before beginning a Six Sigma project.
Opportunity/Problem What is wrong or not working
Opportunity for potential improvement
Potential Impact Why the project should be done
Technical and nancial benets
Goal (Target) What is to be delivered, expressed as a
measurable goal
Scope What is expected to be done and the
boundaries of activity
Schedule Expected timeline for the entire project
and for each phase
Resources People who need to be involved
Material resources required
p Figure 4. A supplier-input-process-output-customer (SIPOC) map provides details about the process. This SIPOC map represents the dibutyl phthalate
(DBP) process used as an example.
Raw Materials Charging
- Molten PA Addition
- Heat-Up, Reaction, Reactor Discharge
Washing Steps
- Neutralization
- Water Washes
Steam Stripping
- Continuous Distillation
Activated Carbon
Treatment and Filtration
- Treatment
- Filtration and Pump to Storage
Input Metrics Process Metrics Output Metrics
Reaction Time
Number of Washes
Stripping Steam Rate
AC Treatment and Filtration Time
Unit Manufacturing Cost
(rework yes/no?)
Phthalic Anhydride
Melt Color
Acid Number
% Assay (GC)
% Assay
Weight of Raw Materials
Reaction Temperature
Acid Number (end of reaction)
NaOH Concentration
Acid Number (after washes)
Stripping Temperature, Vacuum
% Assay (GC)
Specific Gravity
Color (APHA)
Acid Number (mg KOH/g)
Flashpoint (C)
Overall Process Time
Process Time
Rework Time, if any
Supplier A
Supplier B
Supplier C
Supplier D
Supplier E
Supplier F
Local Distributor
Plant Facilities
Phthalic Anhydride (PA)
p-Toluenesolfunic Acid (pTSA)
Activated Carbon
Sodium Hydroxide
Cooling Water
Dibutyl Phthalate (DBP)
Wet Butanol
Aqueous Waste
QC Lab
Bulk Customers
Copyright 2011 American Institute of Chemical Engineers (AIChE)
56 March 2011 CEP
Back to Basics
diffculties. Following these steps carefully can make a
major difference in execution.
Phase 2
In the Measure phase, the frst step is to establish the
important process parameters and product properties that
need to be closely monitored. These include the input
variables (e.g., properties and amounts of raw materi-
als, intermediates, and catalysts), process variables (e.g.,
temperatures, pressures, fows, and residence times), and
output variables (e.g., product yield, product purity or other
property measurements, process throughput, etc.).
One of the most important activities in the measure
phase is to identify the key input (Xs), process (Xs), and
output (Ys) variables. One or more output variables refect
the symptom of the problem that needs to be resolved or
optimized (e.g., low product yield). Key variables are the
inputs (including process inputs) and outputs that could be
important for the problem at hand. These are identifed frst,
during the measure phase. Later, in the analyze phase, these
will be narrowed down to the critical inputs and output
variables, which will be the focus of the problem-solving or
optimization effort.
The cause-effect relationship is demonstrated when the
output variables are a function (or effect) of some input and
process variables: Y = f(X).
The SIPOC map is one tool that helps to identify the
key Xs and Ys, because it contains a reasonable amount of
process detail. The value stream map (e.g., a detailed PFD
or P&ID) is also useful to identify the Xs and Ys, because it
has more-detailed and more-specifc information about the
process. Another useful tool for establishing probable cause-
effect relationships is the Ishikawa cause-effect, or fshbone,
diagram (Figure 5).
The dibutyl phthalate (DBP)
process described in the SIPOC
map in Figure 4 is making a
product whose color is out of
the specifed range. The fsh-
bone diagram for this problem
is derived from known data and
information gathered through
brainstorming and previous
experience. From the SIPOC
and the fshbone diagram, the
key Xs that may affect Y, the
product color, can be identifed.
The cause-effect relationship is
represented as Y
= f(X
, X
where: Y
= DBP color, and
= DBP purity; and the Xs can
be raw materials quality and process and equipment condi-
tions, some of which are shown in Figure 5.
Many problems involve multiple related outputs; in this
case, the color problem (Y
) is most likely a result of impuri-
ties in the product (Y
The next step is to determine which parameters are
most important to measure and what types of measurement
methods, sensors, and instruments will be used. A method or
description of how the measurement is performed is referred
to as operational defnition.
Correct data are critical for sound decision-making, and
a good measurement system is essential to gathering the
correct data. Be sure that your measurement systems are
adequate. Even if a measurement has been in place and is
considered routine, it may be necessary to validate the mea-
surement system.
The next step in the measure phase is to perform a mea-
surement system analysis (MSA). You can determine which
measurements require an MSA based on the importance of
the measurement and previous experience with the method,
or after reassessing the data and the method.
An MSA evaluates the amount of variation associated
with the measurement system and compares that to the
process variation and specifcations. The main MSA tool
for doing this, Gage R&R, assesses the repeatability and
reproducibility of the measurement relative to the process
variations. Repeatability relates to the results obtained by the
same operator on the same sample, whereas reproducibility
relates to the results obtained by different operators.
Variation less than 10% indicates an acceptable measure-
ment system; 10% to 30% may be acceptable, but over 30%
signals an unacceptable measurement system.
Statistics software packages such as Minitab and JMP
have standard MSA programs. Less-expensive Excel-based
RM Addition
Raw Materials
Inert Gas
Bottom Valve
Reactor Seal
Reaction Time
pTSA Color
(color space)
PA Impurities
PA Color
APHA (color)
Lab Bias
Procedures/Execution Equipment Color Measurement
Reactor Discharging
Synthesis Conditions Raw Materials
p Figure 5. A shbone diagram for the DBP process example illustrates the possible causes of off-spec color.
Copyright 2011 American Institute of Chemical Engineers (AIChE)
CEP March 2011 57
versions are available online in statistical process control
(SPC) software packages (13).
The fourth step in the measure phase is to prepare a data
collection plan. This plan (which may take many forms)
should list: the historical measurements and data that need to
be measured and collected, in what form, and for what period
of time; and new data to be collected, the measurement
methods to be used, the form of the data, and the timeline
for collecting them. New data could be measurements taken
previously that need to be repeated on the same samples or
new samples, or new measurements and/or methods.
With the data collection plan in place, you can begin the
actual collection of the data. Six Sigma tools can be used for
organizing data, but the simplest method is to collect and
organize the data in a spreadsheet. Tools such as a control
chart (Figure 6) and a Pareto chart (Figure 7), which are
needed for the analyze phase, are often prepared during the
measure phase to organize the data and perform preliminary
data analysis.
The control chart in Figure 6 summarizes the DBP yield
for more than 70 batches and shows that the average yield
(X-bar) is 81.4%. If, for example, the target yield is 85%,
clearly there is a problem, even though statistically the
process is in control i.e., within the upper (UCL) and
lower control limits (LCL) of 3 sigma. Such a process is
described as in control but not capable, which means that the
results are consistent but not at target.
Figure 7 is a Pareto chart for several input and process
variables monitored during the production of DBP. A Pareto
chart is useful for recognizing the vital few among the
trivial many a reference to the 80/20 Pareto principle
rule of thumb that 80% of the problems are caused by 20%
of the possible causes.
The fnal step of the measure phase establishes a quanti-
tative assessment of the current situation i.e., a baseline
process capability.
A simple way to look at the capability of a process is to
compare the variations in the process with the specifcation
limits. The process capability index, C
, is the ratio of the
specifcation limits (voice of customer) and the control limits
(process variation):
( ) C
2 p
requires both the USL and LSL, and if the mean of
the distribution is not centered, C
will not refect reality.
Thus, the most commonly used measure of capability is C
It is calculated based on a specifcation limit and the mean
(X-bar) of the process data. If both USL and LSL are speci-
fed, use the limit to which the mean is closest (lowest C

3 3
or pk
v v
^ h
8 B
of 1.33, which corresponds to SQL = 4 (or four
sigma), has traditionally been used as the minimum target
for a manufacturing process. A Six Sigma quality level is
equivalent to a C
= 2.
Quantifying the current process capability will help in
assessing future improvements by allowing a comparison
of the C
before and after the implementation of proposed
Phase 3
The objective of the Analyze phase is to determine
the root cause of the problem (or the opportunity) to be
The frst analyze step is a detailed data and process
analysis using some of the Six Sigma tools that are most
p Figure 6. A control chart of DBP yield for 70 lots prepared using Minitab
software shows process yield with the upper and lower control limits, and
the average yield (X-bar).
8 h
p Figure 7. A Pareto chart shows the number of occurences of
variability in raw materials properties and process time.
1 8 15 22 29 36 43 50 57 64 71
Lot Number
UCL = 98.70
X = 81.42
LCL = 64.14

Copyright 2011 American Institute of Chemical Engineers (AIChE)
58 March 2011 CEP
Back to Basics
relevant to chemical engineering work (Table 3).
Many common data-analysis tools such as descrip-
tive statistics, control charts, Pareto charts, and regression
analysis are available in spreadsheets (e.g., Excel) and
can easily be used for Six Sigma projects. Understanding
the magnitude and types of variations, as well as where
and when variation occurs, can lead to the root cause of the
Several hypotheses may be considered when searching
for the root cause of the problem. Good data and process
analysis will help you develop the hypotheses and narrow
them down.
Hypothesis testing has many chemical engineering
applications. The most common questions it seeks to answer
are: Are two sets of data statistically the same or different?
Are the results statistically the same or different after the
introduction of a new raw material or switching to another
vendor or reactor? Do changes that have been made in an
attempt to improve the process lead to a statistically different
set of results?
Several relatively simple methods for hypothesis testing
(described in most statistics books) are:
one-sample t-test, which compares the sample mean to
a historical mean or target
two-sample t-test, which compares the means of two
one-way analysis of variances (ANOVA), which com-
pares the means of more than two samples
two-way ANOVA, which compares two samples based
on multiple factors (means and other parameters).
All of these techniques evaluate the mean(s) of the
sample(s), and address two hypotheses: the null hypothesis,
, that the means of the two populations are equal:
: m
= m
and the alternative hypothesis, H
, that the means of the two
populations are different:
: m
In statistics, sample refers to a subset of a popula-
tion, X-bar is the mean of a sample, and m is the mean of a
population. A t distribution (as in t-test) is a probability
distribution for a relatively small set of samples. ANOVA
employs statistical models and procedures to compare the
means of two or more samples or populations by analyzing
the variances (i.e., the sum of the squares of the total, error,
and samples variances).
Hypothesis testing can be very useful in establishing the
root cause of a problem. The time spent becoming familiar
with these techniques will be well spent.
Among process analysis tools, team brainstorming is
one of the best ways to generate good ideas. Process maps
(SIPOC maps and enhanced PFDs or P&IDs) and Ishikawa
cause-effect diagrams are also useful.
Non-value-added analysis is a basic methodology to
identify process steps and process times that do not yield
any positive effect and do not add value to the process or
product. Beyond process steps and times, it can also be
applied, for example, to process measurements or analy-
ses that may not be required and only add delays or costs.
Lower cost and higher productivity (shorter process time)
are the expected outputs.
The non-value-added methodology scrutinizes each
process element with regard to:
customer value added (CVA). These steps provide a
direct value to the customer (e.g., synthesis and purifcation),
and are candidates for optimization.
business value added (BVA). These steps are required
to complete the process or make a product (e.g., quality
control, maintenance, change-overs, and transfers), and are
also candidates for optimization.
non-value-added (NVA). These activities are not
required to meet customer or business needs. They are a
waste in the process (e.g., rework, storage/inventory, hold-
ing/delays, and extra process time). Try to eliminate these.
You can also identify and evaluate probable causes by
analyzing the risk points in the current process using the
failure mode and effects analysis (FMEA) technique. FMEA
is often also used in the improve phase, which will be dis-
cussed in the next section.
To conduct a FMEA, list all the factors (i.e., the key input
and process Xs) that you have determined to possibly contrib-
ute to the problem in the frst column of Table 4. List possible
failures or errors for each factor in the second column, and
the effects that each failure or error could produce (i.e., the
impacts on the key output variables, Ys) in the third column.
Then, in the fourth column, rate the severity of each poten-
tial failure on a scale of 1 (low) to 10 (high). Next, consider
potential causes of each failure (ffth column) and the exist-
ence of any controls to prevent them (seventh column), and
enter an appropriate rating for each. Finally, calculate a risk
Table 3. Data analysis and process analysis tools help to
determine the root cause of a problem.
Data Analysis Tools Process Analysis Tools
Descriptive Statistics
and Data Distributions
Ishikawa Cause-and-Effect
(Fishbone) Diagram
Control Charts Process Maps (PFD, P&ID)
Pareto Charts Brainstorming
Regression Analysis Non-Value-Added Analysis
Hypothesis Testing Failure Mode and
Effects Analysis (FMEA)
Copyright 2011 American Institute of Chemical Engineers (AIChE)
CEP March 2011 59
priority number (RPN) by multiplying the three ratings. The
factors with the highest RPNs should be further scrutinized.
Additional details on the use of FMEA can be found in
numerous books and websites on Six Sigma. A good place to
start is the FMEA Info Centre (4).
The most important and sometimes the most diffcult
step of the analyze phase is the identifcation of the most
probable cause or causes of the problem. There are no gen-
eral guidelines for this step because each problem is unique
and often very specifc, however many of the tools discussed
here can be used to arrive at the most probable Y = f(X).
If possible, obtain proof of the proposed cause-effect
relationship. Depending on the details of the problem, this
can be done by experimenting at various scales (bench, pilot
plant, or manufacturing) through simple experiments
that vary one factor at a time (OFAT), or through a design of
experiments (DOE) matrix of experiments that vary multiple
factors (input or process variables) to confrm the expected
effect (5).
The analyze phase concludes with the documentation of
all the fndings, including a summary of the rationale and
proof of the root cause(s) of the problem.
Phase 4
After some ideas for solutions have been summarized in
the analyze phase, the emphasis shifts to the Improve phase,
where the focus is on selecting and demonstrating one or more
of the solutions. Although the specifcs will differ from case to
case, the frst step is to generate
and develop potential solutions
i.e., to identify, clarify, and, if
necessary, expand the ideas that
have been proposed.
It is essential to establish
criteria for evaluating the solu-
tions. Consider quality, cost,
and delivery (QCD) as well as
safety factors when developing
specifc criteria relevant to the
issue at hand. These evaluation
criteria and other known infor-
mation will be used in the next step, the selection of the best
solution or a few alternative solutions.
The Pugh matrix is a table-based tool used for select-
ing the best solution after the pool of potential solutions has
been narrowed down to a reasonable number (e.g., fewer
than 10). There are several versions of the Pugh matrix;
Table 5 is an example of the weighted Pugh matrix.
To use the Pugh matrix, list the evaluation criteria in the
frst column; if necessary, the list can be ranked and short-
ened to make it more workable. Rank the importance of
each criterion, on a scale of 1 (low) to 5 (high), in the second
column. Enter the current (baseline) process or product in
the third column, followed by the potential solutions in the
remaining columns. Next, rate the solutions as better than
(+), equal to (0), or worse than () the current baseline for
each evaluation criterion. These can also be weighted (e.g.,
+2) if the effect is expected to be very strong. Then multiply
the importance ranking (15) and the solution rating (+, 0,
or ) for each evaluation criterion to obtain an overall score
for each solution, total them and compare the solution to
the baseline.
This can be done as a standard table in a spreadsheet.
Templates are available on many Six Sigma and related
websites, some for free (6).
The fourth step in the improve phase is risk assess-
ment a key requirement. Although similar to the FMEA
conducted in the analyze phase, an improve phase FMEA
focuses on the proposed changes and their expected impacts.
The impacts include those on the main outputs (Ys) that
Table 4. The FMEA technique allows the engineer to identify and evaluate probable causes of process errors
by analyzing the risk associated with key inputs and process variables.
Key Inputs
and Process
Mode (Error)
Failure Effects
(Impacts on
Key Output
Risk Priority
Table 5. Importance rankings (weights) are assigned to criteria associated with
solutions to a process error in a weighted Pugh matrix. The solutions are rated as
better than (+), equal to (0), or worse than () the current baseline.
Proposed Solutions
Evaluation Criteria
(Quality, Cost,
Delivery, Safety)
Process or
Solution 1 Solution 2 Solution 3 Solution 4
Copyright 2011 American Institute of Chemical Engineers (AIChE)
60 March 2011 CEP
Back to Basics
need improvement, as well as the impacts on other process
or product parameters that may be impacted by the planned
changes in inputs (Xs).
Therefore, the contents of the frst column contain only
the proposed changes, and an action recommended col-
umn is added to address the potential issues.
The fnal step in the improve phase is to demonstrate the
solution, which is required before you can proceed to full
implementation. If the demonstration is unsuccessful, a new
solution will need to be found.
The demonstration of the solution the pilot can be
undertaken at various levels: modeling/simulation, bench-
scale demonstration, pilot plant experiments, or temporary
changes at the manufacturing scale. The pilot can be a trial at
a smaller physical scale (in a pilot plant facility) or a smaller
time scale (in a production facility). It is sometimes worth-
while to perform a DOE for the pilot, although many times
the demonstration is a straightforward confrmation of the
proposed solution(s). The pilot provides the data that con-
frm the expected improvements, reduces the risk of failure,
and may expose additional issues that need to be addressed.
This will make the full-scale implementation more effective.
The pilot results need to be compared to the historical
results to confrm that indeed an improvement was achieved.
Various statistical tools can be used for this:
descriptive statistics and histograms
two-sample t-test to compare the mean of the improved
process with the baseline
ANOVA to compare the mean of the improved process
with two or more sets of baseline results
comparison of the improved process capability (C
with the baseline C
A summary of such comparisons is normally necessary
to obtain the necessary approvals to move to an interim full-
scale implementation in the fnal phase, control.
Phase 5
The last phase in the DMAIC protocol is Control
a systematic full-scale implementation of the proposed
solution(s), which includes developing standard operating
procedures, implementing process changes, monitoring the
improved process, and transitioning the project to the pro-
cess owner.
The development of a control plan, which will be used to
monitor the improved process or product and guide correc-
tive action if necessary, is one of the most important steps in
the control phase. A control plan should contain the follow-
ing elements:
measurements what to measure and the measure-
ment methods
sampling plan at what points in the process, frequen-
cies, and what to measure for each sample
monitoring plan tools for comparing sampling data
to pre-improvement baseline data (control charts, process
capability, two-sample t-test, risk management, etc.)
control plan documents electronic or paper docu-
ments summarizing the collected data.
The fnal step is to assess and confrm the expected tech-
nical and fnancial benefts. This is very important, as it will
highlight the accomplishments and potential advantages of
using a similar approach for other problems or opportunities.
Going forward
Many companies in the chemical and related industries
have embraced Six Sigma and routinely use it in their opera-
tions. Because chemical engineers are data-driven, we are
well-suited to practicing Six Sigma.
Although it has been adapted and implemented in
several forms, the basic methodology, the DMAIC phases,
and many of the tools remain the same. Furthermore, the
approach can be used effectively even by engineers who
have no formal Black Belt or Green Belt training. This
article is intended to be a motivation and starting point for
such an initiative.
GEORGE LIEBERMANN, most recently a Senior Engineering Fellow at Xerox
Research Centre of Canada, is a Xerox Lean Six Sigma Black Belt and a
Black Belt in Design for Lean Six Sigma (E-mail: george.liebermann@ or After a career in the
chemical industry in Romania, Israel, and at Uniroyal Chemical (now
Chemtura) in Elmira, Ontario, he joined Xerox in 1981 and has held
several senior scientic and managerial positions, including process
engineering manager. He contributed to the scale-up and commer-
cialization of many toner, ink, and photoreceptor materials and has
over 40 U.S. patents, 15 of which cover materials or processes used
in Xerox products. He has received many awards, including the Xerox
Presidents Award, the 2007 Ontario Professional Engineers Engineer-
ing Excellence Award, and AIChEs 2006 Process Development Practice
Award and 2009 Lappin Award. He was the founding chair of the AIChE
Process Development Group/Division. He is a Fellow of AIChE, of the
Canadian Academy of Engineering, and of the Chemical Institute of
Canada. He received his PhD in chemical engineering at the Polytechnic
Institute Timisoara in Romania.
Literature Cited
1. Sigmazone, SPC XL Software for Microsoft Excel,
2. QI Macros, Measurement System Analysis Gage R&R,
3. BPI Consulting, SPC for Excel: Statistical Analysis Software,
4. FMEA Info Centre, FMEA Failure Mode and Effects
5. Louvar, J. F., Simplify Experimental Design, Chem. Eng.
Progress, 106 (1), pp. 3540 (Jan. 2010).
6. Lean Software Engineering, Pugh Decision Matrix,
Copyright 2011 American Institute of Chemical Engineers (AIChE)