**Comparative Analysis of Optical Radiation Emitted by Electric Arc Generated at AC and DC Voltage**

#### **Łukasz Nagi \* , Michał Kozioł and Jarosław Zygarlicki**

Faculty of Electrical Engineering, Automatic Control and Informatics, Opole University of Technology, Proszkowska 76, 45-758 Opole, Poland; m.koziol@po.edu.pl (M.K.); j.zygarlicki@po.edu.pl (J.Z.)

**\*** Correspondence: l.nagi@po.edu.pl

Received: 7 June 2020; Accepted: 25 September 2020; Published: 2 October 2020

**Abstract:** The article presents a comparison of the spectra of electromagnetic radiation emitted by an electric arc. The spectrum ranges from ultraviolet through visible light to near infrared. Spectra from electric arcs were compared for different frequencies of generating current and for direct current. Characteristic peaks for each measurement were described, and the percentage of individual components of light emitted through the arc was presented. An electric arc is an undesirable phenomenon in many areas, and its detection and control depends largely on its source. There are also areas where an electric arc is used. A better understanding of the physical phenomena involved in different arcs can help optimize the use of the electric arc. Safety and economy through the elimination of parasitic energy shares i.e., in the welding arc can be based on the control of the arc by controlling its optical spectrum. The optical method used in this study is one of the methods of electrical discharge detection in electrical devices and systems.

**Keywords:** electric arc; gas insulation; arc welding; optical method; spectrophotometer; electromagnetic radiation; arc lamps

#### **1. Introduction**

The main aim of the research presented in this article was to determine and analyze the spectrum of optical radiation emitted by an electric arc in air for DC and different AC voltage frequencies. The shape of the spectrum and range of the radiation correspond with the applied voltage. Electric arc is an undesirable phenomenon in many areas. Uncontrolled electrical discharges from Corona Discharges (CD) to Partial Discharges (PD), breakdowns to the arc, or rapid changes in conditions allow the insulation medium to ionize and create a plasma channel for the arc. This is particularly dangerous for electrical equipment or the aviation industry. The aerospace industry is developing toward optimizing fuel consumption for this purpose by reducing the weight of aircraft. The transition from pneumatic or hydraulic systems to electrical systems is associated with a number of electrical failure problems. The next generation of aircraft that are to be fully electric (more electric aircrafts, MEA) will probably have to deal with the problems of choosing insulating materials for cables or protecting equipment, modules, and electric conductors against damage associated with electrical discharges and the detection of related damage and electrical phenomena. In aeronautics, as in other industries, the early detection of electrical discharges improves safety. In the paper [1], the authors presented a method of CD detection with a cheap CMOS (Complementary Metal-Oxide-Semiconductor) camera recording the spectrum of electromagnetic radiation in the UV (Ultraviolet) and VIS (Visual light) ranges. For measurements of this range of radiation, devices such as spectrophotometers with a larger range of radiation recording can also be used: not only the wider UV range, but also the near infrared (NIR) light frequencies [2]. This is a well-known method for the detection and measurement of partial discharges in power devices [3]. The methods of apparent charge measurement, acoustic emission method [4],

X-ray recording in Gas Insulated Switchgear (GIS) [5], or measurement of UHF signals [6] are also used. However, in aircraft conditions at reduced pressure, the methods based on electromagnetic radiation detection—UHF or Optical methods—have the best efficiency. The atmospheric conditions are important in such measurements, as well as the arrangement of elements at the test site, which may constitute obstacles to the signals [7]. The electric arc, which is a consequence of CD–PD breakdown, causes not only damages in electrical equipment, but also damages to the insulation, which increases the probability of occurrence of more such events and, consequently, failures. Along with the increase of voltages, which is required with the increased consumption of electric energy, the electric stresses in the insulations of wires also increase. The risk of electric arcs also increases. A comprehensive review of the knowledge and needs concerning arc tracking is presented in the article [8]. In the paper [9], the authors analyzed the electric arc spectrum for alternating current (AC). In the earlier mentioned article [1], the research on CD spectra were related to direct current (DC). This article presents research on electric arc spectra coming from both DC and AC for selected frequencies of current generating an arc in the air under normal pressure.

There are also areas in industry where the arc is used in many processes. Generated in appropriate conditions and controlled by the selection of many parameters, the electric arc is a phenomenon without which many industries could not work. One of them is welding. The electric arc connects the workpiece with a suitable admixture and produces a weld. Transformer (classic) welding machines usually supply the electrodes (so-called lagging) with alternate voltage–network voltage. They operate at a low frequency, which is unfavorable, because the arc near the sinusoidal transition is extinguished by 0V, and as a result, an uneven weld is obtained on the welded material, and the material is easily melted [10]. Recent methods of current inversion technology and the use of microprocessor controller technologies have resulted in new alternatives for old-fashioned methods: manual metal arc (Manual Metal Arc, MMA) welding and TIG (Tungsten Inert Gas) welding techniques [11]. The TIG method is associated with a non-fusible wolfram electrode and a negative voltage power supply. In MMA-type inverter welders, there is a lagging electrode usually supplied with positive voltage. The use of constant voltage improves the quality of the weld and enables a more accurate control of the arc, using electronic circuits that precisely regulate the voltage, current, and circuit induction of the electrode. In TIG welding, it is recommended to limit the energy released during the arc fracture, as the increased arc discharge energy is accompanied by the erosion of the non-flammable electrode [12]. In all types of welders, a significant part of energy is radiated to the environment in the form of electromagnetic radiation as a result of uncontrolled stimulation of the arc shielding gas atoms. It is important for these processes to limit this radiation by controlling the arc supply. This will limit the impact of this radiation on the health of workers and reduce losses in a broad sense from the economic point of view [13]. The second of the hypotheses put forward in these studies is the assumption that it is possible to increase the efficiency of welding equipment by modifying the parameters of the current supplying the electric arc. The supply of a high-frequency voltage to the arc allows for more precise control of the arc formation process and the reduction of excitations of the shielding gas atoms, thus reducing the parasitic emission of electromagnetic waves. Moreover, the differentiation of electric arcs based on spectra can be applied in arc furnaces. Diagnosis of the insulation of such devices is based on PD detection. The optical method is one of the diagnostic methods. Efficient differentiation of whether the signal comes from a generated arc or from a damaged insulation is important to ensure operational safety. This problem was partially addressed in [14].

Recording of the spectrum of electromagnetic radiation generated by electric arcs has so far focused on DC arcs and was performed by means of UV cameras, visible light cameras, and spectrophotometers. Arcs with currents from 50 to 200 A were studied [15] as well as arcs with current peak levels from 10 to 100 kA with short peak duration: about 15 µs [16]. The first of them presents the results recorded for the whole process related to the arc formation and its development until its expiry. In the article by Martins et al. [16], the camera was synchronized with the arc generator trigger. The development of this research is presented in the article [17]. The authors investigated arcs generated by the current

with peak values from 100 to 250 kA. In the study, a high-speed camera from the visible light range was used. The use of the spectrophotometer extends the spectrum of electromagnetic radiation by UV and NIR. The article [18] presents the use of a system with a spectrophotometer to record the spectrum of an electric arc generated on a Yacob ladder powered by direct current. The color of the DC arc changes from blue to violet and then yellow, creating a flame as the current increases, which is all in the visible light range [19]. Spectra from outside the visible light range are presented in the literature as residual, mainly for direct current supply. There is no work that would analyze the spectra for AC and DC in such a wide range. Arc power supply for different frequencies gives surprising results and is presented in our previous article [9]. The main aim of the research presented in this article was to determine and analyze the spectrum of optical radiation emitted by an electric arc in air for DC and different AC voltage frequencies. The shape of the spectrum and range of the radiation correspond with the applied voltage.

The article is an extension of comparative tests of DC and AC electric arcs. The aim of the authors was to present generalized experimental studies on the DC electric arc, in relation to the AC electric arc, with a wide spectrum of supply voltage frequencies. The research was of laboratory and cognitive character, and the parameters of electric arc supply were not optimized for any of the applications proposed in the article.

#### **2. Materials and Methods**

In this study, two systems were used to generate an electric arc. Block diagrams of the systems are shown in Figures 1 and 2. The first system (Figure 1) generated a high-frequency electric arc of a given frequency. The second system (Figure 2) was prepared to generate an arc with a DC-powered spark gap.

**Figure 1.** Block diagram of a system to generate an electric arc at a given frequency.

The circuit from Figure 1 consisted of a functional generator (Tektronix AFG1022), whose signal output was connected to the input of a power amplifier using an integrated amplifier on the Texas Instruments OPA541 circuit. The output of the power amplifier supplied the primary winding of the high-voltage transformer adapted to work with high frequency by using a ferrite core. The transformer's gear ratio was 1:500. The transformer's secondary winding fed directly to the spark plug terminals located inside the shielded chamber, which also accommodates the spectrometer probe to record the electromagnetic radiation emitted by the electric arc. The optical input surface of the spectrometer was positioned parallel to the generated electric arc at a distance of 1 cm. The optical radiation emitted by the electric arc was recorded with an optical spectrophotometer type HR4000 from Ocean Optics, which allowed registration in the range from 200 to 1100 nm. The transmission of the optical signal from the emission source to the spectrophotometer took place using a polymer optical fiber (POF) with the same spectral range as the spectrophotometer. The optical system was calibrated by the instrument manufacturer and included compensation for the spectral transmission characteristics of the POF optical fiber. The spectrometer data output was connected to a PC.

The system in Figure 2 was created by modifying the first system in which the output of the secondary winding of the high-voltage transformer was connected to a single-half rectifier with a ripple filter. The first end of the high-voltage transformer secondary winding was connected to the anode of a high-voltage rectifier diode (BY16 Diotec Semiconductor), whose cathode was connected to the first electrode of a 1nF high-voltage ceramic capacitor and the first spark plug. The second end of the secondary winding of the high-voltage transformer was connected to the second electrode of the 1nF high-voltage ceramic capacitor and the second spark plug.

**Figure 2.** Block diagram of a system for generating an electric arc with a DC-powered spark gap.

The sinusoidal voltage signal from the function generator was fed to the power amplifier, in which it was amplified with current and from the amplifier output fed the primary winding of the high-voltage transformer. The stimulated high-voltage transformer induced in its secondary winding a sinusoidal voltage of adjustable frequency and 4 kV amplitude. In the first circuit, the alternating high voltage directly supplied the spark gap. In the second circuit, the high alternating voltage was regulated to DC and filtered from the ripple and then fed to the spark plug inputs. An electric arc was formed in the spark gap, the light spectrum of which was recorded by a probe and optical spectrometer system. Data from the spectrometer were sent to a PC, where they were archived and further analyzed in the Matlab calculation program.

The measurements were taken in a laboratory with constant ambient conditions, using the systems described above. Arc initialization took place by shortening the electrodes to each other and then extending them over a distance of several millimeters. The measurements were made at an ambient temperature of 22 ◦C, humidity of 45%, and air pressure of 1000 hPa. The constant voltage of the spark supply from the second system was 4 kV, and the same voltage amplitude was also set for alternating voltages in the first system. During the measurements, none of the electrodes of the spark plug was earthed: the spark plug worked on floating reference potential.

For the purpose of interpretation and analysis of recorded optical signals, the following division of the spectral range was assumed: ultraviolet radiation (UV, from 200 to 380 nm), visible light (VIS, from 380 to 780 nm) and near-infrared (NIR, from 780 to 1100 nm). Each measurement cycle was preceded by a background light calibration procedure.

#### **3. Results and Discussion**

Figure 3 shows the registered spectra of optical radiation in the UV-VIS-NIR range for an electric arc generated at 4 kV AC at different frequencies. The obtained optical spectra of electric arcs for different frequencies of current presented in Figure 3 are mostly similar. The same arc release voltage, 4 kV, was used in the test. The characteristic peaks for each frequency are mainly in the ultraviolet range. However, some differences appear for very different frequency values of the arc generating current. For low frequencies, the spectrum range can be seen in the near-infrared part and partially in the visible light range (Figure 3a). For increasingly higher frequencies (Figure 3b,c), the visible light and NIR components disappear. For frequencies of 100 and 130 kHz (Figure 3d,e), this range is already marginally recorded. It can also be noted that the component in the VIS range for 380–440 nm from a clear peak at low frequencies clearly becomes a Pareto and may not be an important descriptor for this frequency range. Characteristic wavelengths for the presented graphs are collected and presented in Table 1.

**Figure 3.** *Cont*.

**Figure 3.** Examples of spectral characteristics for an electric arc generated in the air at a voltage of 4 kV AC with a frequency: 5 kHz (**a**); 30 kHz (**b**); 40 kHz (**c**); 100 kHz (**d**); 130 kHz (**e**).

Figure 4 shows the radiation spectrum in the UV-VIS-NIR range for the electric arc generated by DC. The whole range of spectral lines with characteristic wavelengths of higher intensity than the continuous spectrum can be observed. Table 1 shows the characteristic line lengths for this spectrum. Compared to the spectra obtained from arcs generated by alternating voltage, the main part of the electromagnetic wave being recorded is in the visible light range. In addition, a greater share of NIR can be observed than it was visible for arcs generated by AC. The comparison of the spectra to better illustrate the differences is shown in Figure 5.

**Figure 4.** Examples of spectral characteristics for an electric arc generated in the air at a voltage of 4 kV DC.

**Figure 5.** Comparison of spectral characteristics for an electric arc generated in the air at a voltage 4 kV of: 5 kHz AC, 130 kHz AC and DC.

Using the assumptions of quantum mechanics, we can describe optical radiation as a stream of photons, in which each elementary particle is a carrier of energy. The amount of energy that a single photon has can be determined by the following equation:

$$E = hv\tag{1}$$

where *E*—energy of a single photon (J), *h*—Planck constant 6.626 × 10−34 (J·s), υ— wave frequency (1/s). The optical radiation wave frequency is expressed by the formula:

ݒℎ = ܧ

$$
v = c\lambda \tag{2}$$

where υ—wave frequency (1/s); *c*—speed of light in vacuum 2.998 × 108 (m/s); λ—wavelength (nm).

From the number of photons recorded by the spectrophotometer for each wavelength, it is possible to estimate the relative amount of energy that this component emits (*Ew*). For this purpose, the number of photons (n) is included in Equation (1): ߣܿ = ݒ *υ λ*

$$E\_w = n!v\tag{3}$$

The sum of the energies of all components makes it possible to determine the energy for the whole analyzed spectral range (UV-VIS-NIR). Thus, the energy calculated is not the total energy of the optical radiation emitted by the electric arc. It is a relative value estimated on the basis of recorded spectral characteristics, which, for better illustration, is presented as a percentage of individual spectral ranges (Figure 6). ݒℎ݊ = ௪ܧ

When analyzing the spectral ranges for the energy emitted, it can be concluded that for an electric arc generated at alternating voltage, regardless of its frequency, ionizing radiation dominates in the ultraviolet range. Its share is about 60–70% of the emitted energy in the form of optical radiation. The remaining part is the energy of visible light photons, whose share is about 20–30% and a small share, about 2–6% of infrared radiation energy.

Slightly different energy distribution was observed for the optical radiation emitted by the electric arc generated at DC voltage. In this case, the energy of ionizing radiation is only 30% of the energy emitted as optical radiation. However, the dominant share is for the energy of visible light photons. Infrared radiation energy is very low and represents only 1% of the energy of the whole spectral range.

**Figure 6.** Percentage of optical radiation energy for the analyzed spectral ranges for an electric arc generated in the air at a voltage of 4 kV AC with a frequency 5 kHz (**a**); 130 kHz (**b**); and DC voltage (**c**).


**Table 1.** Characteristic peaks from electromagnetic spectrum from arc.

Differences in the energy shares of spectral ranges as well as in spectral characteristics obtained for the arc generated at alternating and DC voltage result, among others, from physicochemical phenomena occurring at the atomic level. The high frequency of alternating voltage significantly reduces the transmission of energy needed to stimulate gas atoms (in this case, it is air). It causes a rapid change in polarization, which causes an increase in electron acceleration and at the same time slowing down the electrons. This has a significant impact on the kinetic energy transmitted by them, which, as a result of the rapid polarization, cannot reach a sufficiently high value to interact with air particles. Only part of the electrons reaches a sufficiently high level of energy at which, as a result of the collision, ionizing radiation is emitted and registered in the ultraviolet range. This is clearly visible on alternating voltage spectra, where a small number of band spectra, characteristic for the activity of high-energy particles, are observed (Figure 3).

In the case of the arc generated at DC voltage, the transmitted energy is supplied continuously. This maintains a constant and stable electric field in which moving electrons increase their kinetic energy. When the electron collides with gas particles, it transfers most of its energy to their excitation, which can be observed in the form of numerous band spectra occurring in the spectral characteristics (Figure 4).

#### **4. Conclusions**

On the basis of the results obtained, significant differences in the recorded spectra of optical radiation emitted by the electric arc generated at alternating and direct voltage were shown. The percentage share of optical radiation energy was estimated for particular spectral ranges. For example, the results obtained may be useful in the design of arc-protective systems. They can also be a starting point for designing electronic and power-electronic systems of arc generation with given physical and

physicochemical properties. In turn, this can translate into new designs, prototypes, and constructions of different devices. The generalized studies presented in the article may be a guide for scientists dealing with welding arcs, arc lamps, or chemical catalyst systems, improving their energy efficiency and functional parameters. Research on the physical properties of the electric arc, correlated with the frequency parameters of the voltage generating the electric arc, can also be applied to the design of electric actuators, their electrical bundles, and power supply systems, which are used in challenging environmental conditions conducive to the formation of an undesirable electric arc, e.g., at reduced atmospheric pressure in aircraft.

**Author Contributions:** Conceptualization, Ł.N., M.K. and J.Z.; methodology, M.K. and J.Z.; software, M.K. and J.Z.; validation, J.Z. and Ł.N.; formal analysis, Ł.N. and M.K.; investigation, J.Z.; resources, Ł.N., M.K. and J.Z.; data curation, Ł.N., M.K. and J.Z.; writing—original draft preparation, Ł.N., M.K. and J.Z.; visualization, J.Z. and M.K.; supervision, Ł.N.; project administration, Ł.N.; funding acquisition, Ł.N and M.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was co-financed by funds of the National Science Centre Poland (NCS) as part of the PRELUDIUM research project No. 2014/15/N/ST8/03680 and the PRELUDIUM Research Project No. 2017/25/N/ST8/00590.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Ageing Tests of Samples of Glass-Epoxy Core Rods in Composite Insulators Subjected to High Direct Current (DC) Voltage in a Thermal Chamber**

**Krzysztof Wieczorek 1,\* , Przemysław Ranachowski <sup>2</sup> , Zbigniew Ranachowski <sup>2</sup> and Piotr Papli ´nski <sup>3</sup>**


Received: 1 November 2020; Accepted: 16 December 2020; Published: 20 December 2020 -

**Abstract:** In this article, we presented the results of the tests performed on three sets of samples of glass-reinforced epoxy (GRE) core rods used in alternating current (AC) composite insulators with silicone rubber housing. The objective of this examination was to test the aging resistance of the rod material when exposed to direct current (DC) high voltage. We hypothesized that the long-term effects of the electrostatic field on the GRE core rod material would lead to a gradual degradation of its mechanical properties caused by ionic current flow. Further, we hypothesized that reducing the mechanical strength of the GRE core rod would lead to the breakage of the insulator. The first group of samples was used for reference. The samples from the second group were subjected to a temperature of about 50 ◦C for 6000 h. The third group of samples were aged by temperature and DC high voltage for the same time. The samples were examined using the 3-point bending test, micro-hardness measurement and microscopic analysis. No recordable degradation effects were found. Long-term temperature impact and, above all, the combined action of temperature and DC high voltage did not reduce the mechanical parameters or change the microstructure of the GRE material.

**Keywords:** DC high voltage; composite insulator; glass-reinforced epoxy core; 3-point bending test; mechanical strength; micro-hardness

#### **1. Introduction**

Progress in the construction of high voltage power converter systems and the dynamic development of electricity generation systems from so-called renewable sources have resulted in an increasing interest in the transmission of electricity through high voltage direct current (HVDC) transmission lines [1,2]. The high cost of converter stations—from alternating current (AC) to direct current (DC) and vice versa—are compensated by a significant reduction in loss of energy transmitted over long distances via HVDC transmission lines, especially when compared to systems that operate at AC voltages and have lower construction costs [3].

Current high voltage lines are more often equipped with modern composite insulators. The main advantage of this is surface hydrophobicity. In polluted environments, this property makes it impossible to create water paths that conduct leakage currents for housing composite insulators. Silicone elastomer insulator housings have the ability to hydrophobize surface pollution and regenerate temporarily in lost surface properties. Compared to ceramic and glass insulators, they are significantly lighter and, in many countries, cheaper. However, the use of HVDC for the transmission of electricity unfortunately raises some technical problems. The constant electric field, forced by the HVDC line, has a significant impact on the integrity of dielectric materials usually produced for AC applications [4]. Compared to systems operating at alternating voltages, electrostatic phenomena may cause changes in degradation processes, electrical strength [5,6] and even a four-fold increase in the accumulation of surface soiling [7]. Under such conditions, when the leakage current flows through the polluted surface without passing through the zero curve of voltage and current, the ignition of non-extinguishing concentrated surface discharges may occur. This can lead to degradation of the composite housing. The ageing process in the presence of HVDC results in an increased accumulation of spatial charge in areas where material is more degraded, thus contributing to a stronger distortion of the electric field [8]. This phenomenon may cause intensification of partial discharges. In addition, if the glass-reinforced epoxy (GRE) core material is exposed to the long-term electrostatic field, then the ionic current flow may cause gradual degradation of mechanical properties [9]. This process could be activated at increased temperatures. Electrolysis of the carrying material (glass fibers) could then lead to the breakage of such insulators and, consequently, to serious failures. forced by the HVDC line, has a significant impact on the integrity of dielectric materials usually produced for AC applications [4]. Compared to systems operating at alternating voltages, electrostatic phenomena may cause changes in degradation processes, electrical strength [5,6] and even a four-fold increase in the accumulation of surface soiling [7]. Under such conditions, when the leakage current flows through the polluted surface without passing through the zero curve of voltage and current, the ignition of non-extinguishing concentrated surface discharges may occur. This can lead to degradation of the composite housing. The ageing process in the presence of HVDC results in an increased accumulation of spatial charge in areas where material is more degraded, thus contributing to a stronger distortion of the electric field [8]. This phenomenon may cause intensification of partial discharges. In addition, if the glass-reinforced epoxy (GRE) core material is exposed to the long-term electrostatic field, then the ionic current flow may cause gradual degradation of mechanical properties [9]. This process could be activated at increased temperatures. Electrolysis of the carrying material (glass fibers) could then lead to the breakage of such insulators and, consequently, to serious failures. This experiment aimed to test the mechanical strength of GRE core rod samples after 6000 h of

*Energies* **2020**, *13*, x FOR PEER REVIEW 2 of 14

significantly lighter and, in many countries, cheaper. However, the use of HVDC for the transmission of electricity unfortunately raises some technical problems. The constant electric field,

This experiment aimed to test the mechanical strength of GRE core rod samples after 6000 h of aging at a temperature of about 50 ◦C and in the presence of HVDC. aging at a temperature of about 50 °C and in the presence of HVDC.

#### **2. Materials and Methods 2. Materials and Methods**  In the ageing comparative tests, we used samples of GRE material cut from the carrying rod of a

In the ageing comparative tests, we used samples of GRE material cut from the carrying rod of a typical high voltage composite insulator for AC lines. Fibers in the rod were made of ECR-glass (ECR—type of glass electrical chemical reinforced) [10,11]. Apart from silicon (SiO2,), these types of fibers contain calcium from CaCO<sup>3</sup> (as flux and stabilizer) and aluminum from Al20<sup>3</sup> (to improve chemical resistance), which are used in the glass composition. Typical ECR-glass contains over 58% SiO2, about 22% CaO and less than 12% Al203, as well as smaller amounts of other additives. The presence of mobile sodium cations (Na+) was excluded. typical high voltage composite insulator for AC lines. Fibers in the rod were made of ECR-glass (ECR – type of glass electrical chemical reinforced) [10,11]. Apart from silicon (SiO2,), these types of fibers contain calcium from CaCO3 (as flux and stabilizer) and aluminum from Al203 (to improve chemical resistance), which are used in the glass composition. Typical ECR-glass contains over 58% SiO2, about 22% CaO and less than 12% Al203, as well as smaller amounts of other additives. The presence of mobile sodium cations (Na<sup>+</sup> ) was excluded. The samples had a diameter of 24.0 mm and a length of 120.0 mm. The first group, marked with

The samples had a diameter of 24.0 mm and a length of 120.0 mm. The first group, marked with the letter A, were fresh reference samples and reflected the material's initial structure. The samples of the second series, marked with the letter B, were subjected to a temperature of about 50 ◦C ± 2 ◦C for 6000 h. This temperature was selected following the measurements of the insulator housing temperature made on a sunny, cloudless day with an ambient temperature of about 27 ◦C. Measured temperature values under actual insulator operation conditions reached about 46 ◦C, as shown in Figure 1. the letter A, were fresh reference samples and reflected the material's initial structure. The samples of the second series, marked with the letter B, were subjected to a temperature of about 50 °C ± 2 °C for 6000 h. This temperature was selected following the measurements of the insulator housing temperature made on a sunny, cloudless day with an ambient temperature of about 27 °C. Measured temperature values under actual insulator operation conditions reached about 46 °C, as shown in Figure 1.

**Figure 1.** The surface temperature of the composite insulator housing measured with a thermal imaging camera. **Figure 1.** The surface temperature of the composite insulator housing measured with a thermal imaging camera.

The samples of the third series, marked with the letter C, with electrodes applied to their front surfaces, were subjected for 6000 h to a temperature of about 50 °C ± 2 °C and a DC voltage of 20 kV. The average value of the voltage distribution along the main axis of the typical insulator was about 1 kV/cm. In this research, it was applied twice as high as the electric field strength, i.e., 2 kV/cm. The samples of the third series, marked with the letter C, with electrodes applied to their front surfaces, were subjected for 6000 h to a temperature of about 50 ◦C ± 2 ◦C and a DC voltage of 20 kV. The average value of the voltage distribution along the main axis of the typical insulator was about 1 kV/cm. In this research, it was applied twice as high as the electric field strength, i.e., 2 kV/cm. Increasing the voltage was supposed to accelerate the aging process. Figure 2 shows one sample from each of the three series.

from each of the three series.

from each of the three series.

*Energies* **2020**, *13*, x FOR PEER REVIEW 3 of 14

*Energies* **2020**, *13*, x FOR PEER REVIEW 3 of 14

**Figure 2.** Glass-reinforced epoxy (GRE) material samples of the high voltage alternating current (HVAC) composite insulator carrying rod. From the left: reference sample—group A; thermally aged sample—group B; and DC and thermally aged sample—group C, with visible electrodes attached to the sample's front surfaces. **Figure 2.** Glass-reinforced epoxy (GRE) material samples of the high voltage alternating current (HVAC) composite insulator carrying rod. From the left: reference sample—group A; thermally aged sample—group B; and DC and thermally aged sample—group C, with visible electrodes attached to the sample's front surfaces. **Figure 2.** Glass-reinforced epoxy (GRE) material samples of the high voltage alternating current (HVAC) composite insulator carrying rod. From the left: reference sample—group A; thermally aged sample—group B; and DC and thermally aged sample—group C, with visible electrodes attached to the sample's front surfaces.

The samples were arranged in a special stand and placed in a heating chamber. Figure 3 shows the samples in the heating chamber (photograph taken with a fluke thermal imaging camera). The samples were arranged in a special stand and placed in a heating chamber. Figure 3 shows the samples in the heating chamber (photograph taken with a fluke thermal imaging camera). The samples were arranged in a special stand and placed in a heating chamber. Figure 3 shows the samples in the heating chamber (photograph taken with a fluke thermal imaging camera).

**Figure 3.** Samples placed in the heating chamber. **Figure 3.** Samples placed in the heating chamber. **Figure 3.** Samples placed in the heating chamber.

Samples from all three groups (both reference samples, thermally aged samples and those thermally and voltage aged) were tested using the 3-point bending test. For this purpose, a testing machine (INSTRON 1343) was used, which was extended with controllers and software by MTS. (MTS–Mathematisch Technische Software-Entwicklung GmbH, Berlin, Germany). Special adaptation of the support system was necessary as the tested samples had a cylindrical shape with a diameter of 24.0 mm. Therefore, in the steel rollers on which the samples were based, semi-circular notches with a radius of 12.0 mm and a depth of 10 mm were made at 100.0 mm of spacing. This is illustrated in Figure 4. A relatively low crosshead speed of 0.1 mm/min was set. This corresponded approximately to a force increase of 20 N/s. Samples from all three groups (both reference samples, thermally aged samples and those thermally and voltage aged) were tested using the 3-point bending test. For this purpose, a testing machine (INSTRON 1343) was used, which was extended with controllers and software by MTS. (MTS–Mathematisch Technische Software-Entwicklung GmbH, Berlin, Germany). Special adaptation of the support system was necessary as the tested samples had a cylindrical shape with a diameter of 24.0 mm. Therefore, in the steel rollers on which the samples were based, semi-circular notches with a radius of 12.0 mm and a depth of 10 mm were made at 100.0 mm of spacing. This is illustrated in Figure 4. A relatively low crosshead speed of 0.1 mm/min was set. This corresponded approximately to a force increase of 20 N/s. Samples from all three groups (both reference samples, thermally aged samples and those thermally and voltage aged) were tested using the 3-point bending test. For this purpose, a testing machine (INSTRON 1343) was used, which was extended with controllers and software by MTS. (MTS–Mathematisch Technische Software-Entwicklung GmbH, Berlin, Germany). Special adaptation of the support system was necessary as the tested samples had a cylindrical shape with a diameter of 24.0 mm. Therefore, in the steel rollers on which the samples were based, semi-circular notches with a radius of 12.0 mm and a depth of 10 mm were made at 100.0 mm of spacing. This is illustrated in Figure 4. A relatively low crosshead speed of 0.1 mm/min was set. This corresponded approximately to a force increase of 20 N/s.

The measuring system recorded the force acting on the sample, which was then converted to stress, according to the relation [12]: The measuring system recorded the force acting on the sample, which was then converted to stress, according to the relation [12]: The measuring system recorded the force acting on the sample, which was then converted to stress, according to the relation [12]:

$$
\sigma\_f = \frac{8 \times F \times l}{\pi \times d^8},
\tag{1}
$$

where:

where: where: σ*f*—is the bending stress, in MPa;

*σf*—is the bending stress, in MPa; *F*—force loading the sample, recorded by the measuring system, in N; *σf*—is the bending stress, in MPa; *F*—force loading the sample, recorded by the measuring system, in N;

*l*—distance between supports in the measuring system, equal to 100 mm; *F*—force loading the sample, recorded by the measuring system, in N; *l*—distance between supports in the measuring system, equal to 100 mm; *l*—distance between supports in the measuring system, equal to 100 mm;

*d*—diameter of the cylindrical sample, equal to 24 mm. *d*—diameter of the cylindrical sample, equal to 24 mm. *d*—diameter of the cylindrical sample, equal to 24 mm.

Taking into account the fixed values occurring in relation (1): Taking into account the fixed values occurring in relation (1):

$$
\sigma\_f = 0.0184 \times F \tag{2}
$$

*Energies* **2020**, *13*, x FOR PEER REVIEW 4 of 14

**Figure 4.** The sample of a group A from HVAC composite insulator carrying rod in a mechanical system for strength testing via the 3-point bending test. The notches in bottom supporting rollers are **Figure 4.** The sample of a group A from HVAC composite insulator carrying rod in a mechanical system for strength testing via the 3-point bending test. The notches in bottom supporting rollers are visible.

In addition to the 3-point bending test, we performed a micro-hardness examination of the sample material. It constituted an important supplement to the results of the optical method of the material testing. It also made it possible to independently assess the material homogeneity and cohesion. The measurements were made using the Vickers method (with a typical micro-hardness measurer) at 1 kG load of the indenter. We used a semi-automatic mode of measuring the imprint diameter. It should be emphasized that, apart from the obtained average values, the scatter of results In addition to the 3-point bending test, we performed a micro-hardness examination of the sample material. It constituted an important supplement to the results of the optical method of the material testing. It also made it possible to independently assess the material homogeneity and cohesion. The measurements were made using the Vickers method (with a typical micro-hardness measurer) at 1 kG load of the indenter. We used a semi-automatic mode of measuring the imprint diameter. It should be emphasized that, apart from the obtained average values, the scatter of results (which proves the homogeneity of the microstructure of the material) provides important information.

#### (which proves the homogeneity of the microstructure of the material) provides important information. **3. Results**

visible.

#### **3. Results**  *3.1. Mechanical Strength Tests*

*3.1. Mechanical Strength Tests*  The mechanical characteristics of the 3-point bending test for all 13 tested samples showed a very high similarity (Figures 5–7). Small differences occurred individually for particular samples, but there were no differences in the characteristics of the samples that would be typical for groups A, B or C. For a stress that usually slightly exceeds 300 MPa (294–345 MPa for individual samples), the displacement linearly increased when stress increased. The slope of characteristics was identical for all tested samples. A slight non-linearity at the beginning of the characteristics resulted from the arrangement of the samples in the clamping system. When the stress reached about 300 MPa, the samples broke axially. Long cracks were formed that extended symmetrically from the middle of the samples and were present on the upper and lower side of the specimens, as illustrated in Figures 8 and 9. These cracks usually did not reach the ends of the samples. The formation of cracks, accompanied by a well audible crackle, was reflected by a clear fault, sometimes even two faults, on the samples' mechanical characteristics. This effect occurred for all tested samples and only the length of axial cracks differed. The formation of these cracks can be considered as a critical point. Sample stiffness was clearly reduced and, therefore, the slope of the further part of the The mechanical characteristics of the 3-point bending test for all 13 tested samples showed a very high similarity (Figures 5–7). Small differences occurred individually for particular samples, but there were no differences in the characteristics of the samples that would be typical for groups A, B or C. For a stress that usually slightly exceeds 300 MPa (294–345 MPa for individual samples), the displacement linearly increased when stress increased. The slope of characteristics was identical for all tested samples. A slight non-linearity at the beginning of the characteristics resulted from the arrangement of the samples in the clamping system. When the stress reached about 300 MPa, the samples broke axially. Long cracks were formed that extended symmetrically from the middle of the samples and were present on the upper and lower side of the specimens, as illustrated in Figures 8 and 9. These cracks usually did not reach the ends of the samples. The formation of cracks, accompanied by a well audible crackle, was reflected by a clear fault, sometimes even two faults, on the samples' mechanical characteristics. This effect occurred for all tested samples and only the length of axial cracks differed. The formation of these cracks can be considered as a critical point. Sample stiffness was clearly reduced and, therefore, the slope of the further part of the characteristics already showed some differences for individual samples. There were also deviations from the straightforward course of the characteristics. Additional faults, visible on the characteristics of individual samples, corresponded to

the formation of consecutive cracks and the enlargement of existing ones. This was largely random, hence the significant differences in the characteristics of the different shapes. However, it should be emphasized that there was no apparent link between these discrepancies and the group to which the samples belonged. of individual samples, corresponded to the formation of consecutive cracks and the enlargement of existing ones. This was largely random, hence the significant differences in the characteristics of the different shapes. However, it should be emphasized that there was no apparent link between these discrepancies and the group to which the samples belonged.

**Figure 5.** Mechanical characteristics of the 3-point bending test for reference samples–group A.

*Energies* **2020**, *13*, x FOR PEER REVIEW 5 of 14

*Energies* **2020**, *13*, x FOR PEER REVIEW 5 of 14

of individual samples, corresponded to the formation of consecutive cracks and the enlargement of existing ones. This was largely random, hence the significant differences in the characteristics of the different shapes. However, it should be emphasized that there was no apparent link between these

discrepancies and the group to which the samples belonged.

**Figure 5.** Mechanical characteristics of the 3-point bending test for reference samples–group A. **Figure 5.** Mechanical characteristics of the 3-point bending test for reference samples–group A.

**Figure 6.** Mechanical characteristics of the 3-point bending test for samples subjected to temperature—group B. **Figure 6.** Mechanical characteristics of the 3-point bending test for samples subjected to temperature—group B. *Energies* **2020**, *13*, x FOR PEER REVIEW 6 of 14

**Figure 7.** Mechanical characteristics of the 3-point bending test for samples subjected to high direct current (DC) voltage and temperature–group C. **Figure 7.** Mechanical characteristics of the 3-point bending test for samples subjected to high direct current (DC) voltage and temperature–group C.

**Figure 8.** Long axial crack on top of sample A1, with an indentation caused by a crosshead of the

**Figure 10.** Indentation at the point of operation of the crosshead of the strength testing machine, long

strength testing machine.

**Figure 9.** Long axial crack in the lower part of sample C1.

axial crack and cracks on the side surface of sample C3.

current (DC) voltage and temperature–group C.

current (DC) voltage and temperature–group C.

*Energies* **2020**, *13*, x FOR PEER REVIEW 6 of 14

*Energies* **2020**, *13*, x FOR PEER REVIEW 6 of 14

**Figure 8.** Long axial crack on top of sample A1, with an indentation caused by a crosshead of the strength testing machine. **Figure 8.** Long axial crack on top of sample A1, with an indentation caused by a crosshead of the strength testing machine. **Figure 7.** Mechanical characteristics of the 3-point bending test for samples subjected to high direct

**Figure 9.** Long axial crack in the lower part of sample C1. **Figure 9.** Long axial crack in the lower part of sample C1.

**Figure 10.** Indentation at the point of operation of the crosshead of the strength testing machine, long axial crack and cracks on the side surface of sample C3. Notwithstanding any differences in sample characteristics above the critical point (indicating the load where the breaking of the reinforcement initiates), samples exhibited high repeatability of the maximum stress value. When this value was reached, faults were produced, often with a large decrease in stress. In addition to audible cracklings, this indicates the formation of subsequent cracks that substantially reduced sample rigidity. It should also be noted that the samples did not deflect during the test. The recorded displacement of the crosshead (on the order of a few millimeters) caused the test samples to significantly indent, as illustrated in Figures 8 and 10. Additionally, at higher force values, steel components of the measuring system underwent deflection. After taking the samples out of the clamping system, they did not show the slightest bend. However, most of them had cracks on flat side surfaces, as seen in Figure 11. In Tables 1–3, the values of maximum force and stress were collected for all tested samples. Table 4 shows averaged values of maximum stress, together with standard deviation, for all three groups of samples. **Figure 8.** Long axial crack on top of sample A1, with an indentation caused by a crosshead of the strength testing machine. **Figure 9.** Long axial crack in the lower part of sample C1.

**Figure 10.** Indentation at the point of operation of the crosshead of the strength testing machine, long axial crack and cracks on the side surface of sample C3. **Figure 10.** Indentation at the point of operation of the crosshead of the strength testing machine, long axial crack and cracks on the side surface of sample C3.

*Energies* **2020**, *13*, x FOR PEER REVIEW 7 of 14

**Figure 11.** Cracks on flat side surface of sample C4. **Figure 11.** Cracks on flat side surface of sample C4.

**Table 1.** Maximum force and stress values recorded for reference samples—group A. **Table 1.** Maximum force and stress values recorded for reference samples—group A.


**Table 2.** Maximum force and stress values recorded for samples subjected to temperature—group B. **Sample Designation B1 B2 B3 B4 Table 2.** Maximum force and stress values recorded for samples subjected to temperature—group B.


**Sample Designation C1 C2 C3 C4**  Maximum force (kN) 31.14 31.44 31.33 31.28 **Table 3.** Maximum force and stress values recorded for samples subjected to high DC voltage and temperature—group C.


Average value maximum stress (MPa) 572 ± 14.1 570 ± 17.1 571 ± 9.7 **Table 4.** Average values of maximum stress, including standard deviation, for the samples of all three groups.

**Group of Samples A B C** 


#### *3.2. Microscopic Examination of Samples*

temperature—group C.

Microscopic examinations were carried out on the flat side surfaces of the samples, which were randomly selected from all three groups: A, B and C. For this purpose, fragments of the material were cut out of the selected samples. Further, we made so-called metallographic micro-sections, which were also used in micro-hardness measurements.

The metallographic micro-sections were prepared on a Struers LaboPol-2 polishing machine. The surface of the samples were grinded using SiC abrasive papers, and then polished using Struers DiaPro diamond suspension with the grain diameters of 3 µm and 1 µm. The final polishing was carried out on a colloidal SiO<sup>2</sup> suspension, with a grain size of 0.04 µm (Struers OP-S suspension). After each grinding and polishing step, the samples were washed in an ultrasonic washer in ethyl alcohol.

All recorded mechanical characteristics of the 3-point bending test were collected (Figure 12). The figures and tables illustrate the reproducibility of the results obtained from the 3-point bending test.

**Figure 12.** Mechanical characteristics of the 3-point bending test of all tested samples. **Figure 12.** Mechanical characteristics of the 3-point bending test of all tested samples.

*3.2. Microscopic Examination of Samples*  Microscopic examinations were carried out on the flat side surfaces of the samples, which were randomly selected from all three groups: A, B and C. For this purpose, fragments of the material were cut out of the selected samples. Further, we made so-called metallographic micro-sections, which were also used in micro-hardness measurements. The metallographic micro-sections were prepared on a Struers LaboPol-2 polishing machine. The surface of the samples were grinded using SiC abrasive papers, and then polished using Struers DiaPro diamond suspension with the grain diameters of 3 µm and 1 µm. The final polishing was carried out on a colloidal SiO2 suspension, with a grain size of 0.04 µm (Struers OP-S suspension). After each grinding and polishing step, the samples were washed in an ultrasonic washer in ethyl alcohol. The analysis of microscopic images concluded that the microstructure of the tested material was characterized by high homogeneity. Images from different places on the tested surface, both from the same and different samples, showed no differences. The tight arrangement of fibers and binders (in the form of epoxy resin occupying about 1/3 surface) was analogous to all of the tested observation fields. The glass fiber diameter was also the same (several micrometers) with a small size dispersion. It should be emphasized that there were no differences for individual samples from groups A, B and C. Neither the long-term subjection to a temperature of about 50 ◦C nor the combined action of temperature and DC voltage of 20 kV caused any observable effects in the material's microstructure, which had been found in earlier studies [9]. Microstructure images of the sample material from all three groups are presented in Figures 13–15. There were few small dark chippings in the microstructure elements (i.e., fiber fragments and, less often, binder). They were created during the preparation of the surface of the samples. *Energies* **2020**, *13*, x FOR PEER REVIEW 9 of 14

distinguished more clearly, allowing for more accurate measurements. Image analysis was carried out with a Clemex computer-based analyzer. The blue phase was made up of glass fibers against a dark organic phase. The binary mask that was applied to the image, as shown in Figure 15, allowed **Figure 13.** Microstructure image of sample A2 on its flat side surface, 200×. The ECR glass fibers in section and slightly darker epoxy resin are visible. Dark chipping of fiber fragments and less often of binder are few. **Figure 13.** Microstructure image of sample A2 on its flat side surface, 200×. The ECR glass fibers in section and slightly darker epoxy resin are visible. Dark chipping of fiber fragments and less often of binder are few.

for the quantitative measurement of glass fibers and resin content in the material and facilitated the determination of fiber diameters. Figure 16 shows the size distribution of glass fiber diameters obtained by averaging three microscopic images (A, B and C, one from each sample group). After appropriate reformatting of the 8-bit grey scale images and processing of these images with an optical microscope, the areas with the glass fibers and epoxy resin binder were depicted and distinguished more clearly, allowing for more accurate measurements. Image analysis was carried out with a Clemex computer-based analyzer. The blue phase was made up of glass fibers against a dark organic phase. The binary mask that was applied to the image, as shown in Figure 15, allowed for

**Figure 14.** Microstructure image of sample B2 on its flat side surface 100×. Few dark chipping of fiber

clearly multimodal, with fibers with a diameter of 13.5–16.0 mm dominating.

On all tested micro-sections of samples A2, B2 and C3, glass fibers represented, on average, 67.0%. The differences for individual observation fields did not exceed 2.1%. The average value of the glass fiber diameter was 14.9 mm. The whole distribution was between 11.0–18.5 mm. However, over 90% of the fibers had a diameter in a narrow range, from 13.0 to 16.5 mm. The distribution was

The microstructure of the tested GRE material of the HVAC composite insulator carrying rod was clearly assessed as entirely appropriate, compact and homogeneous. Chipping created during the grinding and polishing process of test surfaces (mainly small fragments of glass fibers) did not exceed 3% of the surface. Tightly arranged glass fibers constituted 2/3 of the material by volume. The areas with only the resin visible in the micro-sections were not numerous and the size did not exceed several fibers. The ECR glass fibers used were uniform. Further, their diameter distribution was

fragments and binder are visible.

quite narrow and they did not raise any objections.

the quantitative measurement of glass fibers and resin content in the material and facilitated the determination of fiber diameters. Figure 16 shows the size distribution of glass fiber diameters obtained by averaging three microscopic images (A, B and C, one from each sample group). **Figure 13.** Microstructure image of sample A2 on its flat side surface, 200×. The ECR glass fibers in section and slightly darker epoxy resin are visible. Dark chipping of fiber fragments and less often of binder are few.

*Energies* **2020**, *13*, x FOR PEER REVIEW 9 of 14

**Figure 14.** Microstructure image of sample B2 on its flat side surface 100×. Few dark chipping of fiber fragments and binder are visible. **Figure 14.** Microstructure image of sample B2 on its flat side surface 100×. Few dark chipping of fiber fragments and binder are visible. *Energies* **2020**, *13*, x FOR PEER REVIEW 10 of 14 *Energies* **2020**, *13*, x FOR PEER REVIEW 10 of 14

**Figure 15.** Microstructure image of sample C3 on its flat side surface, 200×. Few chipping of fiber fragments and binder are visible (**a**) on the right side and (**b**) the same area with a colored binary mask, which enables precise measurements to be taken. **Figure 15.** Microstructure image of sample C3 on its flat side surface, 200×. Few chipping of fiber fragments and binder are visible (**a**) on the right side and (**b**) the same area with a colored binary mask, which enables precise measurements to be taken. **Figure 15.** Microstructure image of sample C3 on its flat side surface, 200×. Few chipping of fiber fragments and binder are visible (**a**) on the right side and (**b**) the same area with a colored binary mask, which enables precise measurements to be taken.

**Figure 16.** Size distribution of glass fiber diameters in the GRE material for the HVAC composite insulator carrying rod. **Figure 16.** Size distribution of glass fiber diameters in the GRE material for the HVAC composite insulator carrying rod. **Figure 16.** Size distribution of glass fiber diameters in the GRE material for the HVAC composite insulator carrying rod.

an important supplement to the results of other tests of the material. The measurements were carried out using the Vickers method, with a Struers Dura Scan universal micro-hardness meter with 1 kG of indenter load (HV1 – Hardness Vickers method, 1 means 1 kG ). The measurements were carried out on the same side surfaces of the samples on which microscopic tests were performed. A semi-automatic mode of measuring the imprint diameter was used. If different diameters were obtained on the same imprint with generally small differences, their length was averaged (Figure 17). The following HV1(International designation of Hardness Vickers method. HV1 means a hardness of 1 kG) values were obtained and averaged over 5 measurements on each sample:

an important supplement to the results of other tests of the material. The measurements were carried out using the Vickers method, with a Struers Dura Scan universal micro-hardness meter with 1 kG of indenter load (HV1 – Hardness Vickers method, 1 means 1 kG ). The measurements were carried out on the same side surfaces of the samples on which microscopic tests were performed. A semi-automatic mode of measuring the imprint diameter was used. If different diameters were obtained on the same imprint with generally small differences, their length was averaged (Figure 17). The following HV1(International designation of Hardness Vickers method. HV1 means a hardness of 1 kG) values were obtained and averaged over 5 measurements on each sample:

*3.3. Micro-Hardness Measurements of Samples* 

*3.3. Micro-Hardness Measurements of Samples* 

Reference sample A2–165 ± 9;

Reference sample A2–165 ± 9;

Temperature-aged sample B2–169 ± 11;

Temperature-aged sample B2–169 ± 11;

Temperature- and voltage-aged sample C3–155 ± 8.

Temperature- and voltage-aged sample C3–155 ± 8.

On all tested micro-sections of samples A2, B2 and C3, glass fibers represented, on average, 67.0%. The differences for individual observation fields did not exceed 2.1%. The average value of the glass fiber diameter was 14.9 mm. The whole distribution was between 11.0–18.5 mm. However, over 90% of the fibers had a diameter in a narrow range, from 13.0 to 16.5 mm. The distribution was clearly multimodal, with fibers with a diameter of 13.5–16.0 mm dominating.

The microstructure of the tested GRE material of the HVAC composite insulator carrying rod was clearly assessed as entirely appropriate, compact and homogeneous. Chipping created during the grinding and polishing process of test surfaces (mainly small fragments of glass fibers) did not exceed 3% of the surface. Tightly arranged glass fibers constituted 2/3 of the material by volume. The areas with only the resin visible in the micro-sections were not numerous and the size did not exceed several fibers. The ECR glass fibers used were uniform. Further, their diameter distribution was quite narrow and they did not raise any objections. *Energies* **2020**, *13*, x FOR PEER REVIEW 11 of 14

#### *3.3. Micro-Hardness Measurements of Samples* The obtained micro-hardness values were high and clearly proved the quality of the tested GRE material of the HVAC composite insulator carrying rod. This confirmed the generally high

Apart from the abovementioned microscopic tests, micro-hardness measurements were carried out for all samples. They allowed to access the cohesiveness and homogeneity of the material, being an important supplement to the results of other tests of the material. The measurements were carried out using the Vickers method, with a Struers Dura Scan universal micro-hardness meter with 1 kG of indenter load (HV1 – Hardness Vickers method, 1 means 1 kG). The measurements were carried out on the same side surfaces of the samples on which microscopic tests were performed. A semi-automatic mode of measuring the imprint diameter was used. If different diameters were obtained on the same imprint with generally small differences, their length was averaged (Figure 17). The following HV1 (International designation of Hardness Vickers method. HV1 means a hardness of 1 kG) values were obtained and averaged over 5 measurements on each sample: evaluation of homogeneity and cohesiveness of the tested material. The imprints were often of a regular shape, allowing for automatic diameter measurements. In the case of less regular imprints, it was necessary to correct the marker setting manually, as illustrated in Figure 17. The typical phenomenon of relaxing the energy of load interaction through microcracks, especially running from the apex of the imprint, was not observed. However, the fibers located inside the area of the indenter operation often cracked. Images of typical imprints obtained on samples from three series are shown in Figures 17–19. Compared to the average hardness of the reference sample, the thermally aged sample showed a slightly higher hardness (2.4%). The temperature and voltage-aged sample had a reduced hardness (6.1%). The differences were small and comparable to the standard deviation values. Therefore, we


**Figure 17.** Image of an indenter imprint on the flat side surface of sample A2, 500×. In order to measure it correctly, the position of the markers was manually corrected. The measured values of the imprint diameter were averaged. There are no common microcracks running from the apex of the imprint, but there are visible fiber cracks as a result of the indenter operation. **Figure 17.** Image of an indenter imprint on the flat side surface of sample A2, 500×. In order to measure it correctly, the position of the markers was manually corrected. The measured values of the imprint diameter were averaged. There are no common microcracks running from the apex of the imprint, but there are visible fiber cracks as a result of the indenter operation.

The obtained micro-hardness values were high and clearly proved the quality of the tested GRE material of the HVAC composite insulator carrying rod. This confirmed the generally high evaluation of homogeneity and cohesiveness of the tested material. The imprints were often of a regular shape, allowing for automatic diameter measurements. In the case of less regular imprints, it was necessary to correct the marker setting manually, as illustrated in Figure 17. The typical phenomenon of relaxing the energy of load interaction through microcracks, especially running from the apex of the imprint, was not observed. However, the fibers located inside the area of the indenter operation often cracked. Images of typical imprints obtained on samples from three series are shown in Figures 17–19. *Energies* **2020**, *13*, x FOR PEER REVIEW 12 of 14 *Energies* **2020**, *13*, x FOR PEER REVIEW 12 of 14

**Figure 18.** Image of an indenter imprint on the flat side surface of sample B2, 500×. Cracks and chipping of fiber fragments and resin damage as a result of the indenter operation are clearly visible. **Figure 18.** Image of an indenter imprint on the flat side surface of sample B2, 500×. Cracks and chipping of fiber fragments and resin damage as a result of the indenter operation are clearly visible. **Figure 18.** 

**Figure 19.** Image of an indenter imprint on the flat side surface of sample C3, 500×. Cracks and chipping of fiber fragments and resin damage in the area of the indenter operation are visible. **Figure 19.** Image of an indenter imprint on the flat side surface of sample C3, 500×. Cracks and chipping of fiber fragments and resin damage in the area of the indenter operation are visible.

**4. Discussion**  Numerous publications that have focused on the use of composite insulators in high DC voltage **4. Discussion**  Compared to the average hardness of the reference sample, the thermally aged sample showed a slightly higher hardness (2.4%). The temperature and voltage-aged sample had a reduced hardness

an ionic current in the GRE core. This process may reduce the mechanical strength and, consequently, the insulator can break and the line would fall to the ground. This problem has been noticed in the case of glass disc insulators [13]. However, our research showed that the mechanical properties of the tested samples of the GRE cores did not deteriorate under the experiment's adopted conditions. Both the long-term subjection to a temperature of about 50 °C and the collective action of temperature and DC voltage of 20 kV did not cause any observable and undesirable effects in the microstructure of the material. A similar statement has been made in earlier studies [9]. The

consequently, the insulator can break and the line would fall to the ground. This problem has been

properties of the tested samples of the GRE cores did not deteriorate under the experiment's

action of temperature and DC voltage of 20 kV did not cause any observable and undesirable effects

analysis of the obtained research results is presented in the conclusion.

analysis of the obtained research results is presented in the conclusion.

(6.1%). The differences were small and comparable to the standard deviation values. Therefore, we can conclude that the material hardness of the tested samples from all three series remain at a similar level. There is no clear impact of ageing, neither as a result of temperature or the combined action of temperature and voltage on the material hardness, especially given the fact that the mechanical strength of all three series of tested samples remained at a similar level.

#### **4. Discussion**

Numerous publications that have focused on the use of composite insulators in high DC voltage lines have mainly examined their surface properties. Meanwhile, an important research issue that has not been mentioned in the literature is the long-term mechanical strength of the GRE cores exposed to high DC voltage. Long-term exposure to high DC voltage can lead to the development of an ionic current in the GRE core. This process may reduce the mechanical strength and, consequently, the insulator can break and the line would fall to the ground. This problem has been noticed in the case of glass disc insulators [13]. However, our research showed that the mechanical properties of the tested samples of the GRE cores did not deteriorate under the experiment's adopted conditions. Both the long-term subjection to a temperature of about 50 ◦C and the collective action of temperature and DC voltage of 20 kV did not cause any observable and undesirable effects in the microstructure of the material. A similar statement has been made in earlier studies [9]. The analysis of the obtained research results is presented in the conclusion.

#### **5. Conclusions**

The mechanical characteristics of the 3-point bending test for all tested samples (reference and aged) showed a high similarity. Small differences occurred individually for particular samples, but there were no differences in the characteristics of samples that would be typical for any of the sample groups—reference A, or aged B and C.

Notwithstanding the relatively small differences in sample characteristics, all tested samples showed a high repeatability of the maximum stress value. The average maximum stress values for the three sample groups (A, B and C) were almost identical.

The microstructure of the tested GRE material of the HVAC composite insulator carrying rod should be assessed as entirely appropriate, compact and homogeneous.

Tightly arranged glass fibers constituted 2/3 of the material by volume. Areas with just resin were not numerous and their size did not exceed several fibers. This is difficult to avoid in the production process. The ECR glass fibers used were uniform; their diameter distribution was quite narrow and did not raise any objections.

The obtained micro-hardness values were high and proved the quality of the tested material of the HVAC composite insulator carrying rod. A small dispersion of results (±6.5%) should be emphasized. This confirmed the overall good evaluation of homogeneity and cohesiveness of the tested material. It can be concluded that the material hardness of the tested samples from all three series (A, B and C) remained at a similar high level. Therefore, there was no clear impact of ageing neither as a result of temperature nor as the combined action of temperature and voltage on the hardness of the GRE material.

On the basis of the mechanical and microscopic tests presented above (performed on reference samples and aged GRE samples of the HVAC composite insulator carrying rod) it was found that there were no registrable degradation effects. Long-term (6000 h) interaction of temperature 50 ◦C and the combined action of temperature and DC voltage 20 kV, did not cause any decrease of GRE material mechanical parameters or change in its microstructure. Thus, the test results presented in [9] were fully confirmed.

**Author Contributions:** Conceptualization: K.W. and P.R.; methodology: K.W., P.R., Z.R., and P.P.; validation: K.W., P.R., Z.R., and P.P.; investigation: K.W. and P.R.; data curation: K.W. and P.R.; writing—original draft

preparation: K.W., P.R., Z.R., P.P.; writing—review and editing: K.W., P.R., Z.R., and P.P. All authors have read and agreed to the published version of this manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **E**ff**ect of Moisture on the Thermal Conductivity of Cellulose and Aramid Paper Impregnated with Various Dielectric Liquids**

**Grzegorz Dombek 1,\* , Zbigniew Nadolny <sup>1</sup> , Piotr Przybylek <sup>1</sup> , Radoslaw Lopatkiewicz <sup>2</sup> , Agnieszka Marcinkowska <sup>3</sup> , Lukasz Druzynski <sup>1</sup> , Tomasz Boczar <sup>4</sup> and Andrzej Tomczewski <sup>5</sup>**


Received: 6 July 2020; Accepted: 26 August 2020; Published: 27 August 2020

**Abstract:** This paper presents the effect of the impact of moisture in paper insulation used as insulation of transformer windings on its thermal conductivity. Various types of paper (cellulose and aramid) and impregnated (mineral oil, synthetic ester, and natural ester) were tested. The impact of paper and impregnated types on the changes in thermal conductivity of paper insulation caused by an increase in moisture were analyzed. A linear equation, describing the changes in thermal conductivity due to moisture, for various types of paper and impregnated, was developed. The results of measuring the thermal conductivity of paper insulation depending on the temperature are presented. The aim of the study is to develop an experimental database to better understand the heat transport inside transformers to assess aging and optimize their performance.

**Keywords:** aramid paper; cellulose; dielectric materials; insulation system; mineral oil; moisture; natural ester; synthetic ester; thermal conductivity; transformers

#### **1. Introduction**

The power system is a collection of devices for the generation, transmission, distribution, storage, and use of electricity. Enabling the delivery of electricity for households, enterprises, and public utilities in a continuous and uninterrupted manner is based primarily on the functional connection and appropriate maintenance of this strategic infrastructure. From the point of view of the transmission and distribution of electricity, power transformers play an important role. Knowledge of the condition of the transformer is necessary to achieve maximum return on investment, as well as to minimize the costs associated with its operation [1,2].

At present, the age of most power transformers working in the power system exceeds 25 years, which may result in the need to revitalize or replace them in the coming years [3,4]. The efficiency of transformers depends mainly on the state of their insulation system [5–7]. The insulation system is a type of "transformer heart". Due to the fact that it consists of solid (cellulose and aramid paper) and

liquid (electrical insulating liquid), it may degrade mainly due to thermal stresses, but also electrical, mechanical, and chemical stresses occurring in the transformer operating state [8,9]. As a consequence, it may lead to transformer failure, i.e., affect its operational safety [10,11].

The degradation of the transformer insulation system is significantly dependent on the temperature inside the transformer, and strictly speaking on the temperature of the hottest spot HS in the transformer. The temperature of the hottest place in the transformer depends on the efficiency of its cooling, i.e., on the temperature increases in its individual elements–windings, oil-paper insulation of the windings, insulating liquid filling the transformer interior, and tank and air washing the tank [12]. The higher these temperatures increase, the higher the hot spot temperature [13]. It should also be noted that the HST temperature is also one of the main factors that limit the transformer load [14,15]. Therefore, it is important to effectively cool these devices by using materials with appropriate properties [16,17]. Due to the fact that the oil-paper insulation is the closest to the windings, i.e., the places with the highest temperature, it is particularly exposed to degradation.

The moisture content [18] is one of the main parameters allowing assessment of the condition of the oil-paper insulation of the transformer. The water content in oil-paper insulation increases with the transformer lifetime. Its presence may contribute to the degradation of the transformer insulation system by accelerating the aging processes occurring in it [19,20], which increases the probability of failure. Due to the role of power transformers and distribution transformers in the power system, this is a particularly significant problem [3,21].

Moisture in transformer windings insulation can cause some problems during operation, e.g., aging and electrical breakdown between its windings. Since transformer paper insulation carries a large portion of moisture, knowledge of water content in this part is essential [1,19,22]. Therefore, appropriate numerical multiphysical approaches [23–25] e.g., with ANSYS or COMSOL, constitutive thermal (e.g., thermal conductivity) and electrical (e.g., resistivity, breakdown voltage) material parameters and models are needed [26].

The main sources of moisture in the transformer are primary moisture, tank leaks, and chemical degradation of cellulose. Primary moisture is due to the highly hygroscopic nature of cellulose. Therefore, paper insulation can be characterized by the relative humidity of up to 8% at the final stage of transformer production [27]. Therefore, transformer manufacturers use a number of treatments that allow a reduction of the moisture content of the insulation to an acceptable level below 1% [28]. Tank leaks usually result from the degradation of gaskets on the bushing, radiator flanges, pumps, piping, etc., but may also be due to holes in the heat sinks or a metal tank [29]. In the event of a leak, there is a risk of moisture being sucked into the transformer through capillary action, which may lead to increased moisture in the transformer insulation system. The degradation of cellulose fibers, resulting from the elevated operating temperature of the transformer, is accompanied by water formation-free hydrogen atoms combined with oxygen to become a source of water and contribute to the increase in moisture in the transformer insulation. Due to the fact that the presence of water intensifies the process of further cellulose degradation along with the increasing transformer exploitation time, an increase in the dynamics of moistening of its insulation is also observed [30].

The impact of moisture on the properties of the transformer insulation system is the subject of research of many scientists around the world. The increase in moisture has a number of negative consequences that are important for the transformer insulation system, but also for its other elements. First of all, reducing the resistivity of insulation materials [31] and the breakdown voltage [32,33] belong to the most important of them. As a consequence, this may lead to a decrease in the initial voltage of partial discharges [30], thus contributing to an increase in the likelihood of an unexpected transformer failure [34,35]. An increase in the moisture level of the paper insulation also contributes to the increase of dielectric losses [36,37], thus contributing to the temperature rise inside the transformer. An increase in the moisture level of the transformer insulation system also leads to an increase in cellulose depolymerization rate [19], which contributes to the weakening of its mechanical properties and, consequently, to a shorter transformer life [38,39]. An increase in moisture can cause an increase

of probability of gas bubble formation in the insulating liquid, which is known as the bubble effect [3]. It can lead to an increase in pressure in the transformer tank, which may be a reason for insulating liquid leakage to soil. This can result in the loss of insulation inside the transformer and environmental contamination. The increase in moisture from 0.3%, which corresponds to the new insulation, to 4.5%, which corresponds to aged insulation, causes a 15-fold reduction in the lifetime of transformer insulation. By contrast, according to other sources, the increase in moisture insulation from 0.1% to 1.0% accelerates the aging process of paper insulation 10-fold. According to [40], the increase in moisture from 0.5% to 5.0% accelerates this process up to 100 times.

As it results from the above considerations, temperature and moisture play an important role in the transformer—both factors interact with each other in a kind of feedback, thus directly affecting the time of failure-free operation of the transformer. The higher the transformer operating temperature and humidity, the shorter the lifetime of the transformer insulation system. The analysis of cooling conditioning properties in various insulating liquids used in the transformer, together with the factors affecting them, has already been presented in the literature [41–43]. However, no information is available on the effects of various factors, including temperature and humidity, on the cooling efficiency of the paper insulation used in the transformer. Transformer paper insulation is a solid material; therefore, the cooling efficiency will depend on its thermal conductivity [16]. Therefore, in this paper, the impact of moisture on the thermal conductivity of paper insulation (cellulose and aramid) impregnated with various insulating liquids (mineral oil, synthetic esters, natural esters) used in transformers was examined and analyzed. These studies supplement the knowledge regarding heat transport in the transformer and can be used to calculate the temperature rise in the transformer, both at the design and operation stages. The influence of moisture on the thermal conductivity of materials is a very important issue, not only in terms of the operation of insulation systems of power devices but also in many other cases, such as the energy efficiency of buildings [44,45].

#### **2. Materials and Methods**

#### *2.1. Used Materials*

Kraft cellulose paper and Nomex® 926 aramid paper (DuPont, Wilmington, Delaware, DE, USA) [46] were used for the research.

Three commonly used types of insulating liquids were used to impregnate samples of both types of paper-mineral oil Nytro Draco (Nynas, Stockholm, Sweden) [47], synthetic ester Midel 7131 (M&I Materials, Manchester, UK) [48], and natural ester FR3 (Cargil, Minneapolis, MN, USA) [49].

Table 1 presents the combinations of papers impregnated with insulating liquids prepared for testing.


**Table 1.** Combinations of papers impregnated with insulating liquids.

In order to determine the impact of the structure of both tested types of paper on the thermal conductivity of the analyzed insulation systems, unimpregnated cellulose, and aramid paper samples were also prepared.

Due to the fact that the main goal was to analyze the impact of moisture on the thermal conductivity of the above-mentioned combinations of papers impregnated with different insulating liquids, samples with different water content in paper (WCP) were prepared. In accordance with [50–52], the water content in paper is defined as a ratio of water weight and dry weight of a paper sample expressed as a percentage.

Measurements of thermal conductivity of the analyzed materials were carried out for temperatures in the range of 25 to 100 ◦C. The lower temperature range resulted, first, from the capabilities of the measuring system. Secondly, for lower temperatures, thermal aspects do not play a significant role in the operation of transformers. First, the upper temperature range also resulted from the limitations of the measuring system. Secondly, situations where the working temperature of the paper insulation exceeds 100 ◦C are rare.

#### *2.2. Preparation of Paper Samples with Various Moisture Content*

To determine the effect of moisture of fibrous materials impregnated with various insulating liquids on their thermal conductivity, it was necessary to prepare samples of cellulosic and aramid materials with different water content. The sample preparation procedure included the following stages: (I) drying of the samples, (II) moistening the samples, (III) impregnation and conditioning the samples, and (IV) measuring the water content of the samples.

Samples of fibrous materials were dried in a vacuum chamber for eight hours, at a temperature of 90 ± 5 ◦C, at a pressure of 0.2 to 0.4 mbar. Such drying conditions enabled the water content of the samples to be reduced below 0.2%. Then, to obtain an appropriate level of water content, samples of materials were moistened in a controlled manner by placing them in a climate chamber forcing an appropriate level of relative humidity and temperature, in accordance with the water sorption isotherms in [53,54]. The moistening time of the samples in the chamber was 72 h. In this way, samples with water content in cellulose and aramid paper above 2.0% and 1.5%, respectively, were prepared. The difference in water content between cellulose and aramid samples is due to the different hygroscopicity of both materials.

Due to the inability to set the relative humidity of the air below 10% in the environmental chamber, the remaining samples with a lower moisture level were obtained by placing them in airtight vessels with the appropriate amount of water calculated on the basis of the sample weight, the volume of the vessels, and the assumed moisture level. Samples in sealed vessels were moistened for 168 h. Samples of fibrous materials prepared in this way were impregnated and conditioned in mineral oil, natural ester, or synthetic ester for a period of about 30 days. Before testing, the water content of fibrous material samples was determined using the Karl Fischer Test KFT method in accordance with the International Electrotechnical Commission standard IEC 60814 [53]. Methanol was used to extract water from fibrous samples.

After the thermal conductivity tests, the water content of fibrous materials was measured by control using the KFT method. These studies did not show a significant impact on the conditions of thermal conductivity tests on the change of water content in samples of fibrous materials. The differences in moisture were within the measurement error of the KFT method and did not exceed the 0.2 percentage point.

#### *2.3. Thermal Conductivity Measurements*

To measure the thermal conductivity coefficient λ of papers impregnated with insulating liquids, the self-developed measuring system presented in Figure 1 was used. This system has been described in more detail in publications [55,56].

Δ*λ*

Δ

**Figure 1.** System for measuring the thermal conductivity coefficient λ λ of solid materials; **a**—cooler, **b**,**e**—auxiliary plates with measuring probes, **c**—test sample, **d**—main heater, **f**—auxiliary insulation, **g**—auxiliary heater, **h**—main insulation.

The used measuring system is based on the idea of measuring the thermal conductivity coefficient λ using fixed methods. The concept of measurement consists of causing in the test sample, of known thickness *d* and surface area S, a thermal disturbance ∆*T*. Based on these quantities, the coefficient of thermal conductivity λ can be determined from the relationship [56]:

$$
\lambda = \frac{P}{S} \cdot \frac{d}{\Delta T} \tag{1}
$$

where *P* is the heat of power source (W), *S* is the surface area of the tested sample (m<sup>2</sup> ), *d* is the thickness of the tested sample (m), and ∆*T* is the temperature drop in the tested sample (◦C).

During the measurement, the tested sample (c) was placed between the main heater (d)—supplied with a constant voltage source and the cooler (a)—supplied with water from an external circuit. In the presented measuring system, the tested sample (c) was placed under the main heater (d) in order to eliminate the effect of convection on the temperature drop in the tested sample. The main heater (d) with the power *P* was designed to generate thermal energy which, penetrating through the sample, causes a decrease in temperature ∆*T* for a given thickness of sample *d*. The surface area of the main heater (and also the auxiliary heater) corresponds to the surface area of the tested sample *S*. The cooler (a) was designed to provide a constant temperature on the bottom surface of the sample. The temperature drop in the tested sample (c) was determined based on the temperature measurement in the auxiliary plates (b), which are made of aluminum and in which the measuring probes (Pt1000 measuring probes connected with the temperature recorder Apek APL 154 (Apek, Warsaw, Poland)) were placed. Auxiliary insulation (f) was placed over the main heater. In turn, an auxiliary heater (g) was placed over the auxiliary insulation to compensate for the heat flow from the main heater (d) in a vertically upward direction. The auxiliary heater (g), like the main heater, is powered by a DC voltage source. The task of the auxiliary heater (g) is to generate such an amount of heat that the indications of the measuring probes placed in the auxiliary plates (e) are the same, which means no heat flow in a perpendicular upward direction. In addition, to isolate heat loss in the measuring system, external insulation was also used (h). Adjusting the settings of both power supplies, as well as recording temperature was done using a special computer program written in the LabView environment (National Instruments, Austin, Texas, TX, USA).

Thermal conductivity error ∆λ was calculated on the basis of the complete differential of Equation (1). The maximal value of the error was smaller than 2% (Appendix A).

#### **3. Experiment Results and Analysis**

#### *3.1. Thermal Conductivity Coe*ffi*cient of Unimpregnated Papers*

Table 2 presents the results of measurements of the thermal conductivity of unimpregnated cellulose paper and unimpregnated aramid paper depending on the temperature. The WCP moisture level of both analyzed types of paper was similar and very low—it was 0.52% for cellulose paper and 0.44% for aramid paper, respectively.

**Table 2.** Thermal conductivity coefficient λ of cellulose and aramid unimpregnated papers at different temperatures *T*.


As can be seen from the presented results, the thermal conductivity of unimpregnated cellulose paper was about 9 to 25% higher (depending on the temperature) than the thermal conductivity of unimpregnated aramid paper. This difference decreases with increasing temperature. Both aramid and cellulose paper are polymers, the first of which is a synthetic polymer and the second a natural one. Polymers are characterized by a low value of thermal conductivity coefficient [57]. The thermal conductivity of this group of materials depends on many factors such as structure, molecular weight, density, and degree of crystallinity. As the polymer structure is ordered, its thermal conductivity increases. Both unimpregnated cellulose paper and unimpregnated aramid paper in their structure contain defective structures among other voids, amorphous areas, and entanglements, which impede the spread of heat [58]. They cause a large dispersion of phonons, which reduces heat transport. In addition, voids are filled with air, whose thermal conductivity is very small (0.025 W·m−<sup>1</sup> ·K−<sup>1</sup> ) [59], smaller than the thermal conductivity of papers. A higher value of the thermal conductivity coefficient of cellulose paper results from its density. Both analyzed types of paper were characterized by the same thickness—the thickness of a single layer of paper was 0.005 mm. The density of cellulose paper was 915 kg·m−<sup>3</sup> and of aramid paper was 709 kg·m−<sup>3</sup> [46]. This means that cellulose paper in its structure has less difficulty in transporting through heat voids, which are filled with air. Due to the fact that the thermal conductivity of paper is a resultant of the thermal conductivity of paper fibers and air trapped between the fibers, the thermal conductivity of unimpregnated cellulose paper is higher than the thermal conductivity of unimpregnated aramid paper.

#### *3.2. E*ff*ect of the Moisture on the Thermal Conductivity of Impregnated Cellulose Paper*

Table 3 and Figure 2 present the results of measurements of the thermal conductivity of cellulose paper impregnated with various electrical insulating liquids (mineral oil, synthetic ester, and natural ester) depending on the moisture content.

Comparing the results of the measurements given in Tables 2 and 3, it can be said that the treatment of cellulose paper impregnation with insulating liquids resulted in an increase in its thermal conductivity, which is associated with the replacement of air trapped in the pores of the paper with an insulating liquid which has about one order of magnitude greater thermal conductivity. This conductivity at 25 ◦C is equal to 0.133 W·m−<sup>1</sup> ·K−<sup>1</sup> for pure mineral oil, 0.158 W·m−<sup>1</sup> ·K−<sup>1</sup> for pure synthetic ester, 0.182 W·m−<sup>1</sup> ·K −1 for pure natural ester [60,61], and only 0.025 W·m−<sup>1</sup> ·K −1 for air.

Based on the measurement results shown in Table 3, it can be concluded that as the temperature increases, the coefficient of thermal conductivity of the impregnated cellulose paper increases, regardless of the type of impregnating liquid. As is well known, the thermal conductivity of pure liquids decreases with temperature (except for water and glycerin). For tested pure insulating liquids in the examined temperature range of 25 to 80 ◦C, a decrease in their thermal conductivity by 0.007 W·m−<sup>1</sup> ·K−<sup>1</sup> was observed regardless of their type [60]. However, the thermal conductivity of unimpregnated cellulose paper increased with increasing temperature, and this increase, in the studied temperature range of 25 to 80 ◦C, was 0.034 W·m−<sup>1</sup> ·K−<sup>1</sup> (Table 2). This was due to the fact that both the thermal conductivity of gases and solids increase with temperature. Thus, it can be concluded that primarily the cellulose fibers were responsible for the heat transfer in impregnated cellulose paper, not the insulating liquid. In addition, the conductivity of cellulose paper impregnated with insulating liquids was higher than the conductivity of pure insulating liquids. The increase in the thermal conductivity of the oil-impregnated paper is similar for all tested liquids. In such complex systems, many factors influence heat conduction. However, a similar effect of the tested liquids on the interactions in the papers was observed [57,58]. Since the paper seems to be primarily responsible for heat conduction a similar effect is observed.


**Table 3.** Thermal conductivity coefficient λ of cellulose paper impregnated by various dielectric liquids depending on the water content of paper WCP, at different temperatures *T*.

**Figure 2.** Thermal conductivity coefficient λ of cellulose paper impregnated by various dielectric liquids *λ* depending on the water content of paper WCP and temperature *T*: (**a**) cellulose paper impregnated by mineral oil; (**b**) cellulose paper impregnated by synthetic ester; (**c**) cellulose paper impregnated by natural ester.

Analyzing the measurement results presented in Table 3, it can be seen that the thermal conductivity of cellulose paper impregnated with various insulating liquids increased with increasing moisture content. Water probably penetrated into the pores of the paper. The increase in thermal conductivity of impregnated paper, caused by an increase in moisture, was associated with about four times greater thermal conductivity of water (about 0.60 W·m−<sup>1</sup> ·K−<sup>1</sup> ) [60] compared to the thermal conductivity of the analyzed pure insulating liquids (average 0.15 W·m−<sup>1</sup> ·K −1 ) [61]. The increase in the thermal conductivity of the impregnated cellulose paper, accompanied by the increase in moisture, was practically independent of the type of insulating liquid. The average value of the increase in thermal conductivity fluctuated in the range of 5 to 7% for all types of analyzed liquids. However, the increase in paper thermal conductivity, caused by an increase in moisture, depended on the measurement temperature. As the temperature increased, the increases in thermal conductivity of the impregnated cellulose paper were becoming smaller.

In summary, moisture in cellulose insulation increased the thermal conductivity of this insulation. Thus, moisture, in addition to many of the disadvantages described at the beginning of this article, also has positive features. Greater thermal conductivity of cellulose insulation will result in more efficient heat dissipation from the transformer windings to the cooling liquid. This in turn can lower the hot spot temperature.

#### *3.3. E*ff*ect of the Moisture on the Thermal Conductivity of Impregnated Aramid Paper*

Table 4 and Figure 3 present the results of measurements of the thermal conductivity of aramid paper impregnated with various insulating liquids (mineral oil, synthetic ester, and natural ester) depending on the moisture content.

Based on the results of the measurements, it can be seen that the thermal conductivity of aramid paper, impregnated with insulating liquids, similar to cellulose paper, was higher than the thermal conductivity of unimpregnated aramid paper (containing only air in pores) [60,61]. However, this increase was smaller than for cellulose paper and at a temperature below 60 ◦C, it did not exceed the thermal conductivity of insulating liquids. It is possible, therefore, that in this case conduction at the liquid-aramid paper interface is less effective.

As in the case of cellulose paper, it can also be seen that the thermal conductivity of all analyzed samples of aramid paper increased with increasing temperature. This means that the heat transfer in impregnated aramid paper, as in the case of cellulose paper, is carried out primarily through the paper fibers, not the insulating liquid.


**Table 4.** Thermal conductivity coefficient λ of aramid paper impregnated by various dielectric liquids depending on the water content of paper WCP, at different temperatures *T*.

*λ* **Figure 3.** Thermal conductivity coefficient λ of aramid paper impregnated by various dielectric liquids depending on the water content of paper WCP and temperature *T*: (**a**) aramid paper impregnated by mineral oil; (**b**) aramid paper impregnated by synthetic ester; (**c**) aramid paper impregnated by natural ester.

Analyzing the results of the measurements in Table 4, it can be seen that the thermal conductivity of aramid paper impregnated with various insulating liquids, like in the case of cellulose paper,

∙ − ∙ − ∙ −

increased with increasing moisture. Water, which has four times greater thermal conductivity than the thermal conductivity of pure insulating liquids, is responsible for this result [61]. It interacts with the paper, probably binds to it, and penetrates into its pores.

The increase in thermal conductivity of impregnated aramid paper, accompanying the increase in moisture, was practically independent of the type of insulating liquid, and its average value was in the range of 3 to 5% for all types of analyzed liquids. It can be associated with the similar influence of all investigated oils (EN, ES, OM) on aramid paper. However, the increase in thermal conductivity of aramid paper, caused by an increase in moisture, similar to cellulose paper, depended on temperature. For higher temperature values, the increases in thermal conductivity of aramid paper were getting smaller.

In summary, the moisture content of aramid insulation slightly increases its thermal conductivity. Thus, as in the case of cellulose paper, a slightly higher thermal conductivity of aramid insulation will result in more efficient heat dissipation from the transformer windings, which will contribute to a lower value of the hot spot.

#### *3.4. Comparison of Thermal Conductivity of Impregnated Cellulose and Aramid Paper in the Context of Their Moisture*

Based on the results of the measurements presented in Tables 3 and 4, it can be stated that the increase in the moisture content of the paper insulation caused an increase in its thermal conductivity, both for cellulose and aramid paper.

The increase in thermal conductivity, caused by moisture, in the case of cellulose paper was 5 to 7%, and in the case of aramid paper, this increase was slightly smaller, equal to 3 to 5%. The reason for this was certainly the upper limit to which the samples could be moistened, which was 5 to 7% WCP for cellulose paper, and only 4% WCP for aramid paper. On this basis, it can be said that moisture in the paper causes a similar increase in thermal conductivity, regardless of the type of paper.

The increase in thermal conductivity caused by moisture was getting smaller as the temperature increased. This regularity was observed for practically all types of impregnating liquid and for both types of analyzed paper (cellulose and aramid).

Figure 4 presents the coefficient of thermal conductivity of paper, depending on moisture, for various types of paper, and impregnating liquid. The values of thermal conductivity are presented for 80 ◦C, as the most typical temperature for paper insulation of transformer windings.

**Figure 4.** Thermal conductivity of paper insulation depending on moisture WCP measured at 80 ◦C, for various types of paper and insulating liquid.

∙ − ∙ −

λ ∙

As can be seen in Figure 4 the thermal conductivity of paper insulation increases linearly with increasing moisture content. On this basis, a linear equation has been proposed, describing the thermal conductivity of paper insulation depending on moisture:

$$
\lambda = \mathbf{a} + \mathbf{b} \cdot (\text{WCP}) \tag{2}
$$

where a is the thermal conductivity of paper insulation for zero moisture (W·m−<sup>1</sup> ·K −1 ), b is the coefficient determining the impact of moisture on the thermal conductivity of a given material (W·m−<sup>1</sup> ·K−<sup>1</sup> ·%−<sup>1</sup> ), and WCP is a percentage of water content in paper insulation (%). The parameters *a* and *b* of the linear equations were obtained for the six combinations of analyzed materials (Table 5).

**Material a ∆***a b* **∆***b* **(W**·**m**−**<sup>1</sup>** ·**K**−**<sup>1</sup> ) (W**·**m**−**<sup>1</sup>** ·**K**−**<sup>1</sup>** ·**%**−**<sup>1</sup> )** CP-MO 0.189 0.002 0.0038 0.0009 CP-SE 0.209 0.001 0.0019 0.0003 CP-NE 0.229 0.001 0.0023 0.0002

AP-MO 0.135 0.001 0.0018 0.0004 AP-SE 0.181 0.001 0.0012 0.0003 AP-NE 0.191 0.002 0.0010 0.0007

**Table 5.** The parameters *a* and *b* of the linear equations for the six combinations of analyzed materials; ∆*a* and ∆*b* mean absolute standard error of parameters *a* and *b*, respectively.

The *b* factor, determining the effect of moisture on the thermal conductivity of a given material, has very different values depending on the type of paper. For cellulose paper, these values (0.0021 <sup>÷</sup> 0.0038 (W·m−<sup>1</sup> ·K−<sup>1</sup> ·%−<sup>1</sup> )) are much higher than for aramid paper (0.0010 <sup>÷</sup> 0.0018 (W·m−<sup>1</sup> ·K −1 ·%−<sup>1</sup> )). This means that the same moisture content of paper insulation causes a greater (about two times) increase in the thermal conductivity of cellulose paper than aramid paper. On this basis, it can be concluded that cellulose paper is more sensitive to moisture than aramid paper from the point of view of its thermal conductivity.

The situation is similar when analyzing the *b* factor depending on the type of insulating liquid. For mineral oil, the *b* factor is about twice as high as for both types of analyzed esters. This means that the same moisture content causes a double increase in thermal conductivity of paper impregnated with mineral oil compared to the thermal conductivity of paper impregnated with synthetic or natural ester. On this basis, it can be said that mineral oil is more sensitive to moisture than the analyzed esters from the point of view of thermal conductivity of paper impregnated with these liquids.

In conclusion, it can be said that the most susceptible paper insulation to changes in its thermal conductivity, caused by moisture, was cellulose paper impregnated with mineral oil, for which the coefficient *<sup>b</sup>* was 0.0038 (W·m−<sup>1</sup> ·K −1 ·%−<sup>1</sup> ). In turn, the least susceptible paper insulation to changes in its thermal conductivity was aramid paper impregnated with synthetic or mineral ester, for which the coefficient *<sup>b</sup>* is four times smaller, equal to 0.0010 <sup>÷</sup> 0.0012 (W·m−<sup>1</sup> ·K−<sup>1</sup> ·%−<sup>1</sup> ).

The obtained equations are useful for determining the temperature field of the transformer at the design stage, and especially during its operation when the moisture content of paper insulation increases over the years. In recent years, advanced diagnostic techniques have been developed very dynamically that enable the determination of moisture content in paper insulation. These methods are based on the phenomenon of dielectric spectroscopy (Recovery Voltage Measurement RVM, Frequency Domain Spectroscopy FDS) [62,63].

#### **4. Conclusions**

Moisture in the paper insulation increased its thermal conductivity, regardless of the type of paper and insulating liquid that the paper has been impregnated with. In the case of cellulose paper, this increase is 5 to 7%, and for aramid paper, the increase in thermal conductivity fluctuates within

3 to 5%. The increase in thermal conductivity of paper insulation, caused by moisture, is associated with a much higher water conductivity (about 0.60 (W·m−<sup>1</sup> ·K −1 )) compared to the conductivity of both unimpregnated paper (0.06 <sup>÷</sup> 0.12 (W·m−<sup>1</sup> ·K−<sup>1</sup> )), as well as used pure insulating liquids (about 0.15 (W·m−<sup>1</sup> ·K −1 )).

The increase in thermal conductivity caused by moisture became smaller as the temperature increased. This relationship was observed for practically all types of impregnating liquids and for both types of analyzed papers.

Based on the obtained results, it was found that the same moisture content of paper insulation caused a greater (about two times) increase in the thermal conductivity of cellulose paper than aramid paper. Thus, cellulose paper was more sensitive to moisture than aramid paper in terms of its thermal conductivity. The same moisture content caused a two-fold increase in the thermal conductivity of paper impregnated with mineral oil compared to the thermal conductivity of paper impregnated with synthetic or natural ester. This means that mineral oil is more sensitive to moisture than the analyzed esters from the point of view of thermal conductivity of paper impregnated with these liquids.

Obtained equations, describing the effect of moisture on the thermal conductivity of paper insulation, can be useful for determining the temperature field of the transformer both at the design stage and during its operation.

**Author Contributions:** Conceptualization, Z.N. and G.D.; methodology, G.D., Z.N. and R.L; software, R.L.; validation, G.D., P.P. and Z.N.; formal analysis, G.D., Z.N. and A.M.; investigation, G.D., Z.N., R.L. and P.P.; resources, G.D. and Z.N.; data curation, G.D.; writing—original draft preparation, G.D., Z.N., P.P., R.L., A.M., L.D., T.B. and A.T.; writing—review and editing, G.D. and Z.N.; visualization, G.D. and Z.N.; supervision, G.D. and Z.N.; project administration, G.D. and Z.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This research was funded by the Ministry of Science and Higher Education, grant number 0711/SBAD/4411 and 0912/SBAD/2004.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A. Thermal Conductivity Error Estimation**

Thermal conductivity error was calculated on the basis of the complete differential of equation, which helps to calculate the conductivity:

$$
\lambda = f(p, q, \Delta T) = \frac{p \cdot d}{\Delta T} \tag{A1}
$$

where <sup>λ</sup> is the thermal conductivity coefficient (W·m−<sup>1</sup> ·K −1 ), *<sup>p</sup>* is the surficial thermal load (W·m−<sup>2</sup> ), *d* is the thickness of sample (m), and ∆*T* is the temperature difference on the sample surfaces (K).

By expanding the λ function into Taylor series around the measurement values and neglecting words higher than the first row and replacing infinitely small increments of variables with independent finite increments, we will get:

$$
\Delta\lambda = \left| \frac{\partial\lambda}{\partial p} \right| \cdot \Delta p + \left| \frac{\partial\lambda}{\partial d} \right| \cdot \Delta d + \left| \frac{\partial\lambda}{\partial \Delta T} \right| \cdot \Delta(\Delta T) \tag{A2}
$$

and

$$
\Delta\lambda = \left| \frac{d}{\Delta T} \right| \cdot \Delta p + \left| \frac{p}{\Delta T} \right| \cdot \Delta d + \left| -\frac{p \cdot d}{\left(\Delta T\right)^2} \right| \cdot \Delta(\Delta T) \tag{A3}
$$

It should be noted, that:

$$p = \frac{\mathcal{U} \cdot \mathcal{I}}{\mathcal{S}} \tag{A4}$$

where *U* is the voltage of heater (V), *I* is the current of heater (A), and *S* is the surface of the sample (m<sup>2</sup> ). It means, that:

$$
\Delta p = \left| \frac{\partial p}{\partial U} \right| \cdot \Delta U + \left| \frac{\partial p}{\partial I} \right| \cdot \Delta I + \left| \frac{\partial p}{\partial S} \right| \cdot \Delta S \tag{A5}
$$

and

$$
\Delta p = \left| \frac{I}{S} \right| \cdot \Delta II + \left| \frac{U}{S} \right| \cdot \Delta I + \left| -\frac{I \cdot U}{S^2} \right| \cdot \Delta S \tag{A6}
$$

It should be noted, that:

$$\mathbf{S} = \mathbf{x} \cdot \mathbf{y} \tag{A7}$$

where *x* is the length of the sample (m) and *y* is the width of the sample (m).

It means, that:

$$
\Delta S = \left| \frac{\partial S}{\partial \mathbf{x}} \right| \cdot \Delta \mathbf{x} + \left| \frac{\partial S}{\partial y} \right| \cdot \Delta y \tag{A8}
$$

and

$$
\Delta \mathcal{S} = \left| y \right| \cdot \Delta \mathfrak{x} + |\mathfrak{x}| \cdot \Delta y \tag{A9}
$$

Summarizing:

$$
\Delta\lambda = \begin{vmatrix}
\frac{d}{\Delta T} \Big| \cdot \left( \left| \frac{I}{S} \right| \cdot \Delta II + \left| \frac{II}{S} \right| \cdot \Delta I + \left| -\frac{I \cdot II}{S^2} \right| \cdot \left( \left| y \right| \cdot \Delta x + |x| \cdot \Delta y \right) \right) + \left| \frac{p}{\Delta T} \right| \cdot \Delta d + \\
& + \left| -\frac{p \cdot d}{\left( \Delta T \right)^2} \right| \cdot \Delta (\Delta T)
\end{vmatrix} \tag{A10}
$$

Parameters *d*, ∆*T*, *I*, *U*, *S*, *p*, *x*, and *y* are measurement results or equipment setting values. All ∆*d*, ∆(∆*T*), ∆*U*, ∆*I*, ∆*x*, and ∆*y* mean accuracy of used equipment and measurers, which are known. It is possible to calculate thermal conductivity error ∆λ on the basis of equations (A10). The value of maximal relative error was smaller than 2%.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
