Page 52 - FDMAsia Jan/Feb 2026

50 MATERIALS JAN/FEB 2026 FDM ASIA | www.fdmasia.com

George Clerk www.georgeclerk.com

At lower moisture contents (4.7%–5%), the mean values Table 2. Multiple linear regression model for dielectric constant.
Predictor B Std. Error β p
ranged from 0.066 to 0.078, with maximum values reaching log10(Frequency,
0.3091. With increasing moisture content, the mean values MHz) −0.281 0.014 −0.260 <0.001
Density (kg 0.005 0.000 0.284 <0.001
and maximums of the tan δ show similar values, with a slight m−3)
Moisture
increase compared to lower moisture contents. content (%) 0.168 0.003 0.808 <0.001
Sample
At moisture contents above 15 percent, the highest maxima temperature 0.078 0.022 0.048 <0.001
were measured, reaching up to 0.4, with the highest mean (deg C)
Model statistics: R2 = 0.704; adjusted R2 = 0.704; F(4, 1664) =
values (0.15–0.21). 991.1; p < 0.001. All predictors showed VIF < 1.05.
On average, higher values of the tan δ were measured at
90 deg C but not significantly compared to 20 deg C. Increasing The model explains 70.4 percent of the variance in ε′
the temperature of the samples from 20 deg C to 90 deg C (adjusted R2 = 0.704, p < 0.001). Moisture content was the
led to an increase in ε′ from 5 to 15 percent (depending on most statistically influential parameter (β = 0.808, p < 0.001),
the sample), while in the case of tan δ this increase was at followed by density (β = 0.284, p < 0.001) and logarithmically

the level of experimental variability. transformed frequency (β = −0.260, p < 0.001).
A log10 transformation of the electric field frequency values The temperature of the samples had a statistically significant
was performed before statistical analysis, as the response effect, although smaller than the other parameters (β = 0.048,
of dielectric properties scales logarithmically with frequency. p < 0.001) No multicollinearity was detected among the
Multiple linear regression showed that all four test parameters predictors (VIF < 1.05).
(frequency, density, moisture content, sample temperature) had For the statistical analysis of tan δ, a multiple linear regression
a statistical effect on ε′, as shown in Table 2. was also conducted using the same test parameters (Table 3).