ERICSSON MEETS SMID Improve

1
ERICSSON MEETS SMIDImprove
April, 2011
2
Agenda

• 4th April
• Ericsson presentation
• Statistical tools in manufacturing
• DMAIC/IDDOV
• 7th April
• Define
• Measure
• 11th April
• Analyze
• 14th April
• Implement
• Control

3
DMAIC Chart

• Define
• Understand the task and its financial impact.
• Task selection matrix
• SMART review
• Stakeholder map
• Risk Management
• SWOT analysis
• Process map
• VOC and break down to CTQs
• 7MT
• Affinity diagram

• Measure
• Develop and execute an appropriate data
  collection method.
• Process map
• Data collection table
• Pareto diagram
• 7QCT
• Measurement system analysis
• Sampling technique
• SIPOC
• Gauge RR, Gauge attribute
• Capability analysis
• Benchmark
• Tagushi loss functions

• Analyze
• Find the root causes.
• Fishbone diagram
• Correlation analysis
• 7QCT
• Hypothesis testing
• Regression-analysis
• DOE
• Anova
• 7MT
• Data transformations
• Simulations

• Improve
• Generate and implement solutions.
• FMEA risk analysis
• Process map
• Poka-Yoke
• Hypothesis testing
• Loss functions
• Cost/Benefit selection
• Pugh Concept Selection

• Control
• Ensure that the results will last.
• Documentation, standardization and training
• 7QCT
• SPC
• Business case verification

Light
Comprehensive
4
FMEA

• FAILURE MODE AND EFFECTS ANALYSIS

5
FMEA
Focus on observable behaviors
6
FMEA

• Why?
• If it can go wrong - it will go wrong!
• Prevention is better than cure
• Where?
• Can be applied to any process
• Who?
• A team of people connected with the process

"A large safety factor does not necessarily
translate into a reliable product. Instead, it
often leads to an overdesigned product with
reliability problems. Failure Analysis Beats
Murphy's LawMechanical Engineering , September
1993
7
FMEA worksheet
8
FMEA severity (SEV) rating

• For each failure mode, decide on the impact on
  the product or operation when it occurs.
• Rate this impact in the column labeled SEV
  (severity).
• Establish your baseline for the analysis, i.e.
  SEV10 means death of a human or machine failure.
• A SEV rating cannot change when improvement
  actions are put in place unless the design has
  changed.

Effect of failure mode
Severity Rating for the Failure Effect.
Output loss from pre-amp
Loss of signal from 2nd RF amp
5
Loss of position, velocity time.
Potential Failure modes
9
FMEA occurrence (OCC) rating

• For each potential failure mode come up with one
  or more potential causes.
• Rate the probability of each potential cause
  occurring and place the rating in the column
  labeled OCC (occurrence).

Potential cause of failure mode
Effect of the failure mode
S
O
E
C
Potential Failure Effects
Potential Causes
V
C
Probability of Occurrence
C1 short
3
1

Receiver Output data loss, track loss
U21 function
10
Loss of position, velocity time
5
10
FMEA detection (DET) rating

• For each potential cause, identify the current
  controls which are in place to prevent or detect
  the failure mode. Rate the ability of each
  current control to prevent or detect the failure
  mode once it occurs. Place the rating in the
  column labeled DET (detection).

Current Controls for each failure mode.

• Ability to prevent or detect the failure mode.

O
D
R
C
E
P
Potential Causes
Current Controls
C
T
N
Test PR-20 HW-5
C1 short
2
2
32
None
U21 function
6
6
288
11
RPN calculation gt action plan

• Multiply the SEV, OCC and DET ratings together
  and place the value in the RPN column. The
  largest RPN numbers should get the greatest
  focus. Any SEV which has a value 10 should have
  attention regardless of the OCC and DET values.
  For those RPN numbers which warrant corrective
  action, recommended actions and the person
  responsible for implementation should be listed.

Risk Priority Number (RPN)
Actions
Potential Failure Effects
Potential Causes
Current Controls
S
O
D
R
E
C
E
P
Resp.
Recommended
V
C
T
N

Motor frame is unstable
3
6
18
Operator Knowledge experience Ticket specifies
- but coded
Darren Wooler
Operator skill/training
1
Coding chart to be issued displayed by machine
SEV OCC DET RPN (3 16 18)
Recommended actions and person responsible for
implementation.
12
Impact of corrective actions

• When making corrective actions use the available
  data to secure the best corrective action at the
  time.
• Correct SEV most probably needs a design change.
• Correct OCC may need a design and process change.
• Correct DET may need a design or process change.

13
Action results

• After corrective action has been taken, place a
  brief summary of the results in the Actions
  Taken block. A new value should be assessed for
  the severity, occurrence and detection of the
  failure mode and root cause with the recommended
  action implemented. Place these values in the
  SEV, OCC and DET columns and calculate the new
  RPN.

New SEV, OCC and DET values.

• Summary of actions completed.

Coding chart to be issued and displayed
Chart displayed operator informed
9
3
1
3
New RPN (risk priority number)
machine all viking as no feet - add to works spec
Darren Wooler
Clarify effect of with feet no feet measure
to drg spec
1
20
4
5
14
Example of a Pareto diagram FMEA analysis

Focus on the critical few high RPNs in your
Pareto diagram and do changes in the operations
to decrease SEVERITY or OCCURANCE alt. increase
opportunity of DETECTION
15
POKA-YOKE

• MISTAKE PROOFING

16
History

• Poka-yoke was invented by Shigeo Shingo in the
  1960s.
• The term "poka-yoke" comes from the Japanese
  words "poka" (inadvertent mistake) and "yoke"
  (prevent).
• The essential idea of poka-yoke is to design your
  process so that mistakes are impossible or at
  least easily detected and corrected.

17
(No Transcript)
18
Poka-yoke major categories

• A prevention device engineers the process so that
  it is impossible to make a mistake at all.
  Prevention devices remove the need to correct a
  mistake, since the user cannot make the mistake
  in the first place.
• A detection device signals the user when a
  mistake has been made, so that the user can
  quickly correct the problem. Detection devices
  typically warn the user of a problem, but they do
  not enforce the correction.

19
What Causes Defects?

• Process Variation From
• Poor procedures or standards
• Machines
• Non-conforming material
• Worn tooling
• Human Mistakes
• Except for human mistakes these conditions can be
  predicted and corrective action can be
  implemented to eliminate the cause of defects

20
Ten types of human mistakes

• Forgetfulness (Not Concentrating)
• Misunderstanding (Jump to Conclusions)
• Wrong identification (View Incorrectly...Too Far
  Away)
• Lack of experience
• Willful (ignoring rules or procedure)
• Inadvertent or sloppiness (Distraction, Fatigue)
• Slowness (Delay in Judgment)
• Lack of standardization (Written Visual)
• Surprise (unexpected machine operation, etc.)
• Intentional (sabotage)

21
Methods for using Poka-yoke

• Poka-yoke systems consist of three primary
  methods
• Contact
• Counting
• Motion-Sequence
• Each method can be used in a control system or a
  warning system.
• Each method uses a different process prevention
  approach for dealing with irregularities.

22
COST/BENEFIT ANALYSIS
23
Costs and benefitsExample screening of ASICs
subject to failure risk related to process factor
2,50
2,00
1,50
1,00
0,50
0,00
260
259
258
257
256
255
254
253
252
251
Process factor screening limit
24
Costs and benefits Finding the optimum
2,50
2,00
1,50
1,00
0,50
0,00
260
259
258
257
256
255
254
253
252
251
Process factor screening limit
25
Hypothesis testing
26
t-Test
Power and Sample Size
Paired t
2 sample t
1 - sample
Formulate Hypothesis
Formulate Hypothesis
Formulate Hypothesis
Plot, Plot
Plot, Plot
Test assumption of normality (Anderson- Darling)
HoData normal HAData not normal p-value gt 0.05
Test assumption of normality (Anderson- Darling)
f-test Continuous normal data Levenes test
non-normal data
Test equality of variances
Ho ?12 ? 22 HA ?12 ? ? 22
Paired t
2 sample t
1 - sample
27
F-test
2 Sample
1 Sample
Formulate Hypothesis
Formulate Hypothesis
Plot, Plot
HoData normal HAData not normal p-value gt 0.05
Test assumption of normality (Anderson- Darling)
f-test Continuous normal data Levenes test
non-normal data
Test equality of variances
Ho ?12 ? 22 HA ?12 ? ? 22
28
SAMPLE SIZE
29
Student Curve - t Distribution

• t curves vary with sample size -they get wider
  and flatter than normal as sample size is reduced
• In a normal curve 95 of sampling distribution is
  contained within ? 1.96 se
• In a t distribution 95 is within ? 2.131 se and
  ? 2.776 se for sample size of 16 and 5
  respectively
• As sample size approaches n 30, the t-
  distribution approaches the normal distribution

30
Hypothesis Testing - Options and Errors
T-test F-testH0 my mx sy sx Ha my ? mx
sy ? sx my lt mx sy lt sx my gt mx sy gt sx
a-risk We reject a null hypothesis that is in
fact true b-risk We fail to reject a null
hypothesis that is in fact false
31
(No Transcript)
32
Power and Sample Size

• The Power of a test is the probability that it
  will allow you to reject Ho when Ho is wrong (Ha
  is really true). (power 1 - ?)
• The following has a direct bearing on power as
  do the following
• the alpha (?) level increases, the power
  increases
• the variability of the population (?) increases,
  the power decreases
• the difference (effect size) increases, the power
  increases
• the sample size increases the power increases

33
Front Panel Manufacturer

• A front panel manufacturer wants to detect
  significant changes in front panel lengths. They
  sample thousands of them because it is cheap and
  quick to do. But this huge sample makes the test
  too sensitive the blue line shows it will sound
  the alarm if the average length differs by a
  trivial amount (0.05).This Power Curve shows they
  are wasting resources on excessive precision. A
  sample size of just 100 will detect meaningful
  differences (0.25) without "crying wolf" at every
  negligible blip.
• You also want the confidence in your results
  that's appropriate for your situation (testing
  seat belts demands a greater degree of certainty
  than testing shampoo). We measure this certainty
  with statistical power the probability your
  test will detect an effect that truly exists

34
Power and Sample size

• "We've always done it this way." That's why a PCB
  manufacturer would sample 10 units to test
  whether their strength meets the target.
• According to the Power Curve, this small sample
  size made their test incapable of detecting
  important effects. They must sample 34 PCBs to
  detect meaningful differences (0.50).

35
Acceptance sampling
36
Acceptance Sampling how to reduce inspection
costs

• Quality inspection department receives a shipment
  of 540 capacitors every week. You need to develop
  a sampling plan to make decisions regarding the
  lot without having to inspect all of the
  capacitors. 
• Because some defects are inevitable, you and your
  supplier decide on quality levels and risks that
  allow some defects while maintaining
  profitability for both of you.

37
Acceptance Sampling (2)

• You and the supplier in common agreement decide
  that the worst quality you are willing to accept
  on a regular basis is 2 defective (AQL) and the
  quality that you want to reject most of the time
  is 8 defective (RQL).

38
Acceptance Sampling (3)

• It is important to notice that
• The sampling plan that that has been suggested is
  a good starting point.
• Sometimes people involved in the sampling
  procedure want you to adjust the sample size and
  acceptance number. In cases like these we should
  try to generate multiple plans at the same time
  and compare OC Curves to find the best plan. The
  following scenarios might have to be considered
• More convenient sample size The inspectors find
  it most convenient to inspect 10 capacitors from
  each of the nine boxes in the shipment. They want
  you to change the sample size from 98 to 90.
• Smaller sample size Looking to save time, your
  supervisor suggests taking a much smaller sample.
  He wants you to reduce the sample size from 98 to
  50.
• Larger acceptance number Your supplier is
  nervous that his shipments will be unfairly
  rejected. He wants you to raise the acceptance
  number and accept at least 10 defective
  capacitors before returning an entire lot.

39
Acceptance Sampling (4)

• The black line represents the original sample
  plan with sample size of 98 (n) and acceptance
  number of 4 (c).The red line represents a
  relatively small departure from the original
  plan, showing a negligible reduction in the
  producers risk and a slight increase in the
  consumers risk. You are willing to change your
  sample size to a more convenient one to keep your
  inspectors happy.
• The green and blue lines represent more
  significant changes to the sampling plan which
  result in more risk than you are willing to
  accept.
• Show your supplier that the resulting consumer
  risk is much too high for you to consider raising
  your acceptance number to 10. Perhaps you will
  evaluate other acceptance numbers between 4 and 10

40
(No Transcript)