Transcript Process and Measurement System Capability Analysis

Process and Measurement
System Capability Analysis
Chapter 8
Introduction
•
•
•
•
•
•
•
•
•
Statistical Techniques
Process Capability Analysis Using a Histogram or a Probability Plot
Process Capability Analysis Ratios
Process Capability Analysis Using a Control Chart
Process Capability Analysis Using design of Experiments
Process Capability Analysis with Attribute Data
Gauge and Measurement System Capability Studies
Setting Specification Limits on discrete Components
Estimating the Natural Tolerance limits of a Process
Statistical Techniques
• Process Capability = the uniformity of a process
• Natural Tolerances= µ+3σ and µ-3σ = UNTL (Upper Natural Tolerance
Limit) and LNTL (Lower Natural Tolerance Limit)
• Normal Distribution includes 99.73% of the variable
• Or 0.27% of the process output will fall outside natural tolerance
limits
• Process Capability Analysis a study to estimate PC (Process Capability)
• Product Characterization- samples from the supplier with no idea of
the process they used or made the product with.
Process Capability Analysis Using a Histogram or a
Probability Plot
• 100 or more observations for stability
• Steps to follow before data collection- Choose machine to be used,
select process operating conditions, select operator and monitor the
collection process
• Use x bar and standard deviation for calculations od capability
• Substitute x bar for µ and s for σ in the formula= x̄ + 3s
Histogram Example
Cp and Cpk > 1 but < 1.33
Taken from: http://www.qimacros.com/quality-tools/histogram-examples/
Probability Plotting
• Alternative to Histogram which can determine the shape of the
distribution
• Advantage is divide the range into intervals which often produces
reasonable results for small samples
• Ranked data versus cumulative frequency on special paper with
vertical scale with distribution as a straight line
Probability Plot Example
Data not normally distributed so 3σ will not work
Taken from: http://www.statit.com/support/quality_practice_tips/normal_prob_plot.shtml Authors Robert and
Marilyn Hart
Capability Ratios
• Cp is the USL (Upper Specified Limit) minus the LSL (Lower Specified
Limit) divided by 6σ and measures potential capability
• Cp = USL-LSL/ 6σ
• Sample Standard deviation s or Rbar/D2 which is C-hatp=USL-LSL/6σhat
• Off center process- Cp measure the spread of the specification
relative to six sigma not of the account of where the mean is locate
• Cpk = min{Cpu, Cpl}, Cpu= USL-µ/3σ, Cpl = µ-LSL/3σ
• Cpk measures actual capability and is based on normal distribution of
the process output
PCR Continued
• Cpc = USL-LSL/6*sprt((Pi/2)*E|X-T|) T is the target =1/2(USL+LSL)
• Cpm =USL-LSL/6t t-sqrd = E[(x-T)sqrd] then Cpm =USL-LSL/6*sqrt
sigma sqrd+(mu-T) sqrd which also = Cp/sqrt 1+ξ sqrd
• Confidence Intervals and Tests on PCRs
• C-hat p or C-hat pk- Cp replace σ with s to find interval in Cpk = [1Zalpha/2*sqrt1/9nc-hatsqrdpk=1/2(n-1))≤ Cpk if ≤C-hat pk change –
to +
• Cpc is USL-LSL/s*sqrt (pi0/2)*c
PCR continued
• Testing Hypothesis about PCRs
• Ho: Cp=Cpo (or the process is not capable)
• H1: Cp≥Cpo (or the process is capable)
• Formulate the statistical test in the terms of C-hat p, so the we will
reject Ho if C-hat p exceeds the critical value c.
PC Analysis Using a Control Chart
• They do not necessarily display potential capability and do not
address statistical control
• Regard them as a primary technique of PC analysis.
• 2 charts can be used Attribute and Variable
• Xbar and R should be used whenever possible because stronger and
better information comes from them.
• They allow both instantaneous and variability across time to be
analyzed
PC Control Chart continued
• Xbar Chart – Centerline= Xbarbar, UCL= Xbarbar+ Asub2Rbar, LCL=
Xbarbar- Asub2Rbar
• R Chart- Centerline Rbar, UCL= Dsub4*Rbar, LCL=Dsub3Rbar
• Mu-hat = Xbarbar sigma-hat= Rbar/Dsub2
Xbar and R Chart Example
Taken From: http://www.minitabknowledge.com/2012/12/x-bar-and-r-chart.html#.Uxy-XGbD99A
author Minitab.com
PC Analysis Using Designed Experiments
• Is a systematic approach to varying the input control variables
• Used in isolating and estimating the sources of variability in a process
• Experiments can be designed to sample a various number of
machines with several applicators to obtain an estimation of the
variances
PC Analysis with Attribute Data
• Process Performance is measured by this data
• Nonconforming units or defectives, or nonconformities or defects
• DPU (Defects Per Unit) = Total number of Defects/Total number of
units
• DPMO (Defects per million opportunities) = Total number of defects/
Number of units x Number of opportunities
Gauge and Measurement System Capability
Studies
• Basic Concepts 1. Determine how much of the variability is the gauge
or instrument, 2. isolate those components of variability in the
measuring system, 3. Assess the capability of the instrument or gauge
• Second step in DMAIC problem solving process
• Utilize the 2 R’s of the measurement system capability (MSC),
Repeatability and reproducibility.
• Another important aspect of MSC is Linearity- reflects the differences
in observed accuracy and /or precision experienced over the range of
measurements made by the system.
G&M continued
• Stability and Bias 2 of the problems with linearity due to the
calibration and maintenance issues.
• Stability is the different levels of variability in different operating
regimes
• Bias reflects the difference between observed measurements and a
true value from master or gold standard. Or from a different
measurement technique known to produce accurate values
• Uses Xbar and R charts
G&M continued
• Precision to Tolerance ratio= P/T= k sigma-hat sub gauge/USL-LSL
• Compares the estimate of gauge capability to the width of the
specifications or the tolerance band (USL-LSL) for t he part that is
being measured.
• Discrimination Ratio for the gauge to capable the DR must exceed 4
• Gauge precision and accuracy- precision is the variability in the
measurement and the accuracy is the ability of the instrument to
measure the true value.
The Analysis of Variance Mode
• Measures the R&R (repeatability and Reproducibility) of and
experiment.
• Also known as ANOVA the standard model for a gauge R&R
experiment
• Can be used to estimate variance components by partitioning the
variability in the measurements into component parts
• SS-total= SS-parts+SS-operators+SS-p*o+SS-error
ANOVA continued
• MS-p= SS-parts/p-1, MS-o=SS-operators/o-1, MS-po=SS-p*o/(p-1)(o1), MS-E+SS-error/po(n-1)
• Expected value are; E(MS-p)= σsqrd + nσsqrd + bnσsqrd, E(MS-o)=
σsqrd + nσsqrd + anσsqrd, E(MS-po) = σsqrd + nσsqrd-po, E(MS-E) =
σsqrd
• σ-hatsqrd= MS-E, σ-hatsqrd-po= MSpo-MSE/n, σ-hatsqrdo =MsoMspo/pn, σ-hatsqrdp= MSp-Mspo/on
Setting Specification Limits on discrete
Components
• Linear Combination is the variables of the dimension of a component
where asubc is constants. The all follow a straight line.
• Example: If the xi are normally and independently distributed with a
mean or mu and variance sigma sqrd, then y is normally distributed
with mean or mu = to summation of and mu and variance = to
summation a sqrd and sigma sqrd
• Nonlinear combination is a combination of means mu1, mu2,…mun,
are the nominal dimensions associated with the components x1, x2,
…xn and then by expanding the right-hand side of the equation about
the mu’s.
Estimating the Natural Tolerance Limits of a
Process
• Uses concepts such as 100(1-alpha), mu+z sub alpha/2*sigma.
• Tolerance limits based on the Normal Distribution= xbar+z sub
alpha/2*s
• Confidence limits are used to provide and interval estimate of the
parameter of a distribution
• Tolerance limits are used to indicate limits between which we can
expect to find a specified proportion of a population.
• Nonparametric Tolerance limits- interval that are based on the
distribution of the extreme values in a sample from an arbitrary
continuous distribution.
End
Questions?