Design and Analysis

of Experiments

Chapter 6

The 2k Factorial Design

(part 2)
Dr. Tran Thanh Hung
Department of Automation Technology,
College of Engineering, Can Tho University
Email: tthung@ctu.edu.vn
 Unreplicated 2k Factorial Designs
• These are 2k factorial designs with one
observation at each corner of the “cube”
• Single replicate 2k factorial design =
unreplicated 2k factorial design
• These designs are very widely used
• Risks…if there is only one observation at
each corner, is there a chance of unusual
response observations spoiling (làm hỏng)
the results?
• Modeling “noise”?

2
 Spacing of Factor Levels in
the Unreplicated 2k Factorial
Designs

If the factors are spaced too closely, it increases the chances that
the noise will overwhelm (lấn át) the signal in the data
 increase the distance between the low (-) and high (+) levels of
the factor 3
 Unreplicated 2k Factorial Designs
• Lack of replication causes potential problems
in statistical testing
– Replication provides an estimate of “pure error” (a
better phrase is an internal estimate of error)
– With no replication, fitting the full model results in
zero degrees of freedom for error
• Potential solutions to this problem
– Using high-order interactions to estimate error
– Normal probability plotting of effects (Daniels,
1959)
– Dropping entire factors from the model and other
methods
4
 Example 6.2: an Unreplicated 2k
Design

• A 24 factorial was used to investigate the

effects of four factors on the filtration rate
(tốc độ lọc) of a resin
• The factors are A = temperature, B =
pressure, C = concentration (nồng độ) of
formaldehyde, D= stirring rate (khuấy)
• Experiment was performed in a pilot plant

5
The Resin Plant Experiment

6
 The Resin Plant Experiment

The process engineer is interested in maximizing the filtration rate.

Current process conditions give filtration rates of around 75 gal/h, using the
concentration of formaldehyde, factor C, at the high level.
The engineer would like to reduce the formaldehyde concentration as much
as possible but has been unable to do so because it always results in lower 7
filtration rates.
Estimates of the Effects

8
 The Normal/Haft Normal
Probability Plot of Effects

large effects

small effects

9
ANOVA Summary for the Model

10
Model residuals

satisfy

11
Model Interpretation – Main
Effects and Interactions

12
 Model Interpretation – Response
Surface Plots

With concentration at either the low or high level, high temperature

and high stirring rate results in high filtration rates
13
 Example 6-3 : The Drilling
Experiment

A = drill load, B = flow rate , C = rotational speed, D = type of

drilling mud
y = advance rate of the drill 14
Normal Probability Plot of Effects –
The Drilling Experiment

15
Residual Plots

Resi duals vs. P redict ed

2.5 8625

1.4 4875

R es id uals
0.3 1125

- 0.8 2625

- 1.9 6375

1.69 4. 70 7.7 0 10.7 1 1 3.71

P re d i cte d

16
 Residual Plots

• The residual plots indicate that there are

problems with the normality and equality of
variance assumptions.
• The usual approach to this problem is to use
a transformation on the response
• Power family transformations are widely
used 
y y
*

• Transformations are typically performed to

– Stabilize variance
– Get normality
– Simplify the model 17
 Selecting a Transformation

• Empirical selection of 
• Prior (theoretical) knowledge or experience can
often suggest the form of a transformation
• Analytical selection of  …the Box-Cox (1964)
method (simultaneously estimates the model
parameters and the transformation parameter
)

18
Effect Estimates Following the
Log Transformation

y  ln y
*

Three main effects are large

No indication of large
interaction effects
What happened to the
interactions?

19
ANOVA Following the Log Transformation

20
Following the Log Transformation

21
 Addition of Center Points
to a 2k Designs
• Based on the idea of replicating some
of the runs in a factorial design
• Runs at the center provide an estimate
of error and allow the experimenter to
distinguish between two possible
models: k k k
First-order model (interaction) y   0    i xi    ij xi x j  
i 1 i 1 j i
k k k k
Second-order model y   0    i xi    ij xi x j    ii xi2  
i 1 i 1 j i i 1
22
 Consider a 22 design, 1 run at each corners
nc runs at the center point (0,0)

yF : average of the four runs yF  yC  no "curvature"

at corners (-,-), (+,-), (-,+),(+,+) nF nC ( yF  yC ) 2

yC : average of the n c runs SS Pure Quad 

nF  nC
at (0,0) nF :number of
factorial design points
The hypotheses are:
k
H 0 :   ii  0
i 1
k
H1 :   ii  0
i 1
23
 Example 6.2: add center points

Consider experiment in example 6.2.

Four center points are added to this
nC  4
experiment, and at the points x1= x2 = x3=
Usually between 3 and 6
x4 = 0 the four observed filtration rates center points will work well
were 73, 75, 66, and 69.

The average of these four center points is

70.75.

The average of the 16 factorial runs is

70.06.

 no strong curvature

24
 ANOVA for Example 6.3

4
Can not reject H 0 :  ii  0
i 1
25
 Central Composite Design
(Thiết kế hỗn hợp trung tâm)
If curvature is significant, augment the design with axial runs
to create a Central Composite Design (CCD). The CCD is a
very effective design for fitting a second-order response
surface model.

26
 Practical Use of Center Points

• Use current operating conditions as the

center point
• Check for “abnormal” conditions during the
time the experiment was conducted
• Check for time trends
• Use center points as the first few runs when
there is little or no information available about
the magnitude of error
• Center points and qualitative factors?

27
Center Points and Qualitative Factors

28
 Thực hành chương 6
Bài 2: Thí nghiệm con lắc
• Vào link sau: https://myphysicslab.com/pendulum/pendulum-en.html?SHOW_CLOCK%3Dtrue%3BMASS
%3D5%3BDRIVE_FREQUENCY%3D0%3BGRAVITY%3D10%3BLIMIT_ANGLE%3Dfalse%3BRUNNING%3Dfalse%3BFIRING
%3Dtrue%3BTIME_GRAPH_LINE_2.Y_VARIABLE%3D-1%3BANGLE%3D0.8135232029158118%3BANGLE_VELOCITY%3D-
1.7125364752265197%3BTIME%3D0%3B

• Thiết kế thí nghiệm 23, 1 lần lặp, với các yếu tố sau:
Yếu tố Mức - Mức +
Length 1m 2m
Mass 5 kg 10 kg
Initial angle 0.5 rad 1 rad
• Chạy mô phỏng, đo chu kỳ dao động của con lắc (đơn vị: s)
• Dùng Minitab phân tích kết quả thí nghiệm, tìm công thức tính chu kỳ
Lưu ý:
- Nhập trực tiếp giá trị cho length, mass, bấm nút small để chọn 0.5 rad,
medium để chọn 1 rad cho initial angle. Các giá trị khác cố định.
- Nếu phân tích số dư không ổn, thử dùng bộ chuyển đổi y  y
* 2
 Nút bắt đầu/Dừng
Nút khởi động lại t
Thực hành chương 6
Đồng hồ

Chổ nhập giá trị

Chọn góc Chọn góc

bắt đầu 0.5 bắt đầu 1

• Khá khó để đo đúng 1 chu kỳ.

• Có thể đo 10 chu kỳ rồi tính ra
thời gian 1 chu kỳ.