Professional Documents
Culture Documents
3
Design of Experiments
Engineering Graduate School
Universidad de Cartagena
José Colina Márquez, PhD
Exercise No.1
A paper of the Fire Safety Journal (“The effect of the nozzle design on the stability and performance of turbulent water sprinklers”, vol. 4) describes the experiment that studied a geometric form factor for several
different designs of nozzles with six levels of flow rate at the sprinkler outlet. The main interest was focused on the potential difference between the nozzles’ design and the flow rate was considered as the disturbing factor.
The data gathered is shown as follows:
1. Does the nozzle design affect the form factor? Compare the nozzles with a dispersion diagram and an analysis of variance, using α = 0.05 for this analysis.
factor? Plot a graph of the average form factor for each type of nozzle
and compare it with an escalated
b=6 Blocks
a=5 levels
N=30 Treatments 0.00583 0.00578
0.00002 0.00048 SSError 0.05887
0.00191 0.00090
0.00672 0.00116
0.00200 0.00160
SStratamient 0.09885 0.00292
SSBloques 0.06416
Ho: μ1=μ2=μ3=μ4=μ5
H1: the mean of the treatments are different (at least one of the means is different)
Analysis of variance table for the model with an interest factor (5 levels) and a disturbing factor (6 blocks)
Variation Sum of Degrees of Mean
Fo Fc P-value
source squares freedom square
Treatments 0.09885 4 0.0247 8.3949 2.8661 0.000382
Blocks 0.06416 5 0.0128 4.3592 2.7109 0.007597
Error 0.05887 20 0.0029
Total 0.22188 29
1. Does the nozzle design affect the form factor? Compare the nozzles with a dispersion diagram and an analysis of variance, using α = 0.05 for this analysis.
Given that Fo> Fc, the null hypothesis that the variances are equal is rejected. And it is concluded that the that effectively the potential differential between the nozzle design affects the form factor. In
addition, it is concluded that since Fo of the blocks is lower than that of the treatments, it reconfirms that the variable that most affects or influences our response variable is the design of the nozzles,
with the flow rate (disturbing factor) being less incident .
However, although it is not a factor of interest for the study, the flow rate also has a significant effect (p-value <0.05) on the response variable, although not greater than thenozzles design.
1 0.78167 0.70000
2 0.85333 0.60000
Average
3 0.90167 0.50000
2. Which designs of the nozzles are different regarding to the form factor? Plot a graph of the average form factor for each type of nozzle and compare it with an escalated t-
distribution.
Grupos
Nivel Casos Media Homogéneo
s Mediante la prueba de múltiples rangos: Diferencia mínima significativa (LSD) de Fisher, se tiene que letras diferentes
1 6 0.781667 A
representan diferencias estadísticas significativas, con un nivel de confianza del 95%. En este sentido, se puede concluir que el
5 6 0.813333 A diseño de boquilla 1 es dieferente al diseño 3 y 4. Por su parte el diseño 2 es dieferente al diseño 4. Asímismo el diseño 3 es
2 6 0.853333 AB dieferente al 5 y el diseño 4 dista del 5. Lo anterior se eviencia en la siguinete tabla de contrastes
3 6 0.901667 BC
4 6 0.94 C
Contraste Sig. Diferencia +/- Límites
1 -- 2 -0.0716667 0.0834164
1 -- 3 * -0.12 0.0834164
1 -- 4 * -0.158333 0.0834164
1 -- 5 -0.0316667 0.0834164
2 -- 3 -0.0483333 0.0834164
2 -- 4 * -0.0866667 0.0834164
2 -- 5 0.04 0.0834164
3 -- 4 -0.0383333 0.0834164
3 -- 5 * 0.0883333 0.0834164
4 -- 5 * 0.126667 0.0834164
* indica una diferencia significativa.
Homework No.3
Design of Experiments
Engineering Graduate School
Universidad de Cartagena
José Colina Márquez, PhD
Exercise No.2
The yield of a chemical process was studied with a Graeco-Latin square design, by using five processing time (Latin lette
conclusions with a significance level α = 0.05.
Acid concentration
Batch 1 2 3 4 5
1 Aα=26 Bβ=16 Cγ=19 Dδ=16 Eε=13
2 Bγ=18 Cδ=21 Dε=18 Eα=11 Aβ=21
3 Cε=20 Dα=12 Eβ=16 Aγ=25 Bδ=13
4 Dβ=15 Eγ=15 Aδ=22 Bε=14 Cα=17
5 Eδ=10 Aε=24 Bα=17 Cβ=17 Dγ=14
Niveles (p)
5
N
25
7396
SS TRATA LETRAS SS TRATA
LATINAS 342.8 LETRASGRIEGAS
SSTotal:
Acid concentration
Batch 1 2 3 4 5
1 676 256 361 256 169
2 324 441 324 121 441
3 400 144 256 625 169
4 225 225 484 196 289
5 100 576 289 289 196
Sumatoria 1725 1642 1714 1487 1264
SSTotal 436
Ho: μ1=μ2=μ3=μ4=μ5
H1: the mean of the treatments are different (at least one of the means is different)
Analysis of variance table for the model with an interest factor (5 levels) and a disturbing factor (6 blocks)
Filas o renglores
10 4 2.5 0.4273504274
Columnas
24.400000000001 4 6.1 1.0427350427
Error
46.799999999999 8 5.85
Total 436 24
cessing time (Latin letters) and five catalyst concentrations (Greek letters). This study was carried out with five batches of raw mat
Acid concentration
Batch 1 2 3 4
1 26 16 19 16
2 18 21 18 11
3 20 12 16 25
4 15 15 22 14
5 10 24 17 17
89 88 92 83
Sumatoria Columnas
(Sumatoria
7921 7744 8464 6889
Columnas)^2
6889
8281
7921
7225
6724
37040
7396
12
Sumatoria
7832
SS ERROR = SSTotal-SSColumnas-SSfilas-SSTrat letras latinas-
SSError 46.8
Fc P-value
Given that the processing time factor has its Fo>
establishes that the means of the treatments are
3.8378533545559 0.000941019668464 alternative hypothesis that at least some is different
be stated that the processing time has a significa
chemical process.
3.8378533545559 0.728900141619128 Furthermore, given that the P value of the mention
concluded that it has a statistically significant effe
3.8378533545559 0.785447372185274 with a confidence level of 95.
3.8378533545559 0.442543418593072
ed out with five batches of raw material. Analyze the results and build
tion (Sumatoria
Sumatoria Filas
5 Filas)^2
13 90 8100
21 89 7921
13 86 7396
17 83 6889
14 82 6724
430 37030
78
6084 37102
SS FILAS (o reglones) 10
SS COLUMNAS 24.4
mnas-SSfilas-SSTrat letras latinas-SSTrata let grietas
sing time factor has its Fo> Fc, the null hypothesis that
means of the treatments are equal is rejected and the
hat at least some is different is accepted. In this way it can
rocessing time has a significant effect on the yield of a
chemical process.
t the P value of the mentioned factor is less than 0.05, it is
a statistically significant effect on the response variable,
with a confidence level of 95.0%.