ENG 360

Applied Statistics for Engineers

TERM PROJECT

Topic: Optimization of Biodiesel Production at Low Temperature

Student Name(Class ID): Olcay Kaan Sert (16)

Tunahan Dilaver Derebaşı (27)
Tayyip Emre Kahriman (23)
Box-Behnken Design
Design Summary
Factors: 3 Replicates: 1

Base runs: 17 Total runs: 17

Base blocks: 1 Total blocks: 1

Center points: 5

𝟏𝒔𝒕 Data sheet:

Response Surface Regression: Biodiesel Yield(%) Experimental

versus Reaction Time; Catalyst Amount; Methanol/Oil Ratio
Coded Coefficients
Term Coef SE Coef T-Value P-Value VIF
Constant 15,836 0,434 36,52 0,000
Reaction Time -0,319 0,343 -0,93 0,383 1,00
Catalyst Amount -0,119 0,343 -0,35 0,739 1,00
Methanol/Oil Ratio 13,035 0,343 38,02 0,000 1,00
Reaction Time*Reaction Time 0,258 0,473 0,55 0,602 1,01
Catalyst Amount*Catalyst Amount 0,553 0,473 1,17 0,280 1,01
Methanol/Oil Ratio*Methanol/Oil Ratio 0,381 0,473 0,81 0,447 1,01

Reaction Time*Catalyst Amount 0,108 0,485 0,22 0,831 1,00

Reaction Time*Methanol/Oil Ratio 0,665 0,485 1,37 0,213 1,00
Catalyst Amount*Methanol/Oil Ratio 2,080 0,485 4,29 0,004 1,00

Some of the parameters given and calculated according to the table have a high P-value. And
this prevents us from reaching the optimum value.
Model Summary
S R-sq R-sq(adj) R-sq(pred)
0,969665 99,53% 98,92% 93,36%

As the R2 value approaches 100%, we can say that the relationship is very strong, in which
case the accuracy rate of the predictions to be performed on the model will also perform well.

2
Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Model 9 1381,74 153,53 163,28 0,000
Linear 3 1360,22 453,41 482,22 0,000
Reaction Time 1 0,81 0,81 0,86 0,383
Catalyst Amount 1 0,11 0,11 0,12 0,739
Methanol/Oil Ratio 1 1359,29 1359,29 1445,67 0,000
Square 3 2,41 0,80 0,85 0,508
Reaction Time*Reaction Time 1 0,28 0,28 0,30 0,602
Catalyst Amount*Catalyst Amount 1 1,29 1,29 1,37 0,280
Methanol/Oil Ratio*Methanol/Oil Ratio 1 0,61 0,61 0,65 0,447

2-Way Interaction 3 19,12 6,37 6,78 0,018

Reaction Time*Catalyst Amount 1 0,05 0,05 0,05 0,831
Reaction Time*Methanol/Oil Ratio 1 1,77 1,77 1,88 0,213
Catalyst Amount*Methanol/Oil Ratio 1 17,31 17,31 18,41 0,004
Error 7 6,58 0,94
Lack-of-Fit 3 5,67 1,89 8,28 0,034
Pure Error 4 0,91 0,23
Total 16 1388,32

Regression Equation in Uncoded Units

Biodiesel Yield(%) Experimental = 0,67 - 0,0465 Reaction Time - 5,60 Catalyst Amount
+ 0,405 Methanol/Oil Ratio
+ 0,000072 Reaction Time*Reaction Time
+ 0,553 Catalyst Amount*Catalyst Amount
+ 0,00095 Methanol/Oil Ratio*Methanol/Oil Ratio
+ 0,00179 Reaction Time*Catalyst Amount
+ 0,000554 Reaction Time*Methanol/Oil Ratio
+ 0,1040 Catalyst Amount*Methanol/Oil Ratio

Fits and Diagnostics for Unusual Observations

Biodiesel
Yield(%)
Obs Experimental Fit Resid Std Resid
14 2,720 1,536 1,184 2,44 R

17 26,660 27,844 -1,184 -2,44 R

R Large residual

3
Fig. 1 Pareto Graph Biodiesel Yield (%).

The Pareto plots of the effects serve to visually identify the significant effects and
compare the relative magnitude of various effects.

Fig. 2 Residual analysis & residual plots for Biodiesel Yield (%).

These points on a normal probability plot usually form a straight line, that is, the residuals are
normally distributed.
In the Versus Fits plot, the fitted values of all points show a random pattern of residuals on
both sides of zero line. Each point fluctuates unevenly up and down on the horizontal axis, so
the regression model is normally distributed.
The third figure shows the histogram graph of the residues. This is a discovery tool to show
the general characteristics of the data. If one or two bars are far from the others, then these
points can be considered outliers, since the appearance of the histogram varies depending on
the number of december.
Versus Order, it shows a graph of all the residuals in the order in which the data was collected
and can be used to find non-random errors, especially time-related effects.

4
Fig. 3 Residuals vs Reaction Time

Fig. 4 Residuals Vs Catalyst Amount

5
Fig. 5 Residuals vs Methanol/Oil Ratio

Contour Plot of Biodiesel Yield( vs Catalyst Amount; Reaction Time

Figure 1 Economically comforting results can be obtained if we take as a reference the point where the
efficiency is high and the least amount of catalyst is used in the lowest time interval.

6
Surface Plot of Biodiesel Yield(%) Experimental vs Catalyst Amount;
Reaction Time

Figure 2 We want the yield of the resulting product to be high. That's why the point that is maximized for others
that we see on a 3D graph is: Reaction Time: 60 Catalyst Amount: 0 Yield: 17.2.

𝟐𝒏𝒅 Data sheet:

We have removed the terms with P values above 0.4(“A *A”, “A*B”, “C*C") because they
are insignificant and analyzed them again.
Response Surface Regression: Biodiesel Yield(%) Experimental
versus Reaction Time; Catalyst Amount; Methanol/Oil Ratio
Coded Coefficients
Term Coef SE Coef T-Value P-Value VIF
Constant 16,120 0,290 55,60 0,000
Reaction Time -0,319 0,308 -1,04 0,324 1,00
Catalyst Amount -0,119 0,308 -0,39 0,707 1,00
Methanol/Oil Ratio 13,035 0,308 42,39 0,000 1,00
Catalyst Amount*Catalyst Amount 0,589 0,423 1,39 0,194 1,00
Reaction Time*Methanol/Oil Ratio 0,665 0,435 1,53 0,157 1,00
Catalyst Amount*Methanol/Oil Ratio 2,080 0,435 4,78 0,001 1,00

Model Summary
S R-sq R-sq(adj) R-sq(pred)
0,869791 99,46% 99,13% 97,31%

7
The adjusted R2 value turned out to be higher than in the first analysis. This value contains the
number of predictors in the model to help us choose the right model. The values of R-Sq are
close to thoseof R-Sq (adj) and converge to 100%.
Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Model 6 1380,76 230,13 304,18 0,000

Linear 3 1360,22 453,41 599,32 0,000

Reaction Time 1 0,81 0,81 1,07 0,324

Catalyst Amount 1 0,11 0,11 0,15 0,707

Methanol/Oil Ratio 1 1359,29 1359,29 1796,73 0,000

Square 1 1,47 1,47 1,94 0,194

Catalyst Amount*Catalyst Amount 1 1,47 1,47 1,94 0,194

2-Way Interaction 2 19,07 9,54 12,61 0,002

Reaction Time*Methanol/Oil Ratio 1 1,77 1,77 2,34 0,157

Catalyst Amount*Methanol/Oil Ratio 1 17,31 17,31 22,87 0,001

Error 10 7,57 0,76

Lack-of-Fit 6 6,65 1,11 4,86 0,074

Pure Error 4 0,91 0,23

Total 16 1388,32

Regression Equation in Uncoded Units

Biodiesel Yield(%) Experimental = -1,79 - 0,0275 Reaction Time - 5,46 Catalyst Amount
+ 0,4812 Methanol/Oil Ratio
+ 0,589 Catalyst Amount*Catalyst Amount
+ 0,000554 Reaction Time*Methanol/Oil Ratio
+ 0,1040 Catalyst Amount*Methanol/Oil Ratio

Fits and Diagnostics for Unusual Observations

Biodiesel
Yield(%)
Obs Experimental Fit Resid Std Resid
14 2,720 1,475 1,245 2,34 R
17 26,660 27,782 -1,122 -2,11 R

R Large residual

8
Fig. 6 Pareto Graph Biodiesel Yield 2 (%).

Any factor that exceeded the threshold of critical value is regarded as a significant factor. According
to the critical value in the first analysis, 2. the critical value in the analysis turned out to be lower.

Fig. 7 Residual analysis & residual plots for Biodiesel Yield 2 (%).

9
Fig. 8 Residuals vs Reaction Time 2

Fig. 9 Residuals vs Catalyst Amount

10
Fig. 10 Residuals vs Methanol/Oil Ratio

Contour Plot of Biodiesel Yield( vs Catalyst Amount; Reaction Time

Figure 3 It has the same values as the first one. It is only because we extract some values that the dense parts
seem to shift in different directions.

11
Surface Plot of Biodiesel Yield(%) Experimental vs Catalyst Amount;
Re

Figure 4 While the 3D graph drawn for the first test showed a smooth decrease, this graph first drew a
decreasing and then an increasing image.

12