Statistical Process Control

13
Introduction to Statistical Process Control
The Value SPC Can Add to Quality, Operations, Supply Chain Management,

Process Capability
and Continuous Improvement Programs
Variation and What it Means to be in Control and Capable
Introduction to Control Charts
Hands-on - Control Charts in Minitab -17
Basic Control Charts for Variables
Hands-on - Basic Control Charts for Variables in Minitab -17
Advanced Control Charts for Variables
Hands-on - Advanced Control Charts for Variables in Minitab -17
Control Charts for Attributes
Hands-on - Control Charts for Attributes in Minitab -17
Other control charts
Process Capability
Hands-on - Process Capability in Minitab -17
SPC in Service Sector/ Industries
Case Studies in SPC
Latest trends in SPC including AI
 Discussion Topics

• Process Capability
• Cp
• Cpk, Cpm
• Process Capability - One Sided Specification Limit, Process Capability for Non-Normal
Distributions & Attributes Data

2
 Process Capability

• Process Capability is:

‣ Fitness for use (Juran)
‣ Best for certain customer (defined) conditions (Fiegenbaum)
‣ Degree of excellence with which a product/ service fulfils its intended purpose
(Lawson)
• Customer’s requirements are defined by the customer
‣ Product/ service provider may assist the customer in defining their requirements
• Customer requirements are converted into specifications by either the service provider
or a project management entity (in case of complex products such as ships, aircraft etc)
• However, customer will mostly be interested in the answer to this question: “Did the
product do what I expected to to do?”
3
 Process Capability

In Control but not capable

• Process capability assesses how
well the process, when operating
in control, is able to meet the
specifications
• Process capability is important
because simply being in control is
not sufficient
‣ A process that consistently
produces nonconforming
product can still be in control

4
 Process Capability for Variable Data

• 3 indices are used to compare process performance (control limits) and process
specifications (customer requirements)
‣ Cp
‣ Cpk
‣ Cpm
• Regardless of which index is used
‣ 1st step is to ensure that specifications reflect customers requirements
correctly
‣ 2nd step is to assess that the process is in control
‣ 3rd step is to compare variation of in-control process to spread of specification
using some form of an index 5
 Process Capability for Variable Data

• For processes not in control, sometimes 2 additional indices (Pp, Ppk) are used to
get an initial measure of process capability
• Cp/ Cpk and Pp/ Ppk have identical formulas - the difference being Cpk is calculated
using the WITHIN standard deviation, while Ppk uses the OVERALL standard
deviation
• Since all 3 indices (Cp, Cpk, Cpm) are ratios of specification spread (tolerance/ specs,
specification limits) and process variation (defined by +/- 3σ), only one of the two
actions can improve the ratio:
‣ Decrease process variation
‣ Increase specification spread/ limits

6
 Process Capability for Variable Data

• Higher the value of the index, more capable the process

• Standards for the index value at which the process would be considered capable
varies from industry to industry and org to org
• A generally accepted rule is the Cp=1.33 is the minimum acceptable process
capability value

*The Six Sigma quality program allows the

distribu6on mean to dri8 by ± 1.5 standard
devia6ons. Six sigma quality without the dri8
equates to 0.002 ppm defec6ve. Six Sigma
quality with the dri8 allowed equates to the
o8en quoted 3.4 ppm defec6ve or 3.4
defec6ve parts per million opportuni6es.

7
 Process Capability for Variable Data

Decrease process
variation
Loosen the
specifications
When the process is centered on the
target value of the specification (T),
only two actions can improve
process capability: 8
 Process Capability for Variable Data

If the process is not centered on the process specification, process

capability can also be improved by centering the process on the
specification

9
Why Isn’t Cp=1.00 Good Enough?

Example

10
 Cp

• Cp is the “ratio of tolerance to 6 sigma, or the USL minus the LSL divided by 6 sigma. It is
sometimes referred to as the engineering tolerance divided by the natural tolerance.” (ASQ)
• Cp is an appropriate measure of process capability when:
‣ The specification is two sided—that is it has both an upper and a lower bound
‣ The process is centered on the specification target value
‣ The individual measurements of process output are approximately normally distributed
‣ The process is in control

11
 Calculating Process Capability - Example
Part weight data were collected from a molding operation. 25 samples consisting of 4 parts each were collected from the process. First, sample
means and sample ranges were calculated and used to construct x-bar and R-charts. The process is in control. Next, the individual observations
were analyzed and process mean & standard deviation were calculated and used to analyze process capability. The process distribution is
approximately normal and reasonably centered on the specification target. The organization uses a standard of Cp at or above 1.33 to classify a
process as being capable. The process is capable since Cp is 1.5458.

Control charts using sample statistic Process capability analysis using individual statistics

12
 Cpk

• When the requirement “the process mean is centered on the

LSL Target USL
specification target” is not met, Cpk is used instead of Cp
‣ Cp represents the distance between process centre and
nearest specification limit Cp = 2.0
Cpk = 2.0
‣ It is equal to Cp, when process mean and specification
target are the same
Cp = 2.0
‣ It starts reducing as process mean and specification target Cpk = 1.0
move away
‣ It drops to 0 when as the process mean hits either USL or
Cp = 2.0
LSL Cpk = 0.0
• Cpk calculates 2 indices, Cpl and Cpu - the lower of the 2 is Cpk

Cp = 2.0
Cpk = -1.0

13
 Cpm (Taguchi Capability Measure)

• The main difference between the Cpm and the other

capability indices is that the bias from the target is directly
taken into consideration in the Cpm
• To obtain a good Cpm, a process needs to be right on
target, whereas to obtain a satisfactory Cpk value, a
process simply needs to stay well within specifications
• This might seem to lead to the same conclusion; however,
there is an obvious and major difference: with the Cpm, a
direct penalty is incurred for not being exactly on target
• Cpm is particularly useful where avoiding mismatches is
important - for example a shaft has to be inserted in a hole.
While both components may have excellent Cpk
individually, the resulting assembled system may be out of
spec and suboptimal if exact targets are not considered 14
 Process Capability for Variables Data - One-Sided Specification
Limit

• There are many situations where the specification is one-sided; that is target + 0/–
0.002
• In these cases, the target value is not midway between the LSL and USL. The target
value is the LSL
• In this case, process capability is measured using Cpu (Cpk = Cpu)
• The assumption of normality in the process data must still be satisfied

15
Example - Process Capability One-Sided Specification

16
 Process Capability for non-Normal Distributions

• The use of Cp, Cpk and Cpm require that the process distribution is approximately
normal
• There are many situations where this might not be the case
• It is possible to accommodate nonnormal distributions by identifying the appropriate
distribution and selecting it for use when determining process capability

17
 Process Capability for non-Normal Distributions - Example
A manufacturing company tracks the time it takes to prepare shipments from its factory. They use day as the rational subgroup and x-bar
and s-charts with variable sample sizes to determine the state of control of the process. The control charts indicate the process is in
control. The company has set a goal of 5 minutes per shipment and an upper bound of 10 minutes per shipment and wants to understand
the process’s capability to meet these specifications. Their first attempt to assess process capability used the software’s default setting of
the normal distribution to model their process. The output they obtained shows that the data do not fit the normal distribution, which
renders Cpk meaningless. The graph suggests a Weibull distribution instead. Changing the distribution to Weibull gives a Cpk of 0.9988, well
12

below the target of 1.33. They conclude that the process is not capable as currently designed, and thus initiate process improvement

18
 Process Capability for Attributes Data

• For attributes data, the process capability is defined as the average on the control
chart. For the p chart, the average fraction defective is the process capability
(assuming the process is in control)
• A weakness in capability estimates for attributes data is that they do not suggest why a
system is either capable or not
• For instance, there is no way of knowing whether the system is incapable because it is
not centered, it is too close to a specification limit, or it exhibits too much unit-to-
unit variation

19
Questions?