CHAPTER 9
Statistically-
Based Quality
Improvement
for Variables
Chapter
Overview
• Statistical Fundamentals
• Process Control Charts
• Some Control Chart
Concepts for Variables
• Process Capability for
Variables
• Other Statistical
Techniques in Quality
Management
• Statistical Thinking
• A decision-making skill
demonstrated by the ability to draw
to conclusions based on data.
• Based on three concepts
• All work occurs in a system of
Statistical interconnected processes.
• All processes have variation (the
amount of which tends to be
Fundamentals underestimated).
• Understanding and reducing
variation are important keys to
success.
• Guides people to make decisions
based on data, which needs to be
done in business.
Why Do Statistics Sometimes
Fail in the Workplace?
• Many times, statistical tools
Statistical do not create the desired
result, because many firms
fail to implement quality
Fundamentals control in a substantive way.
• They prefer form over
substance.
Reasons for Failure of Statistical Tools
• Lack of knowledge about the tools;
therefore, tools are misapplied.
• General disdain for all things
mathematical creates a natural barrier to
the use of statistics.
Statistical • Cultural barriers in a company make the
use of statistics for continual
improvement difficult.
Fundamentals • Statistical specialists have trouble
communicating with managerial
generalists.
• Statistics generally are poorly taught,
emphasizing mathematical development
rather than application.
Reasons for Failure of Statistical Tools
(continued)
• People have a poor understanding of
the scientific method.
• Organizations lack patience in
Statistical collecting data. All decisions have
to be made “yesterday.”
• Statistics are viewed as something
Fundamentals to buttress an already-held opinion
rather than a method for informing
and improving decision making.
• People fear using statistics because
they fear they may violate critical
statistical assumptions.
Reasons for Failure of Statistical Tools
(continued)
• Most people don’t understand random
variation resulting in too much
process tampering.
Statistical • Statistical tools are often reactive
and focus on effects rather than
causes.
Fundamentals • People make mistakes with statistics,
because of Type I error (producer’s
risk) and Type II error (consumer’s
risk). These erroneous decisions can
result in high costs or lost future
sales.
• Understanding Process Variation
• All processes exhibit variation.
• Some variation can be controlled
and some cannot.
Statistical • Two types of process variation
• Random
Fundamentals • Nonrandom
• Statistical tools presented here
are useful for determining
whether variation is random.
• Random variation
• Random variation is centered around a
mean and occurs with a consistent
amount of dispersion.
• This type of variation cannot be
Statistical
controlled. Hence, we refer to it as
“uncontrolled variation.”
• The statistical tools discussed in this
Fundamentals chapter are not designed to detect
random variation.
• Nonrandom variation
• Nonrandom or “special cause” variation
results from some event.
• The event may be a shift in a process
mean or some unexpected occurrence.
Statistical
Fundamentals
(random variation)
Statistical
Fundamentals
(nonrandom
variation)
Process Stability
• Means that the variation we observe in
the process is random variation
(common cause) and not nonrandom
variation (special or assignable causes).
• To determine process stability we use
Statistical process charts.
• Process charts are graphs designed to
Fundamentals
signal process workers when nonrandom
variation is occurring in a process.
Sampling Methods
• Process control requires that data be
gathered.
• Samples are cheaper, take less time
and are less intrusive than 100%
inspection.
• Sampling Methods
• Random samples
• Randomization is useful because it
ensures independence among
observations. To randomize means
to sample in such a way that every
piece of product has an equal chance
Statistical
of being selected for inspection.
• Systematic samples
• Systematic samples have some of
the benefits of random samples
Fundamentals without the difficulty of
randomizing.
• Sampling by Rational Subgroup
• A rational subgroup is a group of
data that is logically homogenous;
variation within the data can provide
a yardstick for computing limits on
the standard variation between
subgroups.
Process Control Charts
Statistical Process Control Charts
Tools for monitoring process variation.
The figure on the following slide shows a process control chart. It has
an upper limit, a center line, and a lower limit.
Variables and Attributes
To select the proper process chart, we must differentiate between
variables and attributes.
A variable is a continuous measurement such as weight, height, or
volume.
An attribute is the result of a binomial process that results in an
either-or-situation.
Process Control Charts
Process
VARIABLES
X (process population average)
X-bar (mean or average)
Control R (range)
MR (moving range)
Charts S (standard deviation)
ATTRIBUTES
P (proportion defective)
np (number defective)
C (number conforming)
U (number nonconforming)
Process Control Charts
Central Requirements for Properly Using Process Charts
1. You must understand the generic process for implementing
process charts.
2. You must know how to interpret process charts.
3. You need to know when different process charts are used.
4. You need to know how to compute limits for the different
types of process charts.
A Generalized Procedure for Developing Process Charts
• Identify critical operations in the process where
Process
inspection might be needed. These are operations in
which, if the operation is performed improperly, the
product will be negatively affected.
• Identify critical product characteristics. These are the
Control
aspects of the product that will result in either good or
poor function of the product.
• Determine whether the critical product characteristic
is a variable or an attribute.
Charts • Select the appropriate process control chart from
among the many types of control charts. This decision
process and types of charts available are discussed
later.
• Establish the control limits and use the chart to
continually monitor and improve.
• Update the limits when changes have been made to the
process.
Understanding Control Charts
• A process chart is nothing more than an
application of hypothesis testing where
the null hypothesis is that the product
meets requirements.
Process
• An X-bar chart is a variables chart
that monitors average measurement.
• Control charts draw a sampling
Control
distribution rather than a population
distribution.
• Control charts make use of the central
limit theorem, which states that when
Charts
we plot sample means, the sampling
distribution approximates a normal
distribution.
Process Control Charts
• X-bar and R Charts
• The X-bar chart is a process chart used to monitor the average of the characteristics being measured.
• To set up an X-bar chart
• Select samples from the process for the characteristic being measured.
• Then form the samples into rational subgroups.
• Next, find the average value of each sample by dividing the sums of the measurements by the sample size
and plot the value on the process control X-bar chart. The R chart is used to monitor the dispersion of the
process. It is used in conjunction with the X-bar chart when the process characteristic is a variable.
• To develop an R chart
• Collect samples from the process and organize them into subgroups, usually of three to six items.
• Next, compute the range, R, by taking the difference of the high value in the subgroup minus the low
value.
• Then plot the R values on the R chart.
Process Control Charts
Interpreting Control Charts
Before introducing other types of process charts, we discuss the
interpretation of the charts.
The figures in the next several slides show different signals for
concern that are sent by a control chart, as in the second and third
boxes.
When a point is found to be outside of the control limits, we call this
an “out of control” situation.
When a process is out of control, the variation is probably no longer
random.
Control Chart Evidence for Investigation
Implications of a Process Out of Control
If a process loses control and becomes
Implications nonrandom, the process should be
stopped immediately.
In many modern process industries
of a Process where just-in-time is used, this will
result in the stoppage of several work
Out of stations.
The team of workers who are to address
Control the problem should use a structured
problem solving process.
Some Control Charts
Concepts for Variables
Correction action steps :
• Carefully identify the problem.
• Form the correct team to evaluate and solve the
problem.
• Use structured brainstorming along with fishbone
diagrams or affinity diagrams to identify causes of the
problem.
• Brainstorm to identify potential solutions to problems.
• Eliminate the cause.
• Restart the process.
• Document the problem, root causes, and solutions.
• Communicate the results of the process to all personnel
so that this process becomes reinforced and ingrained in
the organization.
• How Do We Use Control Charts to Continuously
Some Control Improve?
• One of the goals of the control chart user is to
Charts reduce variation.
• Over time, as processes are improved, control
Concepts for limits are recomputed to show improvements in
stability.
Variables • As upper and lower control limits get closer
and closer together, the process improving.
• Two key concepts:
• The focus of control charts should be on
continuous improvement.
• Control chart limits should be updated only
when there is a change to the process.
Otherwise any changes are unexpected.
Process Capability
Process Stability and Capability
• Once a process is stable, the next emphasis is to ensure that the process is capable.
• Process capability refers to the ability of a process to produce a product that meets
specifications.
• Six-sigma program such as those pioneered by Motorola Corporation result in highly capable
processes.
Process Versus Sampling Distribution
• To understand process capability we must first understand the differences between
population and sampling distributions.
• Population distributions are distributions with all the items or observations of interest
to a decision maker.
• A population is defined as a collection of all the items or observations of interest to a
decision maker.
• A sample is subset of the population. Sampling distributions are distributions that
reflect the distributions of sample means.
Capability Studies
• Now that we have defined process capability, we can discuss
Process how to determine whether a process is capable. That is, we
want to know if individual products meet specifications.
• There are two purposes for performing process capability
Capability studies:
1. To determine whether a process consistently results in
products that meet specifications
2. To determine whether a process is in need of monitoring
through the use of permanent process charts.
• Process capability studies help process managers understand
whether the range over which natural variation of a process
occurs is the result of the system of common (or random)
causes.
Capability Studies (continued)
Although different firms use different benchmarks, the
generally accepted benchmarks for process capability are 1.25,
1.33, and 2.0.
We will say that processes that achieve capability indexes (Cpk)
of:
1.25 are capable
1.33 are highly capable
2.0 are world-class capable (six sigma)
1. Select a critical operation. These may be bottlenecks, costly
steps of the process, or places in the process where problems
Five steps in have occurred in the past.
2. Take k samples of size n, where x is an individual observation.
performing • Where 19 < k < 26
• If x is an attribute n > 50, (as in the case of a binomial)
process • Or if x is a measurement 1 < n < 11
(Note: Small sample sizes can lead to erroneous conclusions.)
capability 3. Use a trial control chart to see whether the process is stable.
studies: 4. Compare process natural tolerance limits with specification limits.
Note that natural tolerance limits are three standard deviation
limits for the population distribution. This can be compared with
the specification limits.
5. Compute capability indexes: To compute capability indexes, you
compute an upper capability index ( Cpu), a lower capability index
(Cpl), and a capability index (Cpk). The formulas are:
Cpu = ( USL - μ ) / 3ˆ
Cpl = ( μ -LSL ) / 3ˆ
Cpk = min Cpu, Cpl
where USL = Upper specification limit
LSL = Lower specification limit
μ = Computed population process mean
= Estimated process standard deviation = ˆ = R d 2
The Difference Between
Capability and Stability
• Once again, a process is capable if individual products
consistently meet specifications.
• A process is stable if only common variation is present in
the process.
• It is possible to have a process that is stable but not
capable. This would happen where random variation was
very high.
• It is probably not so common that an incapable process
would be stable.
THE END