Scribd
Upload
7 views

Process Structure and SPC

This document provides an overview of advanced statistical process control techniques including inferential statistics, normality testing, rational subgrouping, capability analysis, control charts, sampling plans, and additional tools such as FMEA. It discusses key concepts such as descriptive versus inferential statistics, normal versus non-normal distributions, variables versus attributes control charts, and sampling to detect process shifts. Examples of various control chart types and calculations are also presented.

Uploaded by

ekdanai.a
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
7 views

Process Structure and SPC

This document provides an overview of advanced statistical process control techniques including inferential statistics, normality testing, rational subgrouping, capability analysis, control charts, sampling plans, and additional tools such as FMEA. It discusses key concepts such as descriptive versus inferential statistics, normal versus non-normal distributions, variables versus attributes control charts, and sampling to detect process shifts. Examples of various control chart types and calculations are also presented.

Uploaded by

ekdanai.a
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 50

Process Structure

and SPC
Advanced applications overview

Mike Miller #7078


Outline
 Inferential Statistics
 Normality testing
 Rational Subgrouping
 Capability Index
 Control Charts
 HTI vs. Shewhart (HTI sampling not random)
 Attribute and varaible
 Sampling
 Detection of shifts
 Reading issues from variation charts
 FMEA
 Additional Tools
Inferential Statistics

 Descriptive statistics is use to describe


the data: (i.e. average, std dev)
 Inferential statistics are used to draw
conclusions: (i.e. testing)
 Components of inferential statistics that
help draw conclusions:
 P-value
 Confidence Intervals
Gage R&R (Limiter Gap)
Right Gap Left Gap
Full Gage Example
Full Gage Example
Full Gage Example
Normality testing

 In order for process control to be valid,


we must be able to show a normal
distribution.
 This can be done using the following
tests:
 Anderson-Darling
 Ryan-Joiner
 Kolmogorov-Smirnov
Normal Model

 Will measure a goodness of fit to a


normal model.
 The tests will determine if the data has
skewness or kurtosis.
Normality testing

 When we test for normality we use


sample data.
 This is because samples are plotted on
the chart not individuals.
 If the chart is an individuals MR chart,
then all data is needed for normality
check.
Normality testing
 To select test:
 Basic Statistics
 Normality test
 Select data column
Normality testing
 To pass the test the P-value must be greater
than 0.05

 If the test does not pass, may lead to


identification of a smaller rational subgroup.
Normal Distribution

 Distribution will take shape of a bell.


Normal Distribution
• Represented by a
bell shaped curve.

• Approximately 2/3
of all checks will be
within 1 Std of
nominal.
• Approximately 1 out
of every 20 points
will land outside of 2
Stds from nominal.

Knowing this information allows us to predict outcomes. If data does not


stay within our prediction, something has changed or something is wrong.
Non-normal distribution

 Sometimes the data may not be normal


in will need to be transformed to a
specific distribution.
 Examples:
 Weibull (Most common)
 Log normal
 Exponential
 Gamma
Non-Normal Distribution
 Our True Position measurements often take on
a weibull distribution.
 Combination of X and Y Axis
 (i.e. Dimple and CuFid)
Rational Subgroups
 In order to define a normal distribution, we
sometimes have to break down to the smallest
subgroup that can establish a normal
distribution.
 Site
 Part number
 Unit
 Tool set
 Additional information can be found in WD
Spec. 92-004107 located in PDM.
Capability
 Next step is to identify your process
capability, this method identifies the
defective parts per million that can be
found in our process.
 Capability can help us identify the level of
continuous improvement needed.
 Higher the CpK the better.
 1.33 is WD minimum requirement for CTF-2
Components of Capability

Dimple symmetry - How the components affect the Capability

In example 1 we have In example 2 we have a In example 3 we are


a good Std and are good Std but we are centered at nominal
centered at nominal shifted from nominal but have a poor Std
Process Control

 If we have good capability HTI is able to


use non-standard limit calculations.
 Due to our sampling plans not always
being random we do not use the
standard shewhart format.
 We take samples off the same strip, using
shewhart control it will cause us to over-
adjust our process.
Standard Shewhart
Process Control

 Due to HTI sampling structure (All


samples on same strip) we use I-MR-R/S
control.
Process Control vs. Shewhart
Sampling to normal
 Same parts different sampling, single strip has
higher variation that has to be captured with
out control chart calculations.
Attribute vs. Variable
 Variable charts are most common at HTI –
uses measurable values.
 X-bar (R or S) charts
 I-MR
 Hybrid of I-MR-R/S
 Attribute charts are “yes/no” charts – condition
either is met or not met.
 P and Np charts
 U and C charts
Variable Chart Calculations
For an X(bar) and R chart For an X(bar) and S chart

Average Average

Range Range

Where X(dbl bar) and R(bar) are calculated

K = number of subgroups

Note: A, B, and D values are constants found on the table provided


Table
X(bar) – S : Gram CIP
Dispersion Charts
 Dispersion is actually the most critical to
view.
 Issues identified are often most complex
 Average charts usually only need an
adjustment.
 Variation within the sample indicates:
 Measurement system issues
 Procedural compliance issues
 Extensive tool or component issues
I – MR : Weld Pulls
Attribute charts

 There are 4 types of attribute charts:


 P (Percent or fraction of nonconforming defects)
 np (Number of rejects in a constant sample)
 c (Defects where n is constant)
 u (Defects per unit, n is not constant)
Formulas for Attribute charts
P – Chart : VM Defects
nP – Chart - Lot Inspection

 Attribute chart
C – Chart : Panel
Inspection
 Counts number of non conformities
 Uses Attribute Data
 Uses consistent sample size
U – Chart : Panel Inspection
 Counts number of non conformities
 Uses Attribute Data
 Can be used with uneven sample size
OoCAP: Out of Control
Action Plan
 Procedure used if
our process goes
outside the
established limits.
 Includes average,
dispersion and
specification non-
compliance.
Sampling plans
 Because we know that 100% sampling
may not always be the most effective.
 We set up sampling plans with
acceptance controls.
 These controls or criteria can be
established from standards that were pre
determined.
 Current audits have us sample 125 parts
in each of our 4800 part lots.
Sampling plans
Sampling plans
Sampling plans
 AQL 0.65 is the
percent defective
expected with a
95% probability.
 AQL = Acceptable
quality level
 LTPD = Lot
Tolerance percent
defective

 If our defect rate is at 1%, we would have a 92% chance of


accepting the lot.
 If our defect rate is at 3%, we would have a 25% chance of
accepting the lot.
Assembly Inspection
Sample plans (Variable)
 With variable data we can also set specific
sample sizes to detect process shifting.
Minitab Power Curve:
 Minimum required
sample size to
detect a mean
shift of 0.0875
degree.
 Using power level
0.9,
 Process sigma of
0.05168
 Proposed sample
size is 8 which is
greater than the
calculated
minimum of 6
parts
FMEA

 Failure – Modes – Effects - Analysis


Cause and Effect
Force Field Analysis
Situation Analysis
Problem Analysis
8D
D1 – Form a Team
D2 – Define the problem
D3 – Contain the problem
D4 – Determine the root cause
D5 – Verify solutions
D6 – Validate solutions
D7 – Prevent from ever happening again
D8 – Congratulate the team

You might also like