Page 53 - FDMAsia Jan/Feb 2026

MATERIALS 51
www.fdmasia.com | FDM ASIA JAN/FEB 2026

Table 3. Multiple linear regression model for loss tangent.
Predictor B Std. Error β p John Douglas
log10(Frequency,
MHz) 0.061 0.002 0.562 <0.001
Density (kg 0.000 0.000 0.094 <0.001
m−3)
Moisture 0.007 0.000 0.352 <0.001
content (%)
Sample
temperature 0.009 0.003 0.056 0.002
(deg C)
Model statistics: R2 = 0.452; adjusted R2 = 0.450; F(4, 1664) =
342.9; p < 0.001. All predictors showed VIF < 1.05.

The model accounts for 45.0 percent of the total variance appears. In contrast, the values of the tan δ increase.
in tan δ (adjusted R2 = 0.450, p < 0.001). Logarithmically This phenomenon indicates greater heat release of the
transformed frequency was the dominant predictor (β = 0.562, p lamellae at higher electric field frequencies tested. However, for
< 0.001), followed by moisture content (β = 0.352, p < 0.001). lamella samples tested at 100 percent RH, the phenomenon
Density (β = 0.094, p < 0.001) and sample temperature (β = differs.
0.056, p = 0.002) were also statistically significant, but with a For these samples, the highest tan δ values were measured

smaller effect. However, multicollinearity was detected among at frequencies below one MHz. This can indicate that maximum
predictors (VIF < 1.05). release of heat of the oak lamellae under high humidity occurs
at frequencies below one MHz.
Frequency-Dependent Dielectric Behaviour & Heat This phenomenon at 100 percent relative humidity results
Release from the saturation of the wood’s porous structure at high
It is evident that each selected parameter has a certain humidity, leading to low-frequency dispersion caused by interfacial
statistical influence on the dielectric properties. The coefficient polarisation at the boundaries between the conductive aqueous

of determination was lower for the loss tangent model. This phases and the insulating cell wall matrices.
may be due to the moisture gradient within the lamellas and When the frequency increases above one MHz, the tan
possible inconsistencies in electrode spacing during testing. δ exhibits similar behaviour to that of other tested samples,
The influence of frequency on the ε′ and tan δ was but its value does not reach the maximum measured below
anticipated. Increasing the frequency caused a decrease in one MHz. Similar behaviour of the tan δ has been reported
the ε′, which was particularly pronounced up to one MHz. in previous studies.
The loss tangent at lower frequencies varied depending on Such a phenomenon occurs because at higher frequencies
the RH, with a clear trend of linear growth observed when polar water molecules inside wood cannot respond effectively, due
the frequency exceeded five MHz. to the limited capabilities of dipolar and interphase polarisation

Therefore, it can be assumed that increasing the frequency mechanisms and the rapid changes in the electric field.
of the electric field surrounding the oak wood results in greater At lower frequencies, water molecules can follow the
heat release. These results are consistent with previous research slower changes in the electric field, resulting in higher values
done on hardwood like poplar and beech. of ε′. As the frequency increases, the ε′, decreases and the
When the frequency increases above one MHz up to 25 tan δ gradually increases, due to the previously mentioned
MHz, there is an initial sharp exponential decrease in the mechanisms, leading to greater heat release.
ε′, after which ε′ stabilises and a very slight linear decrease On average, higher values were recorded for lamella samples