Statistical Process Control (SPC) and Control Charts in Six Sigma Flashcards
Are all processes subject to variation?
Yes.
Define Statistical Process Control
A way to measure, monitor, and control processes.
SPC is a methodology that uses control charts to determine when a process is out of control.
Describe the larger “continuous improvement” philosophy undergirding SPC
Continuous improvement
Inputs > Process/System > Outputs
SPC analyzes the “Process/System” part
Inputs are: People Materials Methods Equipment Environment Measurement
Output = Product/Service
Define common cause variation
Those causes that are inherent to the process and not controllable by process operators.
Define special cause variation
AKA “assignable causes”, include unusual events that the operator can usually remove.
Do the effects of common cause vs special cause variation need to be similar?
No. They often are not.
It is the CAUSES that vary.
What are the three objectives of SPC?
- Monitor the performance of a process
- Identify special and common cause variation
- Control and improve process performance
Describe (1) “monitor the performance of a process” objective of SPC
- Determine process capability and the natural range of the process
- How the process measures up to specifications
- Charts: control, histogram, run charts, check sheets
Describe the difference between “defect” and “control.”
A defect is determined by the process violating the specification limits.
A process would be call unstable/out of control when it violates the control limits/exhibits abnormal behavior.
Describe control limits vs specification limits.
Control limits look inward and are calculated from process data itself.
Specification limits are usually determined by factors external to the process (customer requirements, industry standards, etc).
List the considerations for implementing SPC
- Team members’ buy-in
- Avoiding overanalysis
- Focus on process issues (not human ones)
- Feedback on process behavior
- Control charts
Rational Subgrouping
A small homogenous sample from the process taken in a short space or time, such that every item in the subgroup is produced under similar conditions.
Done so that just normal/random effects are within that group.
Two types of variation
Within subgroup and between subgroup
Minimize within subgroup variation, maximize between subgroup variation.
Rational subgrouping principles
- Observations must be independent
- Obersvations are from a single, stable process
- Observations taken in a time-ordered sequence
Variation of interest
Choose rational subgrouping in a manner that isolates your variation of interest. Ex:
- Shift to shift
- Day to day
- Hour to hour
- Batch to batch
Selecting variables for SPC
Use your judgement, but can generally choose variables that:
- Most difficult to hold
- Tied to customer, organization, or regulatory imperatives
- Represent a critical dimension of the product or process
- Salient or known
Selecting variables for SPC
Use your judgement, but can generally choose variables that:
- Most difficult to hold
- Tied to customer, organization, or regulatory imperatives
- Represent a critical dimension of the product or process
- Salient or known
Most difficult to hold
Directly associated with high defect rates
Known to exhibit a lot of variation
Tied to customer/organizational/regulatory imperatives
- Customer complaints
- Key customer requests
- Standards
Variables representing a critical dimension of the product/process
- Those that effect human safety, environment, or the community
- KPIVs, KPOV (key process input/output variables)
- Variables that have caused processing difficulties
- Variables that help control the process
- Variables that contribute to high internal cost
The Ideal: one variable that’s so critical, you HAVE to do it
If you CAN’T tell, use other tools to narrow down
- analysis of variance
- Pareto analysis
Salient/known variables
- Variables with special cause variation
- Variables that can be measured
- Items that can be counted by the person counting
- A leading indicator
How to use these categories in your own project?
Narrow down to those with biggest effect on:
- Cost
- Customer satisfaction
- Key process variable
Avoid overcharting
When you use more than a few charts, the benefits increase costs, not decrease.
How to apply rational subgrouping
adfsafd
Final considerations of rational subgrouping
Size
- If you have a large subgroup taken over a short period of time, it may contain dependent data
- If you take it over a longer period of time, it may contain special cause variation.
Subgroup size
A sample size of 2-3 items
- Are more economical and assumed to be rational
- Can detect large shifts, but not small shifts as easily
- You need a sufficient history of the process in order to create meaningful control chart
- Need enough subgroups
- Large subgroups may contain dependent data or special cause variation
- Service processes typically have a sample of one
SELECTING CONTROL CHARTS AND DETERMINING LIMITS
SELECTING CONTROL CHARTS AND DETERMINING LIMITS
Controls charts are used for
Analyzing variation in almost any process to
- Control ongoing processes
- Predict range of outcomes expected
- Determine whether process stable/in stat control
- Analyze process variation patterns from special causes
- Determine if goal of QI project should involve making changes to the process
-
Anatomy
Center Line CL
UCL
LCL
Individual data points = measurements/observations
6 steps to create
- Choose appropriate chart for data
- Determine CL
- Determine UCL and LCL
- Draw those lines
- Plot data points
- Analyze & interpret
How to choose the appropriate chart and how to determine CL, UCL, LCL
Important to remember dif between types
Variable data
When quality characteristic can be expressed as numerical data (time, length, degrees, money)
- Can be expressed as decimals
Attribute data
For product characteristics with a discrete response (yes/no, pass fail, good bad)
- Usually for defects
- Ex: off production line as acceptable or not
Visual difference
Control charts for variable data is displayed in pairs
- Control charts for attribute data is displayed singly
Sometimes you’ll have a choice
- Ex: you haven’t collected yet and
- Variable charts are good but also intensive and expensive
Control charts for variable data when subgroup size is greater than one
- X-bar and R chart (or simply R chart or X-R chart)
- X-bar and s chart, aka s chart or X-S chart
Control charts for variable data when subgroup size is one
- Individual/moving range charts AKA ImR charts
- ImR chart is a pair of control charts
Control charts for attributable data
- C & U charts, which follow the Poisson distribution
- Np and p charts for binomial distribution
Control charts for variables data
X-bar and R chart
Typically used when subgroups 2-10 measurements
X-bar and s chart
- For subgroups greater than 10
- Because s (standard deviation) is more accurate depiction of dispersion for large groups vs. R (range)
Individual & moving range chart
ImR chart
Used for variable data that uses individual characteristics
- Plots both individual measurements and moving range between subgroups
Used when sample data isn’t in subgroups, but rather individual observations
- AKA use when the subgroup size is one
For attributes data
- C and u charts follow poisson distribution (can count the number of defects)
- – Ex: defects by day, batch, or machine (use when sample size constant)
- U chart when sample size not constant
NP & P charts (binomial distribution)
- NP -
P
- Most effective when sample size 50 or more
Control chart data point abbreviatiosn
I - individual values X-bar - subgroup averages R - subgroup ranges p & u - proportions or averages np & c- numbers per unit
Variable control limits
SEtting them proportionally to subgroup sizes
- control limits are not straight lines
Constants are involved in the calculation of variable charts
Remember dis
Is mortgage processing time in control?
Won’t be homogenous
- Msut consider each case individually and have subgroup of 1
- So use ImR / moving range chart
-
Invoicing process: counting the number of errors that occur
U chart
