XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 303
sample preparation. The N2 isotherm test comprised 23 rel-
ative pressure data points to examine a wide range of pore
sizes using the Brunauer-Emmett-Teller (BET) method
[27]. Figure 3 shows the steps of the experiment in a flow-
chart. It’s important to highlight that, for a more accurate
assessment of changes in biochar surface area post-mixing,
the surface area of hemp biochar was initially measured in
four replicates before mixing, employing the same method.
The obtained results were subsequently compared with the
findings of the ANOVA analysis.
RESULTS AND DISCUSSIONS
Table 2 displays the twenty combinations of the selected
factors and their ranges, along with their corresponding
measured response (surface area). Table 3 provides the fac-
tors for the reduced quadratic model and fit statistics. The
selection of the reduced quadratic model was guided by
the Akaike Information Criterion (AIC), a method ensur-
ing the model’s suitability without overfitting or underfit-
ting the data. AIC strikes a balance between the model’s
simplicity and goodness of fit, addressing both the risks of
overfitting and underfitting [28]. According to the results,
statistical parameters such as C.V. and predicted R2 was
within an acceptable range, indicating the model’s efficacy
in predicting future values. The adequate precision value
signifies sufficient power in the model, implying minimal
systematic (inherent) errors when compared to the resolu-
tion of the response.
As presented in Figure 4, which includes diagnostic
plots for response, the model’s effectiveness in predicting
future values has been affirmed.
However, the normal plot of residuals in Figure 4
(a), which is the difference between actual and predicted
values based on the regression shows promising accuracy.
The plot of actual vs predicted in Figure 4 (b) indicated
minor deviation from the model, suggesting potential
enhancements through refinement. Figure 4 (c) displays the
Externally Standardized Residuals vs. Predicted, indicating
values within acceptable limits.The Cook’s distance plot in
Figure 4 (d) shows the measure of influence, a combina-
tion of residuals and leverage, which fell within a safe range.
The Leverage vs. plot in Figure 4 (e) visually represents the
influence of each individual data point on determining
coefficients, shows that no specific point had an excessive
impact on the overall model.
Figure 3. The flowchart showing the steps of the experiment
Table 2. The twenty sets of experimental conditions, along
with their respective measured response (surface area)
Run #
Factor 1
Time,
min
Factor 2
Mass,
g
Factor 3
Intensity,%
Response
Surface Area,
m2/g
t 2 11 15 49
2 30 11 15 27
3 2 6 45 36
4 16 1 45 4
5 16 6 45 4
6 30 1 75 3
7 16 6 45 49
8 16 6 45 37
9 30 11 75 36
10 16 6 75 42
11 2 11 75 45
12 30 6 45 49
13 2 1 75 10
14 16 11 45 9
15 16 6 45 9
16 16 6 15 9
17 30 1 15 4
18 2 1 15 3
19 16 6 45 8
20 16 6 45 16
sample preparation. The N2 isotherm test comprised 23 rel-
ative pressure data points to examine a wide range of pore
sizes using the Brunauer-Emmett-Teller (BET) method
[27]. Figure 3 shows the steps of the experiment in a flow-
chart. It’s important to highlight that, for a more accurate
assessment of changes in biochar surface area post-mixing,
the surface area of hemp biochar was initially measured in
four replicates before mixing, employing the same method.
The obtained results were subsequently compared with the
findings of the ANOVA analysis.
RESULTS AND DISCUSSIONS
Table 2 displays the twenty combinations of the selected
factors and their ranges, along with their corresponding
measured response (surface area). Table 3 provides the fac-
tors for the reduced quadratic model and fit statistics. The
selection of the reduced quadratic model was guided by
the Akaike Information Criterion (AIC), a method ensur-
ing the model’s suitability without overfitting or underfit-
ting the data. AIC strikes a balance between the model’s
simplicity and goodness of fit, addressing both the risks of
overfitting and underfitting [28]. According to the results,
statistical parameters such as C.V. and predicted R2 was
within an acceptable range, indicating the model’s efficacy
in predicting future values. The adequate precision value
signifies sufficient power in the model, implying minimal
systematic (inherent) errors when compared to the resolu-
tion of the response.
As presented in Figure 4, which includes diagnostic
plots for response, the model’s effectiveness in predicting
future values has been affirmed.
However, the normal plot of residuals in Figure 4
(a), which is the difference between actual and predicted
values based on the regression shows promising accuracy.
The plot of actual vs predicted in Figure 4 (b) indicated
minor deviation from the model, suggesting potential
enhancements through refinement. Figure 4 (c) displays the
Externally Standardized Residuals vs. Predicted, indicating
values within acceptable limits.The Cook’s distance plot in
Figure 4 (d) shows the measure of influence, a combina-
tion of residuals and leverage, which fell within a safe range.
The Leverage vs. plot in Figure 4 (e) visually represents the
influence of each individual data point on determining
coefficients, shows that no specific point had an excessive
impact on the overall model.
Figure 3. The flowchart showing the steps of the experiment
Table 2. The twenty sets of experimental conditions, along
with their respective measured response (surface area)
Run #
Factor 1
Time,
min
Factor 2
Mass,
g
Factor 3
Intensity,%
Response
Surface Area,
m2/g
t 2 11 15 49
2 30 11 15 27
3 2 6 45 36
4 16 1 45 4
5 16 6 45 4
6 30 1 75 3
7 16 6 45 49
8 16 6 45 37
9 30 11 75 36
10 16 6 75 42
11 2 11 75 45
12 30 6 45 49
13 2 1 75 10
14 16 11 45 9
15 16 6 45 9
16 16 6 15 9
17 30 1 15 4
18 2 1 15 3
19 16 6 45 8
20 16 6 45 16