SPC
outcome when no proper chart stratification is applied to control charts
unable to monitor individual tool performance; control limits aren't sensitive to detect real issues
Z-Score Chart (normalized chart)
w/normalized data instead of original data; allows monitoring of different processes w/a single chart.
good SPC control strategy should have
well defined process/product characteristic to be controlled; sampling plan; control chart set up properly; detailed reaction mechanism
proper ooc commenting allows others to understand
what investigation has been carried out; what has been done to affected process / equipment product; what escalation response is required?
Cpk is a process capability index that takes into account
where the process is running relative to its target
Cpk compares the __ and __ of the car compared to the parking spot
width / location
Cp compares the
width of the car to the width of the parking spot
p-chart
proportion of defective items when sample size changes
5 number summary breaks data into
quartiles: min, 25th%, 50th% [median], 75th%, max
measures of variation
range, standard deviation, inter-quartile range, and box plot summary
after process improvement, process owner should
re-calculate control limit based on data after the improvement
process improvement
reduce common cause from system
sampling plan includes
sampling frequency; sub-level sampling plan; sampling method at every level (e.g., random or fixed)
4 elements of good SPC control strategy
sampling plan, control charts, OOC, reaction mechanism
product disposition includes
scrap, downgrade, rework, or rewrite CQDR
median is better measure of central tendency for __ data
skewed
nonrandom variation caused by
special cause: unexpected event, shock to process
control chart plots process data over time to monitor
stability of process
individual chart versus moving range chart
static versus dynamic, re time
stratification
technique to split data into subgroups to determine if any contain special cause variation
Cp >= 1
textbook def of "capable process"
OOC rules detect patterns
that would be unlikely to occur if there is no process change
UWMA (UnWeighted Moving Average) chart [rolling T chart]
tighter control limits; detect process change w/slow trending or small size of shift.
over-adjustment comes from
treating each deviation from target as if it were the result of a special cause; manual adjustments can become an additional source of variation.
typical process changes we want to detect in a control chart
trend, shift, outliers, cyclical
charts classifications
trend; control
when detecting both big and small shifts using SPC charts, you're required to
turn on 2 or more OOC rules; set up both individuals chart and EWMA/UWMA together
process change investigation looks at
type of change that occurred; when change began; material affected by change; magnitude of change;
np chart
# of defective parts when sample size is constant
c-chart
# of defects on a part
process stability = % OOC =
# of samples w/ OOC rules violations / total samples
3 sigma rule of normal distribution
68% of data within +/- 1 sd; 95% within +/- 2 sds; 99.7% within +/- 3 sds
xbar chart
A plot of sample means over time used to assess whether a process is in control.
S Chart
Control chart that tracks sample standard deviations
Taguchi Loss Function
Cost of quality ("risk") increases as a quadratic function as conformance values move away from the target.
Cpk
Cp, but takes centering into account
reaction mechanism
Out-of-Control Action Plan (OCAP): explains how to react
SPC is part of
PCS - process control systems
Cp (capability index) =
Upper spec limit - lower spec limit / (6*sigma[variation=1/2sd])
what type of control chart used when there are known systematic source of variation present in the process
Xbar-R chart?
R-chart
a control chart that tracks the "range" within a sample; it indicates that a gain or loss in uniformity has occurred in dispersion of a production process
Levey-Jennings chart
a graph used to plot and visualize the results of control samples over time; form of individuals chart and a special case of the common Shewhart X-bar chart
most common OOC rule
any single data point falls outside 3sigma limit from centerline
run-to-run
automatically adjust process recipes to compensate for known disturbances and allow process to run on target
u-chart
average # of defects per unit
purpose of SPC charts
detect changes to a process as they occur rather than detecting them at final product testing, OR, worse, at the customer
EWMA (exponentially weighted moving average) chart
better small trend detection of capability
normal chart to detect
big drift
product vs process control
product measured by spec limits; process measured by control limits.
Cp < 1
cannot fit
important to measure __ and __ together
central tendency / variation
random variation caused by
common cause: inherent to process; expected
CWDR
continue-on quality deviation report
key elements of control chart
control limits [based on statistical distribution of the data]; center line indicating process current running mean; out of control (OOC) rules
M****n chart categories
critical(customer special), area, engineering
trend up or down rule used if
data is independent
shift high or low rule used if
data is independent and median is used as center line
spec limits and control limits are
derived independently
Reaction Mechanisms typically include
description of alarm that failed; primary contact; process change investigation; investigation of root cause; troubleshooting guide; instructions on product disposition; does tool/chamber need to be brought down; documentation
well defined process/product characteristic determined by
design rule; engineering knowledge; customer request
good spc control strategy should have
detailed reaction mechanism for OOC signals; properly set up control chart; sampling plan of collecting data on characteristic for SPC; well-defined process/product character to be controlled
Fault detection
detect abnormal equipment operation through monitoring of tool trace data
problem solving
detect and remove special cause from system
PML
product monitoring list
process considered stable over time when
distribution is consistent and we have an understanding of expected data value
Cp = 1
exactly fits
OOC [control]
for process disposition
OOS [spec]
for product disposition
p[proportion]-chart
fraction non-conforming; monitor changes in avg. quality level; identify control points that call for corrective action
process capability
how well does my process compare against customer's specifications
manually adjusting every run to compensate for last deviation from target; adjusting runs that deviate more than 1S from target
leads to over-adjustment
common sources of variation in a process
lot-to-lot; within-lot; site-to-site; within-unit; unit-to-unit within batch; batch-to-batch; tool to tool
process disposition includes
maintenance, parameter tuning
6 sources of variation [6 m's]
man; machine; method; material; mother nature; measurement system
"central tendency" measure
mean, median, mode
3 parts of SPC
measurement system, detection mechanism, reaction mechanism
type of control charts
measures of central tendency; measures of variation;
IQR is the __ of the data
middle 50%
spc charts help
monitor stability of process or equipment over time; trigger OOC rules; follow reaction mechanism or Out-of Control Action Plan
point out of control limit
most common ooc rule
Attribute Charts
p-chart, np chart, c-chart, u-chart
mode is the __ value of the distribution
peak
Cp > 1
plenty of space
process control is
proactive
control limits are used for
process disposition
spec / alarm limits are used for
product disposition