DIE CAST PROBLEM SOLVING

NADCA PUBLICATION #413

NORTH AMERICAN DIE CASTING ASSOCIATION SCOTT KIRKMAN & STEPHEN UDVARDY
 Although great care has been taken to provide accurate and current information, neither the author(s)
nor the publisher, nor anyone else associated with this publication, shall be liable for any loss, damage
or liability directly or indirectly caused or alleged to be caused by this book. The material contained
herein is not intended to provide specific advice or recommendations for any specific situation.
Any opinions expressed by the author(s) are not necessarily those of NADCA.

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used
only for identification and explanation without intent to infringe nor endorse the product or corporation.

© 2016 by North American Die Casting Association, Arlington Heights, Illinois. All Rights Reserved.

Neither this book nor any parts may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, microfilming, and recording, or by any information
storage and retrieval system, without permission in writing from the publisher.
TABLE OF CONTENTS

I. Introduction 1
II. Variation and Basic Statistics 3
III. Problem Solving Methodologies 7
IV. 8D Step 1: Team Approach 11
V. 8D Step 2: Describe the Problem 15
VI. 8D Step 3: Implement and Verify Interim Containment 21
VII. 8D Step 4: Determine and Verify Root Causes 23
A. Identify Potential Causes 23
B. Select Likely Causes 31
C. Is the Potental Cause a Root Cause? 34
D. Identify Alternative Solutions 47
VIII. 8D Step 5: Verify Corrective Actions 49
IX. 8D Step 6: Implement Permanent Corrective Actions 55
X. 8D Step 7: Prevent Recurrence 57
XI. 8D Step 8: Congratulate Your Team 59
XII. Problem Solving Examples 61
A. Generic Example 61
B. Die Casting Example 1: 2 Factor, 2 Level Full Factorial 63
C. Die Casting Example 2: 5 Factor, 2 Level Fractional Factorial 64
D. Die Casting Example 3: 5 Factor, 2 Level Fractional Factorial 68
E. Die Casting Example 4: 3 Factor, 3 Level Full Factorial 77
XIV. Appendices 83
Chapter I
INTRODUCTION

The State of the Die Casting Industry

The level of technology today in many die casting facilities is largely dependent upon the
process knowledge that exists among the operators, managers, and process engineers
within the facility. This knowledge is typically experiential in nature, and therefore, a common
response to product quality or defect problems has been to make significant process and
die changes such as changes in gating, plunger diameters, die designs, etc. Although these
changes may sometimes be effective, these changes frequently do not resolve the product
quality or defect problems. The problem is a lack of industry skills and standards for analyz-
ing process variation and optimizing critical process variables to minimize or eliminate prod-
uct quality or defect problems.
Furthermore, when one tackles a problem by manipulating process factors to find the
best process for a given defect problem, it is rare that the best process is found. This
is because one typically changes only one process factor at a time until the process
believed to be the best process is found. Since there are nearly an infinite number of
potential process combinations, finding the best process can never happen using this
approach.

The Solution
The lack of success in die cast problem solving is not likely due to a lack of interest or
effort in solving problems. Rather, it is likely due to a lack in understanding the nature
of variation, and the fact that the interactive nature of die casting process parameters
makes solving process problems very difficult. Thus the solution lies in understand the
variation and how to deal with the interaction of process parameters (variables).

This Text
This die cast problem solving text and the accompanying EC-413 course is unique and
progressive in that it addresses variation thoroughly and covers the interactive nature of the
die casting process by using problem solving tools that respect this interactivity. In addition,
this book approaches problem solving using the Eight Discipline Problem Solving Process,
which is required under QS-9000 certification and accepted under TS-16949 specification.

Die Cast Problem Solving Page 1

 Introduction

Page 2 Die Cast Problem Solving

Chapter II
VARIATION AND BASIC STATISTICS

Statistical Thinking
“Statistical thinking is a philosophy of learning based upon the following principles:
• All work occurs in a system of interconnected processes,
• Variation exists in all processes, and
• Understanding and reducing variation are keys to success.”
(Special Publication, ASQC Statistics Division; Spring 1996).

Analyzing this statement in detail provides justification for the problem solving strate-
gies proposed in this course…
1) “Learning” – Learning can be equated to an observer being exposed to an infor-
mative event. Therefore, the probability of learning can be increased by increasing
the probability of observing an event and increasing the probability of an informa-
tive event occurring. Statistical methods enable us to learn more about the factors
that control the quality of our process. The process knowledge gained using sound
problem solving strategies may directly lead to technological improvements that
provide significant quality and cost benefits.
2) “Work occurs in a system of interconnected processes” – In die casting, “in-
teractive processes” may be a more appropriate term to use than “interconnected
processes.” With that adjustment considered, the statement leads one to realize
that the process and design factors that are required to make quality product must
work in symphony together. In die casting, there is a significant interactive effect of
all processes and design characteristics that come together at one instant to en-
able proper filling of the die within 0.005 to 0.200 seconds.
3) “Variation exists in all processes” – Variation occurs naturally in all things to some
degree. The flattest, most perfect cut on the face of a diamond would appear very
rough under our most highly powered microscopes, but this does not mean that the
diamond is not of high value. It only illustrates that variation will always occur, and
understanding the impact of the variation on product quality is what is important.
4) “Understanding and reducing variation are keys to success” – We must first
understand what types of variation are important. Then we can find ways to reduce
these significant process variations, which will likely lead to a decreased production
of sub-standard product and thus an increased production of high-quality product.

Die Cast Problem Solving Page 3

 Variation and Basic Statistics

Average, Variance, Standard Deviation and Sigma

Process variation is understood using the mathematical concept of variance. Variance
is a squared number. Therefore, the variance that occurs in the thickness of a biscuit in
cold chamber die casting is presented in inches squared (in2). Variance of a process is
defined by the formula:
n
V = Σi [(X-bar – Xi)2] / (n-1)

Where: Σ  is a symbol indicating the sum of the statement in the brackets [

] from first sample i to the last sample n.
X-bar is the average of all data points in a sample of the process.
Xi is the individual value of each data point in the sample.
n is the number of data points in the sample.
For example, if we sample 5 biscuits and find they measure 1.5 in (38.1 mm), 1.3 in
(33.0 mm), 1.6 in (40.6 mm), 1.4 in (35.6 mm), and 1.7 in (43.2 mm).
We calculate X-bar as (1.5+1.3+1.6+1.4+1.7) / 5 = 7.5 / 5 = 1.5 in (38.1 mm).
Therefore, V = [(1.5-1.5)2 +(1.5-1.3)2 +(1.5-1.6)2 +(1.5-1.4)2 +(1.5-1.7)2] / (5-1)
V = [0+0.04+0.01+0.01+0.04] / (4) = 0.10 / 4 = 0.025 in2 (16.13 mm2)
To put this into user friendly terms that we can easily understand, we must take the
square root of variance (V) to have a value in units that make sense to use. The square
root of variance is called standard deviation (sd) when considering a sample of data from
a larger population, and standard deviation is called sigma (σ) when considering all the
data from a population.
For example: If our variance (V) is 0.025 in2 (16.13 mm2) in our biscuit example,
then the standard deviation (sd) is equal to the square root of 0.025 in2
(16.13 mm2) or 0.158 in (4.02 mm)
It should be noted that a standard deviation close to 0 indicates that the data points
tend to be very close to the average (also called the mean) of the set of data points,
while a high standard deviation indicates that the data points are spread out over a
wider range of values.

Page 4 Die Cast Problem Solving

 Variation and Basic Statistics

Normal Distribution
The normal distribution is the probability density distribution, which defines most all
types of variation in any process. A normal distribution appears as shown next:
Within the normal distribution curve, 68 percent of the variation falls within –1 to +1
standard deviation from the average; 95 percent of the variation falls within –2 to +2
standard deviations from the average; 99.7 percent of the variation falls within –3 to +3
standard deviations from the average; and 99.994 percent of the variation falls within
–4 to +4 standard deviations from the average. We have accounted for all but 0.3 per-
cent of the process defined by 3 standard deviations, and we typically use 3 standard
deviations to define the normal variation of the process. In the biscuit example above, 3
standard deviations is 3 x 0.154 in. or 0.474 in.

Relationship of Process Variation to Defect Variation

As described earlier, standard deviation is the square root of variance. If we are analyz-
ing the impact of process variation on the variation of a product defect, then we must
consider the squared relationship as well. That is, if we have several process parame-
ters whose variation contribute to the variation of a product defect, we must understand
that reducing variation of one process parameter may not cause a significant reduction
in the defect’s total variation. The standard deviation of the defect variation is equal to
the square root of the sum of the squares of contributing variation from each of the pro-
cess parameters. Therefore, we must consider the variation of all process parameters
when utilizing an effective problem solving process.
For Example:
If: Sd = (Sc12 + Sc22 + Sc32+….)0.5
And three parameters (Fast Shot, Metal Temperature, and Intensification Pressure) con-
tribute to a porosity defect variation as follows:
SFS = 2
SMT = 1
SIP = 2
If SFS is taken as Sc1, SMT is taken as Sc2, and SIP is taken as Sc3, the total porosity defect
variation is Sd = (22 + 12 + 22)0.5 = (9)0.5 = 3.
Assume that through research and development we find that we can put a shot con-
trol valve on the shot end and reduce the Fast Shot standard deviation from 2 to 0.5.
What will be the new porosity defect variation?

Die Cast Problem Solving Page 5

 Variation and Basic Statistics

The answer is Sd = (0.52+12+22)0.5 = (5.25)0.5 = 2.29.

As you can see, the reduction in Fast Shot variation was extremely significant.
However, the corresponding reduction in defect variation was not as significant.
Three important die cast problem solving points can be understood from the above
example:
1) Several factors frequently affect a defect because of the interactive nature of the
die casting process.
2) When several factors contribute to a defect’s variation, significantly improving one
factor may not seem to have a significant effect on the defect.
3) Reducing
 the variation of several parameters is often the only way to assure
significant improvement.
Not taking these points into consideration tend to result in unsuccessful problem solv-
ing efforts in die casting. Becoming aware of these points will make you a better prob-
lem solver.

Page 6 Die Cast Problem Solving

Chapter III
PROBLEM SOLVING METHODOLOGIES

There are several approaches to problem solving that have been offered to the manu-
facturing community. All of these approaches have some commonalities, which are:
1. Describe the problem.
2. Determine and test potential root causes of the problem.
3. Determine potential corrective actions.
4. Implement corrective action(s).
A good problem solving method should include the above steps.

Eight Discipline Problem Solving Process

The Eight Discipline (8D) Problem Solving methodology was developed by the Ford Mo-
tor Corporation. This methodology includes four steps that are not always used in other
methodologies. These steps are:
1. Use a team approach. (8D step 1) Ford thought that a team approach to problem
solving was superior to individual efforts.
2. Implement and verify containment. (8D step 3) This step is to assure that defective
product does not get to the customer during the problem solving process. Therefore,
once a company becomes aware of a defect, they should find a way assure that it is
contained within the manufacturing facility where it is manufactured.
3. Prevent recurrence. (8D step 7) This step is to assure that a problem does not
come back after time.
4. Congratulate the team. (8D step 8) This step is to promote team oriented problem
solving by recognizing those who are involved.
Since the 8D format has been adopted as the preferred problem solving methodology
under the QS-9000 certification, this text will follow the 8D steps.

Die Cast Problem Solving Page 7

 Problem Solving Methodologies

Team Oriented Problem Solving

While Ford created the 8D process, they also developed the Team Oriented Problem
Solving (TOPS) process to work along with 8D. TOPS is a method designed to optimize
the effectiveness of teams. Some of the TOPS techniques will be described in this text.

Eight Phase Problem Solving

The Eight Phase (8 Phase) problem solving process is very similar to 8D except that
teaming is not necessarily utilized, and other steps are not addressed in as much detail.
Since 8D is the better-accepted approach and since the approaches are so similar, 8
Phase will not be referred to within this text.

Dorian Shainin
Dorian Shainin is a problem-solving “Guru” who developed his statistical engineering
methodology while working at Hughes Aircraft. His methods are still being used signifi-
cantly within General Motors and some of his methods will be applied in this text.

Taguchi
Taguchi, a Japanese problem-solving “Guru”, developed a non-interactive design of
experiments methodology. His methods have been successfully used in non-interactive
manufacturing processes for many years. However, his methodologies are of little use in
the die casting industry due to the interactive nature of die casting.

Box, Hunter, and Hunter

Box, Hunter, and Hunter, statistics professors at the University of Wisconsin – Madison,
developed fractional factorial experiment designs that may be formatted to respect in-
teractions within the die casting process. These types of experiments are utilized in this
text in addition to other methods from Box, Hunter, and Hunter.

This Book
The uniqueness of this book lies in that it is specific to the interactive nature of die casting.
We consider a variety of existing problem solving tools such as 8D, Shainin, Taguchi, and
Box, Hunter, and Hunter. The specific problem solving tools selected are those that work
well in reducing interactive die casting problems. The new problem solving format, out-
lined below, is linked to the 8D steps for ease of application and understanding.

Page 8 Die Cast Problem Solving

Problem Solving Methodologies

Die Cast Problem Solving Page 9

 Problem Solving Methodologies

Page 10 Die Cast Problem Solving

Chapter IV
8D STEP 1: TEAM APPROACH

In general, teams have not been well utilized in the problem solving process in die cast-
ing, and therefore, are an important topic of discussion in this text. Teams are an effec-
tive and justifiable means of solving problems, and if employee involvement is a goal,
team oriented problem solving is the answer.

Benefit of Knowledge
The most significant benefit of utilizing teams in the problem solving process is the
cumulative knowledge that teams are capable of providing. By having a well-rounded
team composed of people with the required product, process, and customer knowledge,
problem solving is most likely to be effective.

Benefit of Employee Buy-in

Beyond the obvious benefit of solving a product quality or defect problem, teams may
have a significant benefit with regard to employee buy-in toward manufacturing improve-
ment. Since operators and other floor employees do not always directly see the benefits
of their efforts, they often become disinterested in involvement in improvement activities.
By effectively involving die cast floor employees in teams, employees become more
aware of the benefits of their efforts and more satisfied with their employment.

Taking Advantage of the Experts

The most important aspect of teamwork is to involve the people who understand pro-
cess variation the most. Your first thought may be to involve the process engineers
heavily in problem solving, however, it is the machine operators who likely understand
process variation the most. The machine operators watch the process far more than
anyone else does. They see how it changes over time. Furthermore, they are the ex-
perts when it comes to ideas regarding what may cause the process to vary over time.
Not utilizing the expertise of machine operators in problem solving efforts is the biggest
roadblock to team effectiveness in the die casting industry. Learning to involve operators
in problem solving teams is a key to team success.

Die Cast Problem Solving Page 11

 8D Step 1: Team Approach

Team Development Guidelines

The following guidelines should be used in determining team membership:
1) Each team member must directly gain from the success or lose from the failure
of the problem solving project. People who do not gain or lose from their efforts
should not be involved.
2) Each team must consist of at least four members and no more than eight mem-
bers. If the team has too few members, you may not get buy-in for a solution or be
tapping into enough expertise and perspectives. If there are too many members,
you may not maintain involvement or it may become too cumbersome for the team
to move quickly.
3) Each team member must have time, which is understood by his or her manager,
that is allocated to work on tasks related to the team. If managers and supervisors
do not allocate this time or are not aware of the need to spend time on the project,
tasks will likely not be done.
4) Each team should have at least one operator, one process or product engineer,
one quality engineer or inspector, and one manager. The manager should act as
the team facilitator and break down barriers to problem resolution. If more mem-
bers are to be included in a team, add operators, possibly trimmers, or others that
have something to gain or lose from the completion of the project.

Brainstorming
Brainstorming is a team problem solving tool that is critical to success in problem solv-
ing. The goal of brainstorming is to attain the most creative and effective ideas toward
the resolution of a problem. To get these types of ideas, brainstorming should be con-
ducted in the following manner:
1) Discussions should be enthusiastic and people should be commended openly
for all ideas. Ideas that appear silly or outrageous may not seem valuable at first,
however, these types of ideas may lead to creative solutions. The facilitator should
promote these approaches.
2) Allow only one person to speak at a time. The facilitator should control who speaks
so that all have the opportunity to express their thoughts.
3) Never criticize, evaluate, or compare one idea to another. Just write them all down.
The facilitator should write down ideas and assure that the group follows this rule.
4) The facilitator should make sure the discussion is limited to the problem described.

Page 12 Die Cast Problem Solving

 8D Step 1: Team Approach

Storyboard Brainstorming
Storyboard Brainstorming is a team tool to be used for brainstorming when:
• The sequence of events in the process is not necessarily clear.
• The point at which quality issues/defects occur is not clear.
• There are members within the group who are not likely to become involved or there
are members who will tend to control the problem solving process.
To use Storyboard Brainstorming follow these steps:
1) For each process step or operation, have each team member write down potential
causes of the defect that could originate in each step. Use Post-it notes.
2) P
 ut these potential causes of the defect on a board or wall in sequence of the process.
3) Follow brainstorming rules otherwise.
The benefits of storyboard brainstorming are as follows:
• It enables structured brainstorming for defect resolution of multiple operation processes.
• It enables the group to learn about the entire process.
• It minimizes the impact of dominant group members and promotes the involvement
of typically uninvolved employees.

Die Cast Problem Solving Page 13

 8D Step 1: Team Approach

Page 14 Die Cast Problem Solving

Chapter V
8D STEP 2: DESCRIBE THE PROBLEM

Definition Statement
The definition statement is critical to beginning an effective problem solving project. Once
a problem solving team is created, coming up with a commonly understood definition
statement is the first task for the team. Once developed, the definition statement should
assure that the team directly attacks the cause of the problem defined in the definition
statement. Philosophically speaking, the statement, “A problem well described is a prob-
lem half solved,” describes the importance and benefit of a good definition statement.
The definition statement should have three separate components. It should begin
with the part number or description. The statement should then clearly define the quality
issue, defect or symptom. Once the problem solving project is complete, the statement
will conclude with the third part, which is the proven cause of the defect or symptom.
This sounds very simple, however, it is critical to proper team-based problem solving.
An example of a good prpoblem statement is as follows: Part #1212 and #1213 have
flash in the key-minder hole caused by the key core flexing away from the key-minder
core during the shot.

Pictograph or Defect Pictures

The pictograph is a picture that is used to visually describe the location of a quality issue
or defect. The pictograph should be used with the definition statement to document and
define a quality issue or defect. From here on, the term defect will be used to describe
a discrepancy or discontinuity that impacts quality. This is not to imply that the casting is
defective. Whether the casting is acceptable or unacceptable depends on the customer
requirements and severity of the defects. The problem solving process may be used to
enhance quality level as well as to solve a real problem or non-compliance issue.
The best means of visually illustrating a defect is having the actual products with
defects present at the problem solving meeting. However, the pictograph is required for
long-term defect and problem solving documentation.
Digital photos are a very good way to describe a defect if an effectively clear digital
camera is available. Often the defects are not easy to see on photos.

Die Cast Problem Solving Page 15

 8D Step 2: Describe the Problem

Defect Ranking System

The problem description and pictograph help define the defect problem for resolution.
A ranking system must be established to measure the defect problem. The best ranking
system is a gage of some sort. However, almost all die casting defect problems are sub-
jective in nature and cannot be measured with a standard gage of any sort. Examples of
subjective die casting defects include non-fill, surface appearance, porosity, blisters, warp-
ing, stains, and flash. Therefore, we have to create our own gage that we call the “defect
ranking system.” The following steps must be followed to create a defect ranking system:
Step 1: Determine the number of rank levels.
 se from 2 to 10 defect rank levels. Use more levels for more costly problems.
U
Do not pick so many levels that determining rank level is too difficult. Knowl-
edge of determining the number of rank levels improves with experience.
Step 2: Select a balanced sample of good parts and defective parts that ranges
from the very best defect level to the very worst.
 balanced sample does not infer a normal sample because a normal sample
A
of product may have only a small percentage of defective products. The num-
ber of castings in the balanced sample should be from 5 to 20 times the num-
ber of rank levels.
Step 3: Place all the parts in a row from bad to good.
Step 4: Divide the row of parts into several fairly equal groups.
 he number of groups is equal to the number of rank levels chosen in Step 1.
T
Make sure that the good/bad boundary is between two of your groups. Your
groups will be numbered 0, 1, 2, 3…etc.
Step 5: Select the boundary samples by picking the part that best falls between
the individual groups.
One boundary sample will be the good/bad boundary. This boundary sample
should be the part that is most difficult to determine its acceptability to the cus-
tomer. Your boundary samples will be numbered 0.5, 1.5, 2.5…etc. The number
of boundary samples will be one less than the number of groups. These bound-
ary samples are now your gage. They should be protected and preserved.
Step 6: Complete an Iso-Plot or a Gage Repeatability and Reproducibility (Gage
R&R) study to test the validity of your gage.
An Iso-Plot can be used for most situation, however, you may want to use a
Gage R & R for a more deterministic gage evaluation method. Note that you
may be required to use a Gage R & R if you use your gage as a product valida-
tion tool under QS-9000 or ISO-9000 requirements.

Page 16 Die Cast Problem Solving

 8D Step 2: Describe the Problem

Iso-Plot
An Iso-plot is a tool to compare: 1) two different types of gages, 2) two identical gages,
3) two people using a gage, or 4) any type of comparison of gages. The gage may be a
visual assessment based on the rank level. The Iso-plot is a Dorian Shainin tool that is
shown in the example below:

The following steps are used to create an Iso-plot:

1) Select a sample of parts for usage in your study. (20 to 50 parts)
2) Measure each part using two gages, people, or time frames.
3) Chart the two gage methods with one method on the Y-axis and the other on the
X-axis.
4) Plot one point for each part in your sample.
5) Sketch the shape that represents the extremity of the variation. The shape will be
oblong like a football at a 45 degree angle. (See the following plot.)
The Iso-Plot is evaluated by measuring the following:
• Process variation is determined by measuring the length of the plotted football shape.
• Gage variation is determined by measuring the width of the plotted football shape.
• Repeatability/reproducibility percentage is equal to the gage variation divided by the
process variation.
• Accuracy is determined by calculating the difference of the averages of each method.
• The accuracy percentage is equal to the accuracy divided by the process variation.
• A good repeatability/reproducibility percentage value is below 30 percent.

Die Cast Problem Solving Page 17

 8D Step 2: Describe the Problem

• A good accuracy value is below 10 percent.

Gage Repeatability and Reproducibility

Gage Repeatability and Reproducibility (Gage R & R) is a statistical evaluation tool to
compare two people or two gages measuring the same parts. Gage R & R is most com-
monly used for measuring gage capability for critical casting dimensions. The standard
under QS-9000 in this case is a Gage R&R percentage of 10 percent. When used on
non-critical characteristics, the standard requirement is 30 percent. We typically follow
the 30 percent standard for gage capability because subjective defects and our defect
ranking system are not hard gages. However, you may desire to achieve higher gage
capability for difficult-to-resolve or high-cost defects.
The following describes how to complete a Gage R & R study:
Step 1: Randomly select 5 to 10 parts from the balanced sample of parts.
Select more parts for defects that are more difficult to evaluate.
Step 2: Using the defect ranking system boundary samples, measure the defect
level for all selected parts using two or three different people.
These should be the primary people accustomed to evaluating
product on a day-to-day basis.
Step 3: Measure the same parts in exactly the same way as in Step 2 over a
period of one or two more days.

Day 1 Day 2
Part Inspector 1 Inspector 2 Average Range Part Inspector 1 Inspector 2 Average Range Grand Avg.
1 1 2 1.5 1 1 1 1 1 0 1.25
2 2 2 2 0 2 2 2 2 0 2
3 3 3 3 0 3 4 3 3.5 1 3.25
4 2 2 2 0 4 2 2 2 0 2
5 4 5 4.5 1 5 4 5 4.5 1 4.5

Page 18 Die Cast Problem Solving

 8D Step 2: Describe the Problem

Step 4: Complete the table below with the results of the measurements.
Calculate the following statistics from the above data:
Inspector 1 Average = average of all data points from Day 1 and Day 2 = 2.5
Inspector 2 Average = average of all data points from Day 1 and Day 2 = 2.7
Parts (n) = 5
Trials (r) = 2 (for 2 days)
Inspectors (e) = 2
Grand Mean Average = average of all inspectors = 2.6
Grand Range Average = sum of all ranges/(n x r) = 4/(5 x 2) = 0.4
Difference Between Inspectors = absolute value of (Inspector 1 Average – Inspector 2
Average) = 0.2
Maximum Part Range = High Grand Average – Low Grand Average = 4.5 – 1.25 = 3.25

Calculate Gage Capability:

Error Between Inspectors (Repeatability)
= Grand Range Average x K1 = 0.4 x 4.56 < K1 from constant table in Appendix 1.
= 1.824
Error Within Inspectors (Reproducibility)
= [(Diff. Bet. Insp. x K2)2 – (Repeatability2/(n x r))]0.5 < K2 from constant table in Appendix 1.
= [(0.2 x 3.65)2 - (1.8242/(5 x 2))]0.5
= 0.447
Repeatability & Reproducibility (R&R)
= [(Repeatability)2 + (Reproducibility)2]0.5
= [1.8242 - 0..4472]0.5
= 1.878
Part Variation
= Max Part Range x K3 < K3 from constant table in Appendix 1.
= 3.25 x 2.08
= 6.76
Total Variation
= [(R&R)2 + (Part Variation)2]0.5
= [(1.878)2 + (6.76)2]0.5
= 7.016
% Repeatability
= 100 x Repeatability/Total Variation
= 100 x 1.824/7.030

Die Cast Problem Solving Page 19

 8D Step 2: Describe the Problem

= 26.0%
% Reproducibility
= 100 x Reproducibility/Total Variation
= 100 x 0.447/7.030
= 6.4%
% R&R
= 100 x R&R/Total Variation
= 26.8%
The above would be considered an acceptable gage since the Gage R & R % is less
than 30%. Hence, Iso-Plots and Gage R&R studies provide information on how good
our measurement system is. If we find that our measurement system is not acceptable,
it needs to be improved before we get father down the path of our problem solving. If we
conduct all of the experiments as will be described in Step 4 without a good measure-
ment system, we may not be able to determine if changes we make in the process have
improved or degraded our quality level.
It should be noted that some Gage R&R experts use the following as rules of thumb.
Gage %R&R that is 10% or less means the system is acceptable Gage %R&R less than
30% may be acceptable but there should be a plan to improve the system Gage %R&R
greater than 30% shows that the system needs to be improved
Therefore, although we are considering the 26.8% to be acceptable in the example
above, perhaps a plan to improve it should be adopted.
Microsoft Excel forms for completing Gage R & R studies are included with the
NADCA information available online at [Link]/docs/[Link] and a
Gage R & R calculation template is displayed in Appendix 2.

Page 20 Die Cast Problem Solving

Chapter VI
8D STEP 3: IMPLEMENT & VERIFY INTERIM CONTAINMENT

To assure that non-compliant or defective product does not get to the customer, the 8D
problem solving methodology requires that some sort of defect containment be imple-
mented. This defect containment is only an interim measure, which typically requires
extra inspection activity, to find product defects within the die cast facility. Since contain-
ment is required under the Eight Discipline process and is costly, it should motivate the
die caster to rapidly resolve the defect problem. The extra inspection may also provide
an opportunity to gather quality data that is difficult to gather without inspection. This
data may be very useful during the problem solving process.

Die Cast Problem Solving Page 21

 8D Step 3: Implement & Verify Interim Containment

Page 22 Die Cast Problem Solving

Chapter VII
8D STEP 4: DETERMINE & VERIFY ROOT CAUSES

Step 4 in 8D problem solving, which is “Determine and Verify Root Causes,” is the meat
of any problem solving effort. It is broken down into four sub-steps: 1) identify potential
causes, 2) select likely causes, 3) test to determine if the potential root cause(s) is a
root cause, and 4) select potential solutions to eliminate the root cause(s).

A) Identify Potential Causes

Clue Generation
The tools to identify potential causes are called clue generation tools. Each of these
tools is designed to provide more data, or clues, to determine what is causing a prod-
uct defect. For any defect problem, it is advisable to use multiple clue generation
tools.

Die Cast Problem Solving Page 23

 8D Step 4: Determine & Verify Root Causes

In this exercise, the task is to convert the arrangement of matchsticks (or use tooth-
picks) that form 5 squares into 4 squares by moving only two sticks. Rules:
- There can be no overlap of sticks, no one on top of the other
- No sticks can be removed, all must be used (None of the matchsticks can be
burned)
- All must be used to make the sides of a square
- There cannot be any incomplete squares
- The squares have to be the same size, the side is the length of one stick
Note your thought process as you solve the puzzle. Were there any clues that you
generated that helped solve the problem?
Chronic Cause vs. Special Cause Problem Analysis
To identify potential causes, you must first determine whether the defect problem is
a “chronic cause” or a “special cause” defect problem. A chronic cause problem is a
problem that is always there. In other words, the random normal variation of the pro-
cess may cause the problem or defect to occur at any time. A special cause problem is
a problem that comes and goes entirely due to human or mechanical intervention. We
must determine if the problem is chronic or special cause in order to determine which
tools to use in the problem solving project. If you are not certain whether a defect is
chronic or special cause, you must use tools for both approaches until you ascertain the
problem as one or the other. From this point forward, each tool described will be defined
as a good chronic cause tool or special cause tool for problem resolution.

Page 24 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

TOPS Is and Is Not Questions

The Ford Team Oriented Problem Solving (TOPS) Is and Is Not question methodology
is an excellent tool for beginning to understand special cause defect problems. This is
because these questions assist a problem solving team in exhaustively comparing the
current problem situation with other parts, dies, designs, times, situations, or processes
where the problem may not occur.
For example, two similar die castings exist in your company. You find that one has a
product defect problem. Using the Is and Is Not questions will enable you to more accurately
pinpoint the reason(s) for the problem by comparing the two parts’ processes, dies, designs,
etc. until you have generated an idea of what the differences are among the parts. These
differences then become clues as to what the root cause(s) of the problem might be.
Below you will find the Is and Is Not Questions for use in die cast problem solving:
IS IS NOT
1. Which part(s) has the defect? Which similar part(s) could have the defect,
but does not?
2. What is the defect? What could be the defect, but is not?
3. Where specifically do you see the Where specifically could you have seen
defect(s) on the casting? the defect(s) on the casting, but did not?
4. Where geographically and when in the Where geographically and when in
process flow do you first see the defect? process flow could have you first seen the
defect, but did not or could not?
5. When in time did you first see the When in time could you first see the
defect(s)? (hr. / day / mo. / yr.) defect(s), but did not?
6. Since you first saw the defect, when have Since you first saw the defect, when could
you seen it again? Can you observe any you have seen the defect again, but did not?
trends or patterns? Can you observe any trends or patterns?
7. What is the current state of the defect? What state could or should the defect be in
now, but is not? Is it getting better? Worse?
8. How big is the defect in terms of percentage, How big could the defect be, but is not?
dollars, people, time, or other resources?

Die Cast Problem Solving Page 25

 8D Step 4: Determine & Verify Root Causes

Defect Tally Sheet

The Defect Tally Sheet shown below is a tool that may be used for understanding the
occurrence rate of multiple defects. It is an effective tool to determine whether a defect
is a special cause or chronic cause defect. It also can help you determine if the defect
rate(s) change from hour to hour or shift to shift. Production employees should use the
Defect Tally Sheet to record the number of defects that occur each hour during the shift.

Defect Tally Sheet

Day Shift Hour Defect 1 Defect 2 Defect 3 Defect 4 Defect 5
1
2
3
1 4
5
6
7
8
1
2
3
1 2 4
5
6
7
8
1
2
3
3 4
5
6
7
8
1
2
3
1 4
5
6
7
8
1
2
3
2 2 4
5
6
7
8
1
2
3
3 4
5
6
7
8

Page 26 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

To analyze the defect tally sheet you must create a probability chart (p chart) for each
defect. You may then easily understand the average defect rate and variability for each
defect studied as it is easier to see variability and trends on a chart as compared to a
table of numbers.

Process Flow Diagram

The process flow diagram is a tool to resolve special cause defects. It is only valuable
when the defect is not created in the die casting process, but occurs at some process
step after die casting. It is especially useful for handling damage defects, as shown in
the example, because you can relate the defect rate with each process step.

Process Flow Diagram

Batch Sizes Manufacturing Operations Quality Checks/Audits Defect Mode
Operator hourly
3,000 Die cast Handling damage 1%
Quality shiftly
10,000 Tumble Handling damage 6%
300 Thermal Deburr Quality audit Handling damage 8%
5,000 Cleaning
5,000 Plating Handling damage 12%
5,000 Washing Quality audit
20,000 Ship 100% sort Scrap 12%
1 Assemble

Multi-Vari Study
The Multi-Vari study, developed by Dorian Shainin, is the most valuable clue generation
tool for chronic cause defect problems. The purpose of the study is to determine the
“Time Mode of Variation” of the product defect problem. The “Time Mode of Variation” is
how a defect varies over time. For example, if the Multi-Vari shows that the defect varies
from piece to piece, then the clue is to look for causes that vary from piece to piece. If
the defect would vary from shift to shift, then the clue is that something is changing from
shift to shift. In summary, the reason that the Multi-Vari study is so valuable is that it can
significantly narrow the number of potential causes of a product defect problem. The
procedure for completing a Multi-Vari study is as follows:

Die Cast Problem Solving Page 27

 8D Step 4: Determine & Verify Root Causes

1) Determine the number of castings to be collected each hour and each shift, as well
as the length of the study in days. Make sure castings are collected during all shifts.
Difficult to resolve or expensive defects:
4 pieces/hour, 4 hours/shift, and 4 days/study
Moderately easy to resolve or moderately expensive defects:
3 pieces/hour, 3 hours/shift, and 3 days/study
2) Collect castings under normal process conditions. Assure that operators on all
shifts do not change the way that they run the process during the study. This may
be done by either collecting castings without their knowledge, or preferably by
involving them in the problem solving process so they appreciate the purpose of
the study. Assure that castings are collected in consecutive order over consecu-
tive hours, shifts and days.
3) Process the castings to the point where the defect is typically found.
4) Rank the castings using your defect ranking system.
5) Chart the results with the defect rank on the Y-axis and the casting collection
order on the X-axis.
6) Evaluate the results visually or statistically.
The following are examples of the results of two Multi-Vari studies:
The first example (below and on the cover of this book) shows a complete Multi-Vari
chart. To analyze this chart, you must compare the amount of variation within each po-
tential time mode of variation (i.e., piece-to-piece, hour-to-hour, shift-to-shift, and day-to-
day). On this chart, it appears that piece-to-piece variation is most significant, however
shift-to-shift variation also appears to be somewhat significant.

Page 28 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

In contrast, the charts in the second example below show a Multi-Vari study broken
down into individual piece-to-piece, hour-to-hour, shift-to-shift, and day-to-day charts. In
this example, it can be seen that the greatest amount of variation occurs on an piece-to-
piece basis. The second most significant variation occurs on an hour-to-hour basis. The
issue does not appear to be shift or day related.

Die Cast Problem Solving Page 29

 8D Step 4: Determine & Verify Root Causes

T.O.P.S. Is and Is Not Differences

As discussed earlier in the text, the Team Oriented Problem Solving Is and Is Not ques-
tions generate answers that may be used to generate clues. The differences are obtained
by comparing the answers to each Is and Is Not question, and then brainstorming expla-
nations for the differences in answers. Once brainstorming of these differences is com-
pleted, you should have created a list of differences that explain each set of Is and Is Not
questions. There should be at least one difference brainstormed for each set of questions.
A documentation sheet for the Is and Is Not Differences is included in Appendix 2.

Here is an example of Is and Is Not differences related to the key minder sticking prob-
lem mentioned in the Definition Statement section.
1) Ford ignition cylinders have a different key minder design.
2) The key minder does not snap into place until key is all way in.
3) The grease station is where the key minder is first held forward.
4) The key minder assembly doesn’t occur until the step before the grease station.
5) The defect was first seen between February and April in 1997.
6) There is $460,000 of risk per year in running with the key sticking defect.

Page 30 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Cause and Effect Diagram (Ishikawa)

The Cause and Effect Diagram, shown below, is a basic tool used to brainstorm potential
causes. This tool may be used for special and chronic cause defects, but is typically more
valuable for special cause defects. It is especially valuable in assuring that all possible
reasons for a problem are identified. For example, assume that your group brainstorms
different factors that operators may change from shift to shift because your product defect
problem is varying from shift to shift. Your group may forget to consider that your mea-
surement system may change because you now have different inspectors looking at the
defects from shift to shift. Using the Cause and Effect Diagram would have allowed your
group to consider this potential cause when they may not have otherwise.

To use the Cause and Effect Diagram, write the symptom or problem description in
the box at the far right. Then create categories for the six M’s (Man, Machine, Method,
Material, Measurement, and Management). Finally, brainstorm, with your problem solv-
ing team, potential causes that would fall under each of the six categories.

B) Select Likely Causes

The purpose of the previous section, “Identify Potential Causes,” is to develop clues
regarding the cause of a defect and then brainstorm potential causes using these clues.
This section will provide tools to narrow down the potential causes into a smaller set of
likely causes. The tools identified here are tools designed to enable the problem solving
team to logically eliminate unlikely potential causes.

TOPS Is and Is Not Theories

From the TOPS Is and Is Not Differences, your team should brainstorm theories that explain
the differences. The team should attempt to brainstorm multiple theories for each defined dif-
ference in order to explain the root cause of the differences. The same theories may be used
to explain several differences. These multiple theories are then considered potential root
causes to the defect problem. A documentation sheet for theories is available in Appendix 3.

Die Cast Problem Solving Page 31

 8D Step 4: Determine & Verify Root Causes

An example of Is and Is Not Theories related to the key minder casting problem is as
follows.
T1. Grease causes key minder to stick in inserted position.
T2. Changed die cast core draft on March 18, 1997.
T3. Key minder casting has burrs that catch key minder pocket side wall.
T4. Key minder pocket and casting design allow for too much “slop” which allows key
minder to bind in inserted position.
T5. Key minder pocket and casting design don’t allow enough “slop” which causes
key minder to stick in inserted position.
T6. Key blade has burrs on it that catch on bottom of key minder as key is being
pulled out.
T7. Key bitting pattern allows a steep ledge that catches bottom of key minder as key
is being pulled out.

Notice Theory 2 and reflect back on Difference 5!

TOPS Theories/Differences Matrix

From the TOPS Is and Is Not Theories and Differences, your team should now develop
a Theories/Differences Matrix. This matrix is used to assist you in rating the relative sig-
nificance of the theories that were brainstormed so that you may decide which theories
to test and/or implement.

Differences
Theories D1 D2 D3 D4 D5 D6 D7 D8 Score
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10

In the matrix shown above, your team decides if the theory explains, partially ex-
plains, or does not explain the difference noted for the Is and Is Not Questions. If the

Page 32 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

theory fully explains the difference, then enter a 1 in the corresponding cell. If the theory
partially explains the difference, enter a ½ in the cell. If the theory does not explain the
difference, enter a 0 in the cell. Take the sum of each row to determine the score of
each theory. Those theories with the highest scores are the likely root causes. Address-
ing these root causes in order of score should improve or eliminate the product defect
problem most effectively.
Selecting Likely Causes for Chronic Cause Defects
Differences
Theories D1 D2 D3 D4 D5 D6 D7 D8 Score
T1 0 1/2 0 1 N/A 0 N/A 0 1½
T2 1 1/2 1 1 N/A 0 N/A 1 4½
T3 0 1 1 1/2 N/A 0 N/A 1 3½
T4 0 0 1 0 N/A 1 N/A 1/2 2½
T5 1/2 1/2 1 1/2 N/A 1/2 N/A 0 3
T6 1/2 0 0 1/2 N/A 0 N/A 1/2 1½
T7 1 0 0 0 N/A 1/2 N/A 1/2 2
T8
T9
T10

In the example above, questions five and seven were not answered fully so they did
not have differences (D5 and D7) and theories. Seven theories were brainstormed. When
they were compared to the differences, a score was developed for each theory. From
these scores, Theory 2 (T2) would be the most significant theory to try first. The other the-
ories would be tested in descending score order until the defect is satisfactorily resolved.

Likely causes for chronic cause defects also need to be determined. Brainstorming
using the clues that we have generated will accomplish this task. If you have done a
complete job in clue generation, the brainstorming should be fairly easy.
First, the team leader or facilitator should distribute or display all the information found
during clue generation tools. Using the clues, the team members should brainstorm pos-
sible causes. The possible causes are typically process variables that explain the chronic
variation. The leader or facilitator should document all the ideas from brainstorming.
The next phase of chronic cause problem resolution will be to conduct a designed ex-
periment. Designed experiments containing more than five process variables are cum-
bersome in the die casting process; therefore, we must reduce the number of potential
causes to five or less at this point. To reduce the number of process variables below
five, the problem solving team should first collectively discuss the importance of each
process variable in its relation to the defect. It is important to consider the clues that
your team has generated during this discussion. If the team cannot reduce the number
of process variables below five during this process, a correlation analysis must be used.

Die Cast Problem Solving Page 33

 8D Step 4: Determine & Verify Root Causes

Correlation Analysis
A correlation analysis is rarely used but is a valuable tool in problem solving efforts. The
only time when this analysis must be used is when more than five variables remain after
brainstorming. The goal of using a correlation is to find which variables vary with the de-
fect, and eliminate the variables from consideration that do not vary with the defect. To
complete a correlation study, collect a small sample of castings over the time mode(s) of
variation established in the multi-vari study along with the process variable information.
Then, mathematically calculate your correlation using the Excel correlation spreadsheet
provided within the NADCA on-line files.
To evaluate a correlation study results, compare the calculated correlation values for
all measurable process variables to the defect rank level. The value for a correlation
will be between -1 and 1. If the value is a positive number, it means that you have posi-
tive correlation. A positive correlation indicates that when the defect level is lower, the
process variable level is lower, and when the defect level is higher, the process variable
level is higher. If the correlation value is negative, it means that when the defect is lower,
the process variable level is higher and vice-versa. The closer the correlation value is to
1 or -1, the stronger the correlation. A correlation value close to zero indicates a weak
correlation. Therefore, the process variables with correlation values closer to zero are
less important in your problem solving effort. You should eliminate process variables
from consideration until you have five or fewer variables remaining. The following ex-
ample illustrates how to interpret a correlation.
Correlation Analysis
Porosity Level Fast Shot Velocity Impact Pressure Vacuum Time Intensifier Rise Time
6 84 ips (2.13mps) 2120 psi (14.62 MPa) 4 sec 3 sec
5 84 ips (2.13mps) 2120 psi (14.62 MPa) 5 sec 4 sec
1 80 ips (2.03 mps) 2150 psi (14.82 MPa) 5 sec 1 sec
4 83 ips (2.11 mps) 2130 psi (14.69 MPa) 5 sec 2 sec
7 85 ips (2.16 mps) 2100 psi (14.48 MPa) 1 sec 3 sec
6 85 ips (2.16 mps) 2110 psi (14.55 MPa) 2 sec 1 sec
2 81 ips (2.06 mps) 2140 psi (14.75 MPa) 5 sec 0 sec
0 80 ips (2.03 mps) 2150 psi (14.82 MPa) 1 sec 0 sec
3 81 ips (2.06 mps) 2150 psi (14.82 MPa) 2 sec 1 sec
8 85 ips (2.16 mps) 2110 psi (14.55 MPa) 5 sec 5 sec

Correlation 0.96 -0.93 0.05 0.81

Positive Correlation Negative Correlation Positive Correlation Positive Correlation
Significant Significant Insignificant Somewhat Significant
Eliminate from Test

Castings were produced at various fast shot velocities, impact pressures, vacuum
times, and intensifier rise times. The porosity level of castings produced under each of

Page 34 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

the conditions was assessed. The ranking system was such that 0 indicated the lowest
and 8 indicated the highest porosity level. A correlation of a positive 0.96 of fast shot ve-
locity to porosity level indicates that there is a strong correlation and as fast shot veloc-
ity increases, porosity level increases. The negative 0.93 correlation of impact pressure
to porosity level indicates a strong correlation and that as impact pressure increases,
porosity level decreases. The correlation with the highest absolute value (or the one
that is closest to +1 or -1) is the one with the most significant impact. The positive 0.81
for intensifier rise time is somewhat significant, but less significant than impact pres-
sure. Again, the positive number indicates that as rise time increases, the porosity level
increases. The fact that vacuum time has a correlation number close to 0, indicated that
it is insignificant and can be eliminated from further testing.

C) Is the Potential Cause a Root Cause?

The most critical part of problem solving is to determine whether your theories or process
variables selected truly have an effect on the product defect problem you are trying to
resolve. At this point, if we are resolving a special cause problem, we have theories that
appear to explain the reason for the defect. If we are resolving a chronic cause problem,
we have selected five or fewer process variables that appear to fit the clues that we have
obtained on our product defect problem. Now we have to perform a test to see if we can
“turn the defect off” and “turn it on” with these theories or process variables.
Special Cause Defects
The advantage of special cause problem solving is that you can easily and indepen-
dently test your theories in attempt to resolve the problem. To complete a test on the
theories, you should first baseline your defect level before the test. Then you should
implement the change proposed by the highest scoring theory from the Theories/Differ-
ences Matrix. Finally, evaluate the change in the defect level. If an improvement is seen,
this change should be permanently implemented, given that the change is cost effective.
If the defect is totally eliminated, you need not implement any more theories. However, if
some defect remains, you should try your other theories in order of score until you have
exhausted your theories or you have eliminated the defect.
Interactive Experimentation – Importance of Interactions
Die casting is fairly unique among manufacturing processes. Along with injection molding,
these are the only processes where tens to hundreds of process variables must work in
harmony to produce a product that goes from raw material to finished product in a frac-
tion of a second. This is what makes understanding process interactions imperative to die
casters. The following illustration gives you an idea of how often interactions occur.
An interaction is when two or more process factors work together to affect a product
defect. Interactions are independent of the individual effect of each process parameter
involved. For example, assume we have two process variables, fast shot speed and
die temperature, that both significantly affect porosity on a given casting and the para-

Die Cast Problem Solving Page 35

 8D Step 4: Determine & Verify Root Causes

meters both affect porosity independently. This is not an interaction because they do not
work together to affect porosity.

When process factors are tested in and experiment, each factor is tested at two or
three levels. Assuming a two-level experiment is being done, an interaction is when the
defect result differs when the two process factors are at the same level versus when
they are at different levels. More information on interactions is presented in the example
experiments in Chapter 12 (Problem Solving Examples).
Examples of interactions:
The following IS NOT an interaction:
Fast Shot Velocity Die Temperature Casting Rank
1 1 0
2 1 1
1 2 2
2 2 3

The following IS an interaction:

Fast Shot Velocity Die Temperature Casting Rank

1 1 0
2 1 1
1 2 3
2 2 2

Blocking
Blocking is a powerful tool in designed experimentation. Its value lies in that it reduces un-
explained variation that could be caused by known process variables that are not included

Page 36 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

in the designed experiment. Thus, when conducting an experiment, blocking allows the
effect of the experimental variables to be more apparent. When statistically evaluating your
experiment, proper blocking will raise the statistical confidence of your experiment.
The use of blocking is necessary because some process variables may vary with the
time mode of variation and are not possible to control in the long term due to cost or
lack of technology. Examples of such process variables in die casting may include die
water temperature, plant ambient temperature, air pressure, or operator variation. Be-
fore conducting an experiment, any of the selected process variables that are not pos-
sible to control in the long term should be considered blocked variables.
The blocked variable(s) must remain constant during the experiment. Blocking is ac-
complished by holding the blocked variable(s) at a nominal value during the experiment,
or by waiting for the nominal value to occur before collecting castings. For example, as-
sume you have cycle time variation because you have a manually operated machine. To
block for cycle time, you should use a stopwatch to carefully time the machine to make
sure that the cycle rate is the same during the experiment. If you wanted to block for
fast shot velocity, you might wait until you meet/achieve the nominal value on a process
monitoring system to collect castings for the experiment.

Randomization
Even after clue generation and several effective team meetings, you will likely still have
process variables that the team did not consider that cause some variation in your de-
fect. To minimize the effect of these factors during an experiment, we must randomize
the order of the experiment.
In order to realize the importance of randomization, we must consider the situation where
we do not randomize. For example, assume we complete a two-factor, two-level experi-
ment. Factor 1 is Fast Shot Velocity and Factor 2 is Intensification Pressure. Fast Shot
Velocity will be tested at 90 and 100 inches per second and Intensification Pressure will be
tested at 1200 and 1500 pounds per square inch. The experiment would appear as follows:
Treatment Fast Shot Velocity Intensification Pressure
1 90 ips (2.29 mps) 1200 psi (8.27 MPa)
2 100 ips (2.54 mps) 1200 psi (8.27 MPa)
3 90 ips (2.29 mps) 1500 psi (10.34 MPa)
4 100 ips (2.54 mps) 1500 psi (10.34 MPa)
Assume that an air line blows on a nearby trim press sometime between treatments
2 and 3 of the experiment, and the machines air pressure drops from 100 PSI to 25 PSI
for the duration of your experiment. This loss of air pressure causes your die spray to no
longer atomize, and therefore your die temperature changes significantly. You then ana-
lyze the results of your experiment. The castings show that you have significant poros-
ity on the castings from treatments 3 and 4, but treatments 1 and 2 are much better. Is

Die Cast Problem Solving Page 37

 8D Step 4: Determine & Verify Root Causes

this porosity difference due to the Intensification Pressure or is it due to the air pressure
drop? You will never know because you did not randomize the experiment.
To effectively randomize the experiment, the order of your experimentation should
not have any patterns that would allow other variation to have a detrimental effect on
the experiment. The best way to randomize is to use the random number tables sup-
plied in each Excel experiment spreadsheet file. The random number table will change
every time that you use the software. Therefore, to use the table, begin at the upper left
corner. The first number is the first treatment that should be conducted. The number to
the right should be the second treatment conducted and so on until all treatments in the
experiment are accounted for. If you have a duplicate treatment number, ignore it. If you
complete the first row without completing the experiment, go to the second row.
Three-Level versus Two-Level Experiments
We have the choice of running an experiment with two parameter levels or three param-
eter levels for each process variable. The advantage of two-level experiments is that
they are much quicker to complete and they allow you to use up to five process vari-
ables. The disadvantage of two-level experiments is that they only indicate a better or
worse process. Thus, you do not know which is the best process.
The advantage of three-level experiments is that you are able to understand much
more about the process. A three-level experiment, when used properly, can enable
quick process optimization. The disadvantage of three-level experiments is that you
may only use up to three factors in the experiment and the experiments are very time
consuming. For example, a three-level experiment with three factors requires 27 treat-
ments. A two-level experiment with the same factors only requires eight treatments. The
following chart illustrates the benefit of three-level experiments:

Page 38 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Factor Level Selection

The most significant reason for the lack of success in designed experimentation for
most experimenters is that factor levels are not selected properly. For a two-level experi-
ment, the basic rule is that factor levels should be set at plus and minus three standard
deviations from the current nominal process setting. If process monitoring equipment
does not exist for the process factor, then the rule is that the factor levels should be set
at a small increment above and below the current setting. For a three-level experiment,
use the same rules except use the current setting as the third level. The following charts
illustrates the above concept:

Die Cast Problem Solving Page 39

 8D Step 4: Determine & Verify Root Causes

Selecting process levels that are known to produce castings of unacceptable quality
level is not useful. If the threshold levels that produce unacceptable quality parts are not
known, start with reasonable, not absurd, levels. Also keep in mind that when we select
a level, there is always some variation about that "mean" level/setting.

Confounding
In fractional factorial designed experiments, some combinations of factors are not
tested. When these experiments are used, the intent is always to assume higher level
interactions or sometimes all interactions do not exist. However, when fractional facto-
rial experiments are conducted, these interactions may occur. When these interactions
occur in fractional factorial experiments, the high level interactions are confounded with
main effects or lower level interactions. For example, when a 25-1 fractional factorial
experiment is evaluated, the result of a main effect is confounded with a 4-factor inter-
action. In this case, you cannot tell if the calculated result is due to the main effect, due
to the 4-factor interaction, or due to some combination of the two effects. Because the
likelihood of the 4-factor interaction being significant is low, you assume that it does not
exist. This assumption is the reason that fractional factorial experiments are statistically
risky. However, the benefit of using these experiments often outweighs the risk.

Experiment Designs
In the past, the statistical community referred to the field of statistical problem solving as
design of experiments, or DOE. This term was used because experimenters toiled over
approaches to designing experiments for problem resolution. The act of designing the
experiment is not an issue in die casting due to the fact that we are limited in the num-
ber of effective experimental designs available. This limitation stems from the interactive
nature of die casting, as most experimental designs are for non-interactive problems.
We are further limited in options for effective experimental designs because experi-
ments with greater than five factors and greater than three levels are too large and time
consuming to use. The experiment selection chart below shows the 12 choices of exper-
iments for five or less factors and for two or three levels. To determine which experiment
to use, find the experiments that have the proper number of factors. Then choose the
number of levels desired and determine if confounding will be a concern.
The example experiments in Chapter 12 deal with full factorial and fractional factorial
designs. Full factorials test all possible combinations of factors and levels. For a two-
factor two-level full factorials design there are 4 possible combinations Two levels raised
to the second power (22). For a Three-factor two-level full factorial there are 23 or 8 pos-
sible combinations. A three-factor three-level full factorial would have 33 or 27 possible
combinations. With this number of treatments one can see why using a fractional facto-
rial design, or only a portion of the full fractorial may be desireable.
Below are examples of a two-factor two-level and three-factor two-level full fractorial
design where -1 represents the low level and 1 represents the high level of the factors.

Page 40 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Treatment Factor A Factor B

1 -1 -1
2 -1 1
3 1 -1
4 1 1

Treatment Factor A Factor B Factor C

1 -1 -1 -1
2 -1 1 -1
3 1 -1 -1
4 1 1 -1
5 -1 -1 1
6 -1 1 1
7 1 -1 1
8 1 1 1

Fractional Factorial
ASQ (American Society for Quality) defines fractional factorial design in the following
way: "A factorial experiment in which only an adequately chosen fraction of the treat-
ment combinations required for the complete factorial experiment is selected to be run."
Even if the number of factors, k, in a design is small, full factorial can quickly become
very large. 2K= number of treatments for two-level experiments. If there are 5 factors,
the number of treatments is 32!
An example of how to convert a full fractorial design to a fractional factorial design
is shown below. Full factorial looks like this for two-level three-factor experiment. The
negative 1’s represent the low level and the positive 1’s represent the high levels. A, B
and C represent three factors/parameters.
A B C
-1 -1 -1
1 -1 -1
-1 1 -1
1 1 -1
-1 -1 1
1 -1 1
-1 1 1
1 1 1

To generate the adequately chosen portion of the full fractorial treatments, break
down the full fractorial matrix to a 22 or two-factor two-level matrix, then interact fac-

Die Cast Problem Solving Page 41

 8D Step 4: Determine & Verify Root Causes

tors A and B. Next, let A*B represent the C levels. The 8 treatments is now reduced to
4 or 23-1.

A B A B A*B A B C
-1 -1 -1 -1 1 -1 -1 1
1 -1 1 -1 -1 1 -1 -1
-1 1 -1 1 -1 -1 1 -1
1 1 1 1 1 1 1 1

Experiment Selection Chart

Number of Number of Main Effects 2 Factor Interactions Quality of
Experiment Treatments Type
Factors Levels Confounded with Confounded with Experiment
A 21 2 B vs. C 1 2 N/A N/A Good
B 31 3 B vs. C 1 3 N/A N/A Good
C 22 4 Full Factorial 2 2 Nothing Nothing Good
D 23-1 4 Taguchi F. F. 3 2 All Interactions N/A Bad
E 23 8 Full Factorial 3 2 Nothing Nothing Good
F 24-1 8 Fractional Factorial 4 2 3 Factor Interactions 2 Factor Interactions Mediocre
G 32 9 Full Factorial 2 3 Nothing Nothing Very Good
2w/2
H 22x31 12 Full Factorial 3 Nothing Nothing Good
1w/3
I 24 16 Full Factorial 4 4 Nothing Nothing Good
J 25-1 16 Fractional Factorial 5 5 4 Factor Interactions 3 Factor Interactions Very Good
2w/3
K 32x21 18 Full Factorial 3 Nothing Nothing Good
1w/2
L 33 27 Full Factorial 3 3 Nothing Nothing Good

Replications
More than one casting per treatment must be collected in order to improve the power
of an experiment. Choosing the appropriate number of replications will increase the
likelihood that other types of random variation will not lower the confidence of the ex-
perimental results, assuming that the proper variables were initially selected for the
experiment. A rough guideline for choosing the number of replications is shown on the
following page.

Page 42 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Number of Cost of Prob- Suggested Number

Treatments lem per Year of Replications
2 to 4 < $10,000 5
2 to 4 $10k to $100k 8
2 to 4 > $100k 12
8 to 12 < $10,000 8
8 to 12 $10k to $100k 12
8 to 12 > $100k 15
16 to 18 < $10,000 10
16 to 18 $10k to $100k 15
16 to 18 > $100k 20
> 20 < $10,000 15
> 20 $10k to $100k 20
> 20 > $100k 25
Conducting an Experiment
To complete an experiment, your team should follow these steps:
1) Determine the experiment from the experiment selection chart.
2) Determine factor levels based upon historical data.
3) Decide how the blocks will be applied.
4) Randomize the experiment order.
5) Determine the number of replications for each treatment.
6) Determine a start time for the experiment.
7) Complete the experiment utilizing team members to:
a. collect and mark castings.
b. maintain and monitor blocks.
c. collect process monitoring data.
d. manipulate process factors.
e. document all events during the experiment.
8) If necessary, process the experimental castings to the point where the defect is evident.
9) Measure the defect level using the defect ranking system.
10) Enter the casting defect results into the proper experiment spreadsheet to determine:
a. the best and worst process combinations.
b. the most significant process factors to control.
c. the statistical confidence of your experiment.

Die Cast Problem Solving Page 43

 8D Step 4: Determine & Verify Root Causes

Using Spreadsheets
Included with this document is on-line access that contains 12 Microsoft® Excel spread-
sheets for experiment calculations and documentation. Go to [Link]-
[Link]/docs/education/[Link] to get the course files. Directions for use and application
are provided within the spreadsheet. When using a spreadsheet, always save a copy of
the file first under a different name and use the renamed copy as your working spread-
sheet. Examples of the application of the spreadsheets are in chapter 12.
The spreadsheets can be found at the following link: XXXXXXXXX (website).

Calculating Effects
The spreadsheets noted above automatically calculate all effects for any of the experi-
ments in the selection chart. However, if you are evaluating an experiment manually,
use the following formulas:
Σ 2’s – Σ 1’s
Main Effects: Effect = -------------------------------
(# of Treatments/2)

Where "2" indicates a high level and "1" indicates a low level of the parameters/fac-
tors.
Σ Same – Σ Different
Two-Factor Interactions: Effect = -------------------------------
(# of Treatments/2)

Where "same" indicates all treatments where the factors were all set at a high level,
all set at an intermediate level, and all set at a low level. "Different" indicates the treat-
ments where the factors were not set at the same level.
Σ Even #1’s – Σ Odd #1’s
General for all effects: Effect = -----------------------------------
(# of Treatments/2)

Where "even #1's" indicates the treatments where 2 or 4 factors were set at the low
level and "odd #1's" indicates the treatments where 1, 3 or 5 factors were set at the low
level.
For example, if you are calculating the effect for a two-factor interaction, first calcu-
late the sum of the results of the experiment when the levels of the two factors are the
same. Then subtract the sum of the results of the experiment when the levels of the two
factors are different. Divide this result by one-half of the number of treatments.

Page 44 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Determining Significance
The spreadsheets noted above automatically calculate significance for any of the ex-
periments in the selection chart. However, if you are evaluating an experiment manually
follow these steps:
1) Summarize the results of all the effects.
2) Take the absolute value of each of the effects.
3) Place the effects in descending order.
(The largest absolute value is the most significant effect and so on.)

Determining Best and Worst Process Conditions

To determine the best and worst processes, you cannot necessarily use treatment results.
You must determine the best and worst processes based on the significance of the effects.

To determine the best process, follow these steps:

1) Find the effect with the highest significance.
2) If the effect is a main effect (effect of a single factor, not of an interaction), then a posi-
tive value indicates that the best process for the factor is at level two. If negative, then
the best process for the factor is at level one.
3) If the effect is a two-factor interaction, then a positive value indicates the best pro-
cess for the two factors is the one in which the two factors have the same level.
A negative value indicates the best process for the two factors is the one in which
the two factors have different levels.
4) If the effect is an interaction of three or more factors, then a positive value in-
dicates the best process for the factors in the interaction is one with an even
number of ones. A negative value indicates the best process for the factors in the
interaction is one with an odd number of ones.
5) Find the effect with the next highest significance and return to Step 2.
6) Continue Steps 2 through 4 until all factors are defined.
To determine the worst process, follow these steps:
1) Find the effect with the highest significance.
2) If the effect is a main effect, then a positive value indicates that the worst process
for the factor is at level one. If negative, then the worst process for the factor is at
level two.
3) If the effect is a two-factor interaction, then a positive value indicates the worst
process for the two factors is the one in which the two factors have different lev-
els. A negative value indicates the worst process for the two factors is the one in
which the two factors have the same level.
4) If the effect is an interaction of three or more factors, then a positive value indi-

Die Cast Problem Solving Page 45

 8D Step 4: Determine & Verify Root Causes

cates the worst process for the factors in the interaction is one with an odd num-
ber of ones. A negative value indicates the worst process for the factors in the
interaction is one with an even number of ones.
5) Find the effect with the next highest significance and return to Step 2.
6) Continue Steps 2 through 4 until all factors are defined.
Determining the Best and Worst Processes
Effect Type Best Process Worst Process
+ 2 1
Main Effect
- 1 2
2 Factor + Same Different
Interaction - Different Same
+ Even # of 1's Odd # of 1's
General
- Odd # of 1's Even # of 1's

Process Optimization
When defects exist, even in the best process found by using the problem solving strat-
egies to this point, process optimization may be needed to further minimize defects.
Optimization is often referred to as “response surface optimization” by statisticians. This
reference roots from the way the process is graphed. For example, assume we have
two interacting factors that have been proven critical to a defect. These factors are Lube
Amount and Metal Temperature. We complete a two-factor, two-level experiment and
find the result below:

When reviewing the graph, imagine that the casting rank is represented by height.
To find the peak height of the response surface, move in the direction of highest cast-
ing rank 4 (upper left on the graph). To continue in search of the peak, we must test the
three other points to the upper left.

Page 46 Die Cast Problem Solving

 8D Step 4: Determine & Verify Root Causes

Once we test these three process combinations, we find that the best direction to find
the peak height is now to the lower left. We then test another three process combina-
tions to the lower left. After this test we find that all points have a casting rank of 4.5. If
all points are equal, then the center is likely the peak, or the optimum process. Next we
test the center point of the box and find that the average casting rank is 4.75. We have
now found the optimum process for these two factors, assuming that a sound problem
solving process has been followed to this point. This may be the best possible process.

D) Identify Alternative Solutions

Brainstorming
If you are solving a special cause problem, brainstorming in the team environment is
critical to finding alternative solutions once the root cause(s) has been established. In
brainstorming, your team should identify multiple approaches to eliminating the root
cause(s) to make sure that the best solutions are considered.
Once the team has determined potential solutions, they should analyze them all and
select the best solution(s) required to eliminate the root cause(s).
If you are solving a chronic cause problem, defining alternative solutions consists of
finding technologies and methods to maintain the best process conditions.

Determine Corrective Action Plans

For both chronic and special cause problems, finding the solution(s) to the root cause(s)
may require significant efforts on the part of the team and other individuals. To assure
solutions are implemented in a timely and effective manner, corrective action plans
should be developed and agreed upon by those with the respective responsibilities.
Good corrective action plans will help the team eliminate the defect as soon as possible.

Die Cast Problem Solving Page 47

 8D Step 4: Determine & Verify Root Causes

Page 48 Die Cast Problem Solving

Chapter 8
8D STEP 5: VERIFY CORRECTIVE ACTIONS

Special Cause Problems

To verify special cause problem resolution, we need to prove that we can turn the defect
“off” and back “on” again. This test is the best verification of a solution. If the problem
may not be turned off and on due to die changes or other hardware changes, ensure
that the defect rate and process conditions are documented so that the effect of the
solution(s) is clear.

Chronic Cause Problems

To verify chronic cause problem resolution, the experiment conducted may be evaluated
in one of two ways. ANOVA, which is the statistical technique taught in most design of
experiments courses, does not require more testing of the process. However it is statis-
tically flawed in most die casting situations. The second, B vs. C, does require another
experiment to be conducted but is the more desirable way to confirm results.

Analysis of Variance (ANOVA)

ANOVA is used to evaluate the confidence in the results of a designed experiment.
ANOVA is based upon the statistical F test. The F test compares two normal distribu-
tions to determine if they are significantly different from one another. It does this by com-
paring the magnitude of the effect of a process change to the background variation in
the statistical population of data. The problem with ANOVA as it relates to die casting is
that most defects such as porosity, cracks, non-fill, and casting distortion vary in a one-
tailed distribution and not a normal distribution. The resulting confidence values from the
ANOVA approach are then significantly lower than reality.
A simplified example below will be used to explain how ANOVA works:
Assume we have completed a two-factor, two-level experiment. The two most critical
effects and their relative significance values are:

Effect Name Effect Value

1 Int F.T./G.V. 1.125
2 Gate Velocity -0.875

Die Cast Problem Solving Page 49

 8D Step 5: Verify Corrective Actions

To complete the ANOVA, the Sum of Squares Table must first be completed. This
may be done by first finding the Grand Average of the experiment data. Then by sub-
tracting the average from each data point in the experiment and squaring the result.
These results are referred to as the “difference squares.” Add up all the difference
squares to find the Grand Sum of Squares. The Grand Sum of Squares Table indicates
the total amount of variation that occurred in the experiment.

Experiment Data Table

Trial 1 Trial 2 Trial 3 Trial 4
3 2 4 3
3 2 3 2
1 2 3 1
3 3 1 3
Grand Average 2.438

The experimental data table displays the ranking of the castings produced in the 4 tri-
als or treatments. The grand average is the sum of all 16 data points divided by 16.
Sum of Squares Table
0.316 0.191 2.441 0.316
0.316 0.191 0.316 0.191
2.066 0.191 0.316 2.066
0.316 0.316 2.066 0.316
Grand Sum of Squares 11.938
Each entry in the sum of squares table was calculated by taking the casting rank
level, subtracting from it the grand average of 2.438 and squaring the result. For ex-
ample, the upper left entry of 0.316 was calculated by subtracting 2.438 from 3, which
is 0.562, and squaring 0.562. The result is 0.315844. Rounding to three decimal places
yields 0.316.
The ANOVA table may then be created by using the above data.

ANOVA Table
Effect Sum of Squares Deg. of Freedom Mean Square F-value Confidence
Int F.T./G.V. 1.2656 1 1.2656 1.661 78.0%
Gate Velocity 0.7656 1 0.7656 1.005 66.6%
Residual 9.9063 13 0.7620

Sum of Squares for each effect = (Effect Value)2

Residual Sum of Squares = Grand Sum of Squares – Sum of Squares of all effects
Degree of Freedom for each effect = 1

Page 50 Die Cast Problem Solving

 8D Step 5: Verify Corrective Actions

Degree of Freedom for Residual

= (# Treatments x # Replications) – Degrees of Freedom for all effects – 1
Mean Square = Sum of Squares I Degree(s) of Freedom
F-value = Mean Square of each effect I Mean Square of Residual

Percent Confidence*
= (1 – Fdist (F-value, Degree of Freedom effect, Degree of Freedom Residual)) x 100
*Note: this is an Excel function. The Excel spreadsheet must be used to complete
this calculation.
The F-value is used for the F-test, which is a statistical test. The F-test compares the
experimentally controlled variation to the remaining random uncontrolled variation. The
more experimental control over the process, the higher the resulting confidence.
The confidence column shows the confidence from the F-test. If the confidence is 95%,
there is a 5% chance that the results of the experiment are due to random variation. Typi-
cally, you want to have at least a 75% confidence for each effect. If you are concerned about
the confidence of your experiment, you should complete the B vs. C test below.

Confirmation Run without Statistics (B vs. C Test)

The B vs. C Test should be used if: 1) you do not want to calculate the confidence using
ANOVA; 2) you are not comfortable with your ANOVA results; or 3) you would like to
confirm your experiment results. This test compares the best process to the worst pro-
cess to prove that the results of your experiment are not random.

Die Cast Problem Solving Page 51

 8D Step 5: Verify Corrective Actions

The B vs. C Test is often more powerful than ANOVA in determining the confidence be-
cause you are directly comparing two processes rather than several processes (treatments).
The B vs. C Test works well in the die casting industry because it is based on a one-
tailed distribution. Die casting defects always have a one-tailed distribution. Porosity,
surface defects, mis-runs, heat checks, blisters, etc., are all one-tailed distributions be-
cause they cannot improve beyond perfect. The goal of this test is to show that the B, or
better process, has a tighter distribution with a shorter tail than the C, or current (worse)
process, that has a wider distribution with a longer tail.
To complete a B vs. C Test, you must collect castings during the best and worst pro-
cesses. You should collect three castings minimum and up to 32 castings for both the
best and worst processes. To prove that the best process is truly better than the worst
process, you must first arrange all of the test castings in order from best to worst. Then
you must evaluate the end count of the worst process. The confidence for the confirma-
tion run may be determined by using the following Confidence Table:
B vs. C Confidence Table
Confidence Number of Castings Minimum End Count
14 – 16 7
18 – 28 8
99.9%
30 – 62 9
>64 10
10 – 12 5
99% 14 – 38 6
>40 7
6 3
8 – 30 4
95%
32 – 36 5
>38 6
8 – 16 3
90%
>18 4

If we consider the previous example used in the calculation of ANOVA, trial 3 has the
worst casting results and trial 2 has the best results. The ranking used for this table had 1
indicating the best quality level and 4 indicating the worse quality level. Using the B versus
C analysis, three out of four of the trial 3 castings are worse than all of the trial 2 castings.
Experiment Data Table
Trial 1 Trial 2 Trial 3 Trial 4
3 2 4 3
3 2 3 2
1 2 3 1
3 3 1 3

Page 52 Die Cast Problem Solving

 8D Step 5: Verify Corrective Actions

Therefore, since a total of eight castings are taken from trials 2 and 3 and the end
count is 3, the statistical confidence is 90%. This is much higher than the confidence
calculation (78%) from the ANOVA calculation.

Die Cast Problem Solving Page 53

 8D Step 5: Verify Corrective Actions

Page 54 Die Cast Problem Solving

Chapter 9
8D STEP 6: IMPLEMENT PERMANENT CORRECTIVE ACTIONS

Statistical Documentation
Documentation is critical to maintaining permanent corrective actions. When complet-
ing an involved problem solving project, documenting the defect ranking system and the
gains received through problem solving will enable you to verify the corrective actions at
any time. If the defect problem occurs again, the best tool to re-evaluate the process is
the Multi-Vari Study. Completing another Multi-Vari will allow you to evaluate the current
status of your process and provide a foundation for further problem solving if needed.

Identify New Process Knowledge

During the problem solving process, you will learn much about the process that you
are studying. The previous paradigms that may have limited problem solving should be
changed and this new knowledge should be investigated, communicated, and under-
stood by those in your organization who have responsibilities related to this particular
process.

Apply Lessons Learned

After the new process knowledge is identified, it should be applied. This new knowledge
is often the foundation for technological change within an organization. For example, if
you find that die temperature control would save five percent of defects on a given part,
this may justify die temperature control implementation. Most of the benefit of problem
solving comes from the lessons learned about your die casting process.

Die Cast Problem Solving Page 55

 8D Step 6: Implement Permanent Corrective Actions

Page 56 Die Cast Problem Solving

Chapter 10
8D STEP 7: PREVENT RECURRENCE

Long Term Verification

To prove that your solution(s) is robust enough to eliminate a defect, data must be used
to objectively illustrate this over time. Typically, the best way to verify defect elimination
over the long term is to monitor weekly and monthly quality reports while ensuring that
the procedures or technology changes are maintained on the process.

Take Advantage of Lessons Learned

The resolution of a defect frequently results in lessons learned about your company’s
technology and may be applied to many other processes. To fully take advantage of
what you have learned through your problem solving efforts, find ways to apply this
knowledge to many, or all, of your die casting processes.

Die Cast Problem Solving Page 57

 8D Step 7: Prevent Recurrence

Page 58 Die Cast Problem Solving

Chapter 11
8D STEP 8: CONGRATULATE YOUR TEAM

The team that has helped resolve the defect problem by this point has exemplified the
willingness to work and think above and beyond their typical responsibilities. Because
your company needs to promote involvement and foster teamwork, congratulating the
team that is involved in solving a defect problem is critical. If you want problem solving
efforts to be sustained in your company, finding ways to promote and nurture this activ-
ity is necessary.

Die Cast Problem Solving Page 59

 8D Step 8: Congratulate Your Team

Page 60 Die Cast Problem Solving

Chapter 12
PROBLEM SOLVING EXAMPLES

Generic Problem Solving Example: 2-Factor, 2-Level Full

Fractorial
The following example is from the ASQ Online DOE Tutorial. Attribution for this example
is gratefully paid to ASQ and Dean Christolear. This tutorial and more information on
DOE’s can be found at the ASQ website: [Link]
In this generic example, a DOE procedure utilizing 9 steps is described and shows
the impact of temperature and pressure on the strength of a glue bond. The steps are
as follows:
1. Acquire a full understanding of the inputs and outputs being investigated. A pro-
cess flow diagram or process map can be helpful. Utilize subject matter experts
as necessary.
2. Determine the appropriate measure for the output. A variable measure is prefer-
able. Attribute measures (pass/fail) should be avoided. Ensure the measurement
system is stable and repeatable.
3. Create a design matrix for the factors being investigated. The design matrix will
show all possible combinations of high and low levels for each input factor. These
high and low levels can be generically coded as +1 and -1. For example, a 2 fac-
tor experiment will require 4 experimental runs:

Input A Level Input B Level

Experiment #1 -1 -1
Experiment #2 -1 +1
Experiment #3 +1 -1
Experiment #4 +1 +1

Note: The required number of experimental runs can be calculated using the for-
mula 2n where n is the number of factors.
4. For each input, determine the extreme but realistic high and low levels you wish
to investigate. In some cases the extreme levels may be beyond what is currently
in use. The extreme levels selected should be realistic, not absurd. For example:

Die Cast Problem Solving Page 61

 Problem Solving Examples

-1 Level +1 Level
Temperature 100 F (37.8 C) 200 F (93.3 C)
Pressure 50 psi (0.34 MPa) 100 psi (0.69 MPa)

Temperature Pressure Strength

Experiment #1 100 degrees 50 psi 21 lbs
Experiment #2 100 degrees 100 psi 42 lbs
Experiment #3 200 degrees 50 psi 51 lbs
Experiment #4 200 degrees 100 psi 57 lbs

5. Enter the factors and levels for the experiment into the design matrix. Perform
each experiment and record the results. For example:
6. Calculate the effect of a factor by averaging the data collected at the low level
and subtracting it from the average of the data collected at the high level. For
example:
Effect of Temperature on strength: (51 + 57)/2 - (21 + 42)/2 = 22.5 lbs
Effect of Pressure on strength: (42 + 57)/2 - (21 + 51)/2 = 13.5 lbs
7. The interaction between two factors can be calculated in the same fashion. First,
the design matrix must be amended to show the high and low levels of the inter-
Input A Level Input B Level Interaction
Experiment #1 -1 -1 +1
Experiment #2 -1 +1 -1
Experiment #3 +1 -1 -1
Experiment #4 +1 +1 +1

action. Then the levels are calculated by multiplying the coded levels for the input
factors acting in the interaction. For example:
8. Calculate the effect of the interaction as before.

Page 62 Die Cast Problem Solving

 Problem Solving Examples

Effect of the interaction on strength: (21 + 57)/2 - (42 + 51)/2 = -7.5 lbs

9. The experimental data can be plotted in a 3D Bar Chart.

The effect of each factor can be shown on a Pareto Chart:
The negative effect of the interaction is most easily seen when the pressure is set to
50 psi and Temperature is set to 100 degrees.
Keeping the temperature at 200 degrees will avoid the negative effect of the interac-
tion and help ensure a strong glue bond

Die Casting Example 1: 2-Factor, 2-Level, Full Factorial

The following experiment was conducted to test the effect of die temperature and gate velocity
on surface finish. The two levels determined from historical data for die temperature were 400
and 450 degrees Fahrenheit (204 and 232 degrees Celcius), and for gate velocity were 1400
and 1600 inches per second (35.6 and 40.6 meters per second).
The following experiment design was used:

Treatment Die Temperature Gate Velocity Random Order

1 (1) 400F/204C (1) 1400ips/35.6mps 4
2 (2) 450F/232C (1) 1400ips/35.6mps 2
3 (1) 400F/204C (2) 1600ips/40.6mps 1
4 (2) 450F/232C (2) 1600ips/40.6mps 3

Five castings were collected at each setting during the experiment. The castings
were ranked. Here, the ranking system was 0-5, with 0 being the lowest quality level
and 5 being the highest quality level. The results of the experiment are as follows:

Treatment D.T. G.V. Casting Ranks Average

1 (1) 400F/204C (1) 1400ips/35.6mps 12121 1.4
2 (2) 450F/232C (1) 1400ips/35.6mps 23212 2.0
3 (1) 400F/204C (2) 1600ips/40.6mps 32323 2.6
4 (2) 450F/232C (2) 1600ips/40.6mps 12101 1.0

Die Cast Problem Solving Page 63

 Problem Solving Examples

Trial Die Temp Gate Vel Results 1 2 3 4 5

1 1 1 1.4 1 2 1 2 1
2 2 1 2 2 3 2 1 2
3 1 2 2.6 3 2 3 2 3
4 2 2 1 1 2 1 0 1
Effects Value Absolute value Significance
Die Temp -0.5 0.5 2
Gate Vel 0.1 0.1 3
Int F1/F2* -1.1 1.1 1
Die Temp Gate Vel
Best 1 2 1 2 -1 1
Worst 2 2 1 2
*Int F1/F2 is an abbreviation for interaction between Factor 1 and Factor 2 (between Die temperature and Gate Velocity).

Using the two-factor, two-level spreadsheet provided, the best and worst conditions
and the significance were determined.
Thus, the spreadsheet shows that the best conditions are die temperature at Level 1
and gate velocity at Level 2. This result agrees with Treatment 3, where the best result
was found. Also, the spreadsheet shows that the two-factor interaction between die
temperature and gate velocity is the most significant effect. The main effect of die tem-
perature is the second most significant effect.

Die Casting Example 2: 5-Factor, 2-Level, Fractional Factorial

A team at Unproductive Aluminum Inc. formed to resolve a non-fill problem on a widget
assembly bracket. The team developed a zero to five ranking system for the defects. The
team then conducted a Multi-Vari study and completed the TOPS Is and Is Not Questions.
The clues they generated were that piece-to-piece variation and shift-to-shift variation
were both significant. They knew that the defect was somewhat chronic because they
always had defects at varying levels. However, from the Is and Is Not Questions, they also
felt that the metal delivery and operator practices varied from shift to shift.
Once the team obtained their clues, they met to determine possible causes. They
determined their experimental factors to be:
1) Fast Shot Speed
2) Shot Pressure
3) Metal Temperature
4) Lube Amount
5) Cycle Time
The team blocked for metal level, die temperature, and cooling water temperature. They
blocked for metal level by having the metal deliveryman deliver fresh metal before each
treatment. They blocked for die temperature by monitoring thermocouples in the die next to

Page 64 Die Cast Problem Solving

 Problem Solving Examples

the defect, and internal water flow in local water lines. They blocked for cooling water tem-
perature by conducting the experiment late at night when the outside cooling tower tempera-
ture was consistent and by monitoring the water thermocouple in the cooling tower.
In developing the experimental design, the team used a fractional factorial design due to
testing five factors. The fractional factorial tests for main effects and two factor interactions.
The interaction between three or more factors is ignored.
The experiment design was as follows:

Trial Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
1 1 1 1 1 2
2 2 1 1 1 1
3 1 2 1 1 1
4 2 2 1 1 2
5 1 1 2 1 1
6 2 1 2 1 2
7 1 2 2 1 2
8 2 2 2 1 1
9 1 1 1 2 1
10 2 1 1 2 2
11 1 2 1 2 2
12 2 2 1 2 1
13 1 1 2 2 2
14 2 1 2 2 1
15 1 2 2 2 1
16 2 2 2 2 2

The team completed 16 replications for all 16 treatments. The team then ranked the
casting using their defect ranking system. The results shown in the calculation spread-
sheet were as follows:

Die Cast Problem Solving Page 65

 Problem Solving Examples

Replications 16
Trial Results 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 4.3125 4 4 5 5 5 4 3 4 5 5 5 4 4 4 5 3
2 2.25 2 2 3 3 2 3 2 2 2 3 3 2 2 2 1 2
3 4.3125 4 4 5 4 3 4 5 4 4 5 5 4 4 5 5 4
4 1.75 1 2 2 2 3 2 3 1 2 1 2 2 2 1 1 1
5 1.875 2 2 1 2 2 1 3 2 2 2 1 2 1 2 3 2
6 4.5625 5 5 4 5 4 5 4 5 4 5 4 5 4 5 4 5
7 2.3125 1 2 2 2 3 2 2 3 3 1 3 4 2 2 3 2
8 4.6875 5 5 4 5 4 5 5 4 5 5 4 5 5 4 5 5
9 1.1875 2 2 1 1 1 2 1 0 1 2 1 1 1 2 1 0
10 3.0625 2 3 4 3 2 3 4 3 3 3 4 3 2 3 3 4
11 1.625 2 1 2 1 2 2 2 1 2 2 1 1 3 1 2 1
12 3.75 3 4 4 4 5 4 3 4 4 4 2 4 5 4 2 4
13 4 4 5 3 4 4 4 4 4 4 5 4 3 4 5 4 3
14 1.375 1 2 1 1 1 2 2 2 1 1 1 2 2 2 1 0
15 3.5 3 3 3 4 3 3 3 3 4 3 4 3 5 5 4 3
16 1.5 0 3 2 1 2 1 1 1 1 2 1 2 2 1 2 2

The calculated effects and the significance from the spreadsheet were as follows:

Effects Value Significance

F1 Fast Shot Speed -0.023 13
F2 Shot Pressure 0.102 8
F3 Metal Temperature 0.195 5
F4 Lube Amount -0.758 2
F5 Cycle Time 0.070 10
Int F1/F2 0.008 14
Int F1/F3 0.133 6
Int F1/F4 -0.133 7
Int F1/F5 -0.320 3
Int F2/F3 -0.055 12
Int F2/F4 0.086 9
Int F2/F5 -2.289 1
Int F3/F4 -0.008 15
Int F3/F5 0.211 4
Int F4/F5 0.070 11

Page 66 Die Cast Problem Solving

 Problem Solving Examples

The best and worst processes tested were found to be:

Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 2 2 2 1 1
Worst 1 1 1 2 1
Because there are 16 other potential combinations that were not tested, we need to
evaluate the significance and whether the effect is a positive or negative value to be cer-
tain that we have the actual best and worst conditions. To do this, we first look at the effect
with the highest significance. This is the interaction between Factor 2 (Shot Pressure)
and Factor 5 (Cycle Time). This effect is a negative value. A negative value for an interac-
tion means that the levels for the factors in the best process must be opposite. That is, if
Factor 2 has a level of one, then Factor 5 must have a level of two, or vice versa. Further-
more, this means that the levels for the factors in the worst process must be the same.
Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 2
Best 2 1
Worst 1 1
Worst 2 2
The second significant effect is the main effect of Factor 4, which is Lube Amount.
This effect is a negative value. A negative value for a main effect means that the level
for Lube Amount in the best process must be one. Also, this means that the level for
Lube Amount in the worst process must be two.
Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 1 2
Best 2 1 1
Worst 1 2 1
Worst 2 2 2
The third significant effect is the interaction between Factor 1 (Fast Shot Speed) and
Factor 5 (Cycle Time). This effect is a negative value. A negative value for an interac-
tion means that the levels of the factors in the best process must be opposite. The worst
process must be the same level.
Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 1 1 2
Best 2 2 1 1
Worst 1 1 2 1
Worst 2 2 2 2

Die Cast Problem Solving Page 67

 Problem Solving Examples

The fourth significant effect is the interaction between Factor 3 (Metal Temperature)
and Factor 5 (Cycle Time). This effect is a positive value. A positive value for an inter-
action means that the levels for the factors in the best process must be the same. The
levels in the worst process must be different.
Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 1 2 1 2
Best 2 2 1 1 1
Worst 1 1 2 2 1
Worst 2 2 1 2 2
The fifth significant effect is the main effect of metal temperature. The sign is positive
therefore the best process must be 2 and the worst process must be 1.

Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 1 2 1 2
Best 2 2 1 1 1
Worst 1 1 2 2 1
Worst 2 2 1 2 2

Or

Fast Shot Speed Shot Pressure Metal Temperature Lube Amount Cycle Time
Best 1 1 2 1 2
Worst 2 2 1 2 2
Notice that both the best and worst processes were not tested in the experiment. This
shows the power of a fractional factorial design. That is, the experiment is able to ac-
curately find the best combination of factors by using interactions without testing every
combination…even if the best and worst combinations are not tested.

Die Casting Example 3: 5-Factor, 2-Level, Fractional Factorial

This problem solving exercise was completed on a lock cylinder at a die casting facility.
The problem was that these automotive lock cylinders had internal porosity holes that
connected with the lock tumbler spring pockets. The holes were referred to as “worm-
holes.” The defect would occur if the wormhole was large enough to allow the small
tumbler spring to fall into it. When these tumbler springs fell into the hole, the automo-
bile could not be started because the tumbler would fall in front of the key, which did not
allow the driver to insert the key fully into the ignition.
Until this problem solving exercise was completed, several engineers and outside die
casting consultants had tried to resolve the problem, but were unsuccessful. The defect
was costing the company several million dollars per year and was risking the relation-
ship with the company’s second largest customer.

Page 68 Die Cast Problem Solving

 Problem Solving Examples

To resolve this problem, a problem solving team was developed. The team followed
the Eight Discipline format and used many of the tools that have been discussed in this
book to resolve the problem. The steps taken to resolve this problem were as follows:
1. Team Approach
To begin the problem solving process, the company first established a team.
The team included two operators, two process engineers, one tooling engineer,
one quality engineer, and the die casting manager.
2. Problem Description
The team started the problem solving process by deciding upon a problem state-
ment that was similar to the description provided above. One team member cre-
ated a pictograph that was approved by the team. Then the team created a ranking
system. Because the defect was a multi-million dollar per year problem and the
defect was fairly easy to rank, the ranking system included 11 defect levels. The
team completed a Gage Repeatability and Reproducibility study and determined
that the ranking system achieved an acceptable 30% Gage R&R.
3. Implement and Verify Containment
To ensure that defects did not get to the customer, the team implemented 100% inspec-
tion after the die casting operation and at the end of the assembly of the lock. Even with
these added inspections, a few defective assemblies did make it to the customer.
4. Determine and Verify Root Causes
A. Identify Potential Causes (Clue Generation)
In this phase the team used the TOPS Is and Is Not Questions. From these ques-
tions the team determined that this casting’s defect was unique in the plant. How-
ever, this was the only significant clue found. Using the Multi-Vari study the team
found that Piece-to-Piece variation was significant. This made sense to the team
because it proved the chronic nature of the defect, which is why the Is and Is Not
Questions provided few clues. The Multi-Vari study showing the Piece-to-Piece,
Hour-to-Hour, Shift-to-Shift, and Day-to-Day variation is shown below:

Die Cast Problem Solving Page 69

 Problem Solving Examples

B. Select Likely Causes

When clue generation was completed, the next major step was to brainstorm fac-
tors for consideration in an experiment. To begin, the team reviewed the Multi-Vari
data and the Is and Is Not Questions. Then the team brainstormed eight process
variables (factors) that could easily vary with the time mode of variation shown in
the multi-vari results. These eight factors were:
1) Cavity Cleanliness
2) Cycle Time

Page 70 Die Cast Problem Solving

 Problem Solving Examples

3) Lube Amount
4) Metal Level
5) Metal Temperature
6) Nozzle Temperature
7) Shot Pressure
8) Shot Speed
 ight factors were listed, however, the team had to eliminate three in order to conduct
E
an experiment. The question that the team asked was, “Are any of these eight param-
eters not possible to control due to cost or lack of technology?” The answer was “yes.”
Two parameters, Metal Level and Cavity Cleanliness, could not be changed in the long
term because these factors were standard shop practices for many machines.
 he team now had six parameters remaining. They prioritized which factor was
T
least important to the defect and determined that Cycle Time was least important.
These three factors that are not to be used in the experiment were now to be used
as blocks for the experiment.

C. Is the Potential Cause a Root Cause?

 he experiment the team used was a five-factor, two-level, fractional factorial experi-
T
ment. The experiment tests for main effects and two factor interactions, and ignores
three, four and five factor interactions. The team conducted the experiment by manipu-
lating the five factors and maintaining the blocks. The Metal Level was blocked filling
the furnace to the same level before each treatment. Cavity Cleanliness was blocked
by cleaning the cores and inserts before each treatment. Cycle Time was blocked by
measuring and maintaining the cycle time for each shot during the experiment. Fifteen
replications for each treatment were taken. The experiment design was as follows:
Trial Factor 1 Factor 2 Factor 3 Factor 4 Factor 5
1 1 1 1 1 2
2 2 1 1 1 1
3 1 2 1 1 1
4 2 2 1 1 2
5 1 1 2 1 1
6 2 1 2 1 2
7 1 2 2 1 2
8 2 2 2 1 1
9 1 1 1 2 1
10 2 1 1 2 2
11 1 2 1 2 2
12 2 2 1 2 1
13 1 1 2 2 2
14 2 1 2 2 1
15 1 2 2 2 1
16 2 2 2 2 2

Die Cast Problem Solving Page 71

 Problem Solving Examples

The levels for each factor were chosen using historical data. The levels were:
Trial Lube Amount Shot Pressure Shot Speed Nozzle Temperature Metal Temperature
Level 1 0.10 sec 650 psi (4.48 MPa) 5.5 ips (0.14 mps) 890F (477C) 780F (416C)
Level 2 0.12 sec 750 psi (5.17 MPa) 6.5 ips (0.17 mps) 920F (493C) 800F (427C)

 he team completed the experiment as described above and then ranked the
T
castings using their defect ranking system. The results as shown in the calculation
spreadsheet were as follows:

Trial Results 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 4.47 2 7 7 4 8 7 0 2 2 1 5 8 5 7 2
2 5.07 8 1 1 5 0 0 10 8 8 9 8 5 2 10 1
3 4.40 5 9 0 6 10 0 1 7 7 2 0 0 7 5 7
4 6.33 6 2 10 5 0 10 8 5 8 7 10 2 4 8 10
5 4.73 4 7 0 5 8 7 1 2 8 2 8 8 6 0 5
6 4.60 4 1 10 3 0 4 9 4 2 6 10 2 2 2 10
7 4.60 3 9 0 4 10 7 1 10 10 2 1 1 7 2 2
8 4.67 6 2 10 4 1 2 9 9 1 7 4 7 3 1 4
9 2.67 2 8 2 3 10 3 0 0 0 1 0 1 6 2 2
10 5.40 5 1 9 3 0 8 8 6 10 6 10 5 2 8 0
11 4.87 3 8 2 5 10 2 1 6 6 2 2 8 6 2 10
12 5.07 6 2 10 5 0 8 8 4 6 7 8 2 4 6 0
13 5.40 4 8 0 5 10 8 2 3 10 3 10 4 7 2 5
14 5.47 5 1 8 3 0 10 7 10 9 6 10 0 2 1 10
15 5.20 5 9 1 7 10 9 2 1 0 4 9 5 8 0 8
16 5.87 7 2 4 6 0 9 10 7 7 9 10 4 4 9 0

The calculated effects and the significance from the spreadsheet were as follows:
Effects Value Significance
Lube Amount 0.77 1
Shot Pressure 0.40 6
Shot Speed 0.28 8
Nozzle Temperature 0.13 11
Metal Temperature 0.53 4
Int F1/F2 -0.05 13
Int F1/F3 -0.60 3
Int F1/F4 0.15 10
Int F1/F5 -0.05 14
Int F2/F3 -0.37 7
Int F2/F4 0.12 12

Page 72 Die Cast Problem Solving

 Problem Solving Examples

Effects Value Significance

Int F2/F5 0.05 15
Int F3/F4 0.70 2
Int F3/F5 -0.43 5
Int F4/F5 0.25 9

The best and worst processes tested were found to be:

Factor 1 Factor 2 Factor 3 Factor 4 Factor 5

Best 2 2 1 1 2
Worst 1 1 1 2 1
 he best and worst processes tested agree with the significance shown in the sig-
T
nificance table. This is based upon the following analysis:

Significance Factor(s) Sign Best Worst Agrees with Best/Worst

1 1 + 2 1 Yes
2 3–4 + Same Diff. Yes
3 1–3 – Diff. Same Yes
4 5 + 2 1 Yes
5 3–5 – Diff. Same Yes
*Since factor 2 has not yet been verified we go to #6 in significance.
6 2 + 2 1 Yes

Therefore, in this case, the best and worse processes have already been found.

D. Identify Alternative Solutions

The experiment results showed that the following process changes should be considered:
1. The Lube Amount main effect is positive. Therefore, the Lube Amount should
be set at level two which is 0.12 seconds.
2. The Shot Speed and Nozzle Temperature have a positive interaction. There-
fore, both should be set at level two or both should be set at level one.
3. The Shot Speed and Lube Amount have a negative interaction. Therefore,
since Lube Amount is at level two, then Shot Speed should be set at level one,
which is 5.5 inches per second (0.14 meters per second).. This means that
Nozzle Temperature should also be at level one, which is 890 degrees Fahren-
heit (477 degrees Celsius).
4. The Metal Temperature main effect is positive. Therefore, Metal Temperature
should be set to level two or 800 degrees Fahrenheit (427 degrees Celsius).
5. The Shot Pressure main effect is positive. Therefore, Shot Pressure should be
set to level two or 750 pounds per square inch (5.17 megapascals).

Die Cast Problem Solving Page 73

 Problem Solving Examples

5. Verify Corrective Actions

 lthough the best and worst processes were found, the team needed to prove that
A
the results were not random variation by completing a confirmation run. The team
used the B vs. C test. Twenty parts were taken from the worst process and 20
parts were taken from the best process in the B vs. C test. The parts were ranked
from worst to best process and the results were as shown below:
Worst Process Best Process
1 2
9 3
10 4
13 5
15 6
21 7
22 8
23 11
24 12
25 14
27 16
30 17
33 18
34 19
End 35 20
Count 36 26
=8 37 28
38 29
39 31
40 32

Page 74 Die Cast Problem Solving

 Problem Solving Examples

 rom the B vs. C evaluation table, the above results show a confidence of 99%.
F
Therefore, the experiment results were not random variation. However, some de-
fects still exist in the best process.
6. Implement Permanent Corrective Actions
Statistical Documentation
 he Multi-Vari information from the beginning of the problem solving process
T
shows the original state of the process. The experiment and the B vs. C test show
the current state of the process. If future process problems occur, a new Multi-Vari
could be completed and compared with the experiment and old Multi-Vari results to
evaluate quality and parameter settings.
Identify New Process Knowledge
The following process knowledge was gained from the experiment:
1. The Nozzle Temperature variation was unacceptable.
2. The experiment taught the team that the defect was not a Gas Porosity defect
as had always been thought, but it was a Shrinkage Porosity defect. The team
learned this because more spray would cause more gas porosity and less
shrinkage. Since the defect got better with more spray then the defect must
have been Shrinkage Porosity.
3. The experiment also taught the team that more spray contacting the tumbler
blade causes better quality.
4. The fact that the Shot Speed needed to be lower in the best process, told the
team that the velocity of metal in contact with the defect area was increasing
heat in the area.

Die Cast Problem Solving Page 75

 Problem Solving Examples

Apply Lessons Learned

 rom the new process knowledge gained in the experiment, the applied solutions
F
were as follows:
1. To set the process to the best conditions shown in the experiment.
2. To find a better way to control nozzle temperature. During the experiment the
temperature varied significantly and was difficult to control to the proper level.
The team found that with PID control, they could eliminate 95% of the nozzle
temperature variation.
3. Now that the team knew the defect was shrinkage porosity, this indicated that
the team should find a way to lighten the casting volume in the area of the
shrinkage. To do this, the team spoke with the design engineers of the lock.
The team found that metal savers could be placed in the area as long as the
customer approved the design change.
4. To get more spray contacting the tumbler blade, the team determined that a
rougher surface would hold more lubricant. Therefore, the team decided to try
shot peening of the die to roughen the die surface.
5. To specifically reduce the velocity of the metal in the die without reducing fill
time, the gate area must be increased. The team decided to increase the gate
area and redirect the metal flow away from the defect area.
The results of these changes are shown in the following chart:

Page 76 Die Cast Problem Solving

 Problem Solving Examples

7. Prevent Recurrence
Long Term Verification
 he means of long-term verification in this problem solving example was to evalu-
T
ate the number of defects at the assembly plant and at the customer’s site. Once
the backlog of product made it through the production system, there were no de-
fects found at the assembly plant or at the customer’s site.
Take Advantage of Lessons Learned
 nce this cause and effect relationship was established for wormholes, the team
O
found that some similar defects to wormholes existed on other parts. The knowl-
edge of the resolution to this defect was then communicated to the persons re-
sponsible for the other parts. This resulted in several die and process design
changes on other products at the facility.

Die Casting Example 4: 3-Factor, 3-Level, Full Factorial

This problem solving exercise was completed on a shower handle assembly. The
handle had to be plated and therefore could not have significant surface defects. The
problem was that unacceptable porosity occurred on the casting parting line. Approxi-
mately 20 percent of all products were defective. A five-factor experiment that had been
completed prior to this point had shown that dwell time, hot oil temperature, and vacuum
time were critical. The three-level, three-factor experiment was used to optimize the pro-
cess for minimizing or eliminating the defect. The experiment design was as follows:

Die Cast Problem Solving Page 77

 Problem Solving Examples

Trial Dwell Time Hot Oil Temp. Vacuum Time

1 1 1 1
2 2 1 1
3 3 1 1
4 1 2 1
5 2 2 1
6 3 2 1
7 1 3 1
8 2 3 1
9 3 3 1
10 1 1 2
11 2 1 2
12 3 1 2
13 1 2 2
14 2 2 2
15 3 2 2
16 1 3 2
17 2 3 2
18 3 3 2
19 1 1 3
20 2 1 3
21 3 1 3
22 1 2 3
23 2 2 3
24 3 2 3
25 1 3 3
26 2 3 3
27 3 3 3

The levels for each factor were chosen using historical data. The levels were:
Trial Dwell Time Hot Oil Temp. Vacuum Time
Level 1 5.00 sec 350F (177C) 1.50 sec
Level 2 5.75 sec 400F (204C) 1.75 sec
Level 3 6.50 sec 450F (232C) 2.00 sec

The experiment was completed and the castings were ranked using the previously
established defect ranking system. The results, as shown in the calculation spread-
sheet, were as follows:

Page 78 Die Cast Problem Solving

 Problem Solving Examples

Trial Results 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
1 4.321 4 4 5 5 4 4 3 4 4 5 5 4 4 5 3 4 4 5 5 4 4 4 5 4 5 5 5 4
2 4.357 5 5 4 4 3 5 4 4 5 4 5 5 4 3 4 4 5 5 4 5 4 5 5 4 5 3 5 4
3 4.429 5 5 4 4 5 3 5 5 5 4 4 4 5 4 5 4 3 5 5 4 4 5 4 5 5 4 4 5
4 3.714 3 4 4 3 4 4 5 3 3 4 4 4 4 2 3 4 4 3 4 4 5 3 4 4 3 4 4 4
5 4.286 4 4 4 4 5 5 5 4 4 4 4 5 5 5 3 5 5 5 4 4 4 4 5 3 4 4 4 4
6 3.929 5 4 3 2 5 4 5 4 3 2 5 4 5 4 3 5 5 4 3 2 5 4 5 5 4 3 2 5
7 4.214 4 5 4 5 4 3 4 5 4 5 4 3 5 4 5 4 3 4 5 4 2 5 4 5 4 5 4 5
8 3.821 4 3 4 4 4 5 4 4 2 4 4 3 4 5 4 3 4 4 5 2 4 4 3 5 3 4 4 4
9 4.321 4 3 4 5 4 4 5 5 4 4 5 4 5 3 4 5 3 5 4 4 5 4 5 5 4 5 4 5
10 4.214 5 4 5 3 5 4 5 2 5 4 5 3 5 4 5 4 5 3 5 4 5 2 5 4 5 3 5 4
11 3.964 5 4 3 2 1 5 4 3 2 5 4 3 2 5 4 3 5 4 3 5 4 5 5 5 5 5 5 5
12 4.000 2 3 4 5 3 4 5 3 4 5 3 4 5 3 4 5 3 4 5 3 4 5 4 5 4 5 4 4
13 3.714 5 4 3 2 5 4 3 5 4 3 5 4 3 4 3 4 3 4 3 4 3 4 3 4 3 4 4 4
14 3.536 1 2 3 4 5 2 3 4 5 2 3 4 5 3 4 5 3 4 4 3 4 3 4 4 4 4 3 4
15 4.250 5 4 3 2 5 4 3 5 4 3 5 4 3 5 4 3 5 4 5 4 5 4 5 5 5 5 5 5
16 3.893 2 3 4 5 2 3 4 5 2 3 4 5 3 4 5 3 4 5 4 5 4 5 4 5 4 4 4 4
17 4.107 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 5 4 5 4 5 4 4 4 4 4 4 4 4
18 4.250 3 4 5 3 4 5 3 4 5 3 4 5 4 5 4 5 4 5 4 5 4 5 4 5 4 5 4 4
19 4.000 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 4 4 4 4
20 3.893 3 4 5 3 4 5 3 4 5 3 4 5 3 4 5 3 4 3 4 3 4 4 4 4 4 4 4 4
21 4.107 5 4 3 2 1 5 4 3 2 5 4 5 4 5 4 5 4 5 4 5 4 5 4 5 4 5 4 5
22 3.786 2 3 4 5 2 3 4 5 3 4 5 3 4 5 3 4 5 3 4 5 3 4 3 4 4 4 4 4
23 3.857 5 4 3 2 5 4 3 4 3 4 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4
24 3.464 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 3 4 5 3 4 5 4 4 4 4 4 4 4
25 3.679 5 4 3 2 5 4 3 2 5 4 3 4 3 4 3 4 3 4 3 4 3 4 4 4 4 4 4 4
26 3.571 1 2 3 4 5 2 3 4 5 3 5 3 4 3 4 3 4 3 4 3 4 4 4 4 4 4 4 4
27 3.750 5 4 3 5 4 3 5 4 3 5 4 3 5 4 3 4 3 4 3 4 3 4 3 4 3 4 3 3

The results of a three-level experiment may only be evaluated graphically. Graphs of

casting rank versus dwell time, hot oil temperature, and vacuum time are shown as follows:

Die Cast Problem Solving Page 79

 Problem Solving Examples

The points on the graphs are the average ranking for the treatments related to the
factor levels on the x-axis. For instance, the average quality ranking of about 3.95 on
the Dwell Time vs Casting Quality graph is an average of the rankings for the treatments
that used 5 seconds for dwell time.
Three-level experiments are pooled so the effects can be calculated in the same
fashion as two-level experiments. To do this, the best and worst conditions for each
factor are determined. The best condition for dwell time, for instance, is when the dwell
time is at level 3 or 6.5 seconds. The worse condition is when dwell time is at level 2 or
5.75 seconds. The treatments where dwell time produces an intermediate result, level 1
or 5 seconds, are removed from consideration. In essence, the experiment was convert-
ed from a three-factor, three-level experiment into a three-factor, two-level experiment.
The pooled experiment in this example will appear as shown on the following page.

Page 80 Die Cast Problem Solving

 Problem Solving Examples

Pooled Experiment (To calculate significance and interactions)

Trial Dwell Time Hot Oil Temperature Vacuum Time Results
2 2 1 1 4.357
3 3 1 1 4.429
5 2 2 1 4.286
6 3 2 1 3.929
20 2 1 3 3.893
21 3 1 3 4.107
23 2 2 3 3.857
24 3 2 3 3.464

Next, the pooled experiment matrix is modified to show only two levels — level 1 and
level 2 — for the high and low levels. Therefore, the level yielding the highest quality
ranking for dwell time, originally Level 3 is renamed Level 2. The level yielding the low-
est quality for dwell time was Level 2 originally and is renamed Level 1. The level yield-
ing the lowest quality ranking for vacuum time was originally Level 1 and does not need
to be renamed. The level yielding the highest quality for vacuum time was originally
Level 3 and this is renamed Level 2. See Below.
Pooled Experiment
Trial Dwell Time Hot Oil Temperature Vacuum Time Results
2 1 1 1 4.357
3 2 1 1 4.429
5 1 2 1 4.286
6 2 2 1 3.929
20 1 1 2 3.893
21 2 1 2 4.107
23 1 2 2 3.857
24 2 2 2 3.464

Now the effects can be calculated as is usually done in a two-level experiment. From
this pooled experiment, the calculated effects and significance level are displayed below.

Effects Value Significance

Dwell Time -0.116 4
Hot Oil Temperature -0.313 2
Vacuum Time -0.42 1
Int F1/F2 -0.259 3
Int F1/F3 0.0268 7
Int F2/F3 -0.027 6
Int F1/F2/F3 -0.045 5

Die Cast Problem Solving Page 81

 Problem Solving Examples

The best and worst processes were as follows (original levels in brackets):
Dwell Time Hot Oil Temperature Vacuum Time
T3 Best 2 (3) 1 1
T24 Worst 2 (3) 2 2 (3)

Using the best process in this example three-level three-factor experiment, the cast-
ing defect level went from 20 percent to below two percent. Although the defect was not
eliminated, the cost savings attributed to this experiment were significant.

Page 82 Die Cast Problem Solving

 APPENDIX 1

Day 1 Day 2
Part Insp 1 Insp 2 Insp 3 Avg Rng Part Insp 1 Insp 2 Insp 3 Avg Rng Grd Avg
1 1
2 2
3 3
4 4
5 5
6 6
7 7
8 8
9 9
10 10
Average Average

Inspector 1 Avg Grand Mean Average

Inspector 2 Avg Grand Range Average
Inspector 3 Avg Difference Between Evaluators
Parts (n) Maximum Part Range
Trials (r) Evaluators (e)
Appendix 1
Constants
Error between evaluators (repeatability) Trials (r) K1
= Grand Range Average x K1 = 2 4.56
Error within evaluators (reproducibility) 3 3.05
=[(Diff. Bet. Eval. X K2)2 - (Repeatability2/(nxr))].5 = Eval (e) K2
Repeatability & Reproducibility (R&R) 2 3.65
=[(Repeatability) + (Reproducibility) ]
2 2 .5
= 3 2.70
Parts (n) K3
Part Variation (PV) = Maximum Part Range x K3 = 2 3.65
3 2.70
Total Variation (TV) = [(R&R) + (PV) ]
2 2 .5
= 4 2.30
5 2.08
% Repeatability = 100 x Repeatability / TV = 6 1.93
7 1.82
% Reproducibility = 100 x Reproducibility / TV = 8 1.74
9 1.67
% R & R = 100 x R&R / TV = 10 1.62

Die Cast Problem Solving Page 83

 Appendix 1

Page 84 Die Cast Problem Solving

APPENDIX 2

Is/Is Not Differences

1. A.
B.
C.
D.
2. A.
B.
C.
D.
3. A.
B.
C.
D.
4. A.
B.
C.
D.
5. A.
B.
C.
D.
6. A.
B.
C.
D.
7. A.
B.
C.
D.
8. A.
B.
C.
D.

Die Cast Problem Solving Page 85

 Appendix 2

Page 86 Die Cast Problem Solving

APPENDIX 3

Is/Is Not Theories

T1.

T2.

T3.

T4.

T5.

T6.

T7.

T8.

T9.

T10.

T11.

T12.

T13.

T14.

T15.

Die Cast Problem Solving Page 87

 Appendix 3

Page 88 Die Cast Problem Solving

NORTH AMERICAN DIE CASTING ASSOCIATION
3250 N. ARLINGTON HEIGHTS ROAD
ARLINGTON HEIGHTS, ILLINOIS 60004
P: 847.279.0001 F: 847.279.0002
NADCA@[Link]
[Link]