Flashover Voltage of Epoxy FRP Insulators with Different Surface Roughness and Groove under Nanosecond Pulses in SF₆

Abstract

In order to further improve the insulation performance of fiber reinforce plastic (FRP) materials used in electromagnetic pulse (EMP) simulators, the flashover characteristics of FRP materials with different surface roughness and groove, i.e., those who are easily achieved and have a prominent effect, are investigated in 0.1 MPa SF₆ under nanosecond pulse voltage with a rise time of 20–30 ns. The experimental results show that surfaces with different roughness have no significant influence on the flashover voltages of the FRP insulators, and both the convex grooves made of FRP and the convex grooves with nylon rings inlaid to form projections can improve the surface flashover voltage of epoxy FRP insulators under nanosecond pulse, in which the effect of the former surface is more obvious. For the insulators with convex grooves made of FRP, it is found that the root of the FRP protrusions breaks down after a number of shots with the occurrence of carbonization channels and spots, which is nonexistent for the nylon projections. Combined with the test results of surface characteristics, the surface roughness and the secondary electron emission yield (SEEY) are not key factors of flashover characteristics in SF₆ under nanosecond pulse, arguably due to the fact that the energy needed for an incident electron to ionize an SF₆ molecule is lower than that to excite two secondary electrons. Hence, the flashover performance cannot be improved by adjusting the surface roughness, and the flashover channel is principally governed by the macroscopic distribution of electrical field which can be changed by the convex groove. Breakdown phenomena of FRP protrusions indicate that the bulk insulation performance of resin FRP is weaker compared to pure resin because of its composite structure, as well as the impurities and voids introduced in the manufacturing process. The results are instructive for the design of FRP insulation structures in the compact EMP simulator.

1. Introduction

An EMP simulator, which can generate an extensive and high-amplitude electromagnetic pulse environment [1,2,3], usually consists of the pulsed power source and the radiation antenna. When the simulator works, energy stored in the capacitors is released in a short time via the pulsed power source and is applied to the radiation antenna to generate the EMP in order to meet the standard requirements. With the development of EMP test requests for large-scale test objects, including electronic equipment, the operating voltage of the simulator has to be raised on the premise of generating fast front pulses. Therefore, there is the tough problem centered around how to realize the high voltage insulation in the very compact structure. At present, utilizing FRP materials with excellent mechanical and insulation performance to form a high-pressure insulating environment is a feasible method. During the construction of EMP simulators, FRP materials are frequently used to replace other conventional polymer materials at the expense of higher cost. For long-term consideration, to satisfy the needs of higher operating voltage and more stable working condition of the simulator, the insulation performance, especially the flashover performance of FRP materials, has to be improved sustainedly.

There are many factors which affect the surface flashover voltage of dielectric, such as applied voltage waveform, electrode material, surrounding insulating environment, geometry and material of the insulator, surface characteristics (including SEEY, surface roughness, surface conductivity, and surface traps), and so on [4,5,6,7]. Furthermore, surface modification methods including pre-discharge treatment, fluorination, and coating, etc., are applied to improve the surface flashover performance [8,9,10,11,12]. For engineering practice, whether the method is easy to implement is a major concern for engineers. This paper mainly focusses on the methods of modification of surface roughness and introduction of grooves, which are easily achieved and have a prominent effect. Many scholars have carried out in-depth researches on these factors. Zhao et al., studied the flashover voltage of silicone rubber depending on surface roughness under AC voltage in air, and found that the dry flashover voltage of a sample with high surface roughness increased with surface roughness while the wet flashover voltage was reversed [13]. Xue et al., investigated the effects of surface roughness on surface charge accumulation behavior and surface flashover performance of alumina-filled epoxy resin spacers in SF₆/N₂ mixtures under DC voltage, and attributed the variation of flashover voltage to surface conductivity, surface traps, and creep distance [14]. Guo et al., investigated the influence of surface roughness on the flashover strength of an insulator in a vacuum under impulse voltage, and established a model to explain the experimental phenomena that flashover voltage threshold depicted a rise-and-fall trend with roughness increasing [15]. Taking the surface roughness into consideration, Naruse et al., proposed a method to estimate the surface flashover voltage in a vacuum and the calculation results agreed with the experiments [16]. Iwata’s study showed that the surface flashover voltage in the direction perpendicular to the groove structure is approximately 1.2 times larger than that in the parallel direction under the AC voltage in a vacuum [17]. Liu and Huo et al., applied the fabrication method of micro-grooves on the surface of alumina and polymer insulators, and found that the method can increase the flashover voltage significantly [18,19]. Cheng et al., studied the effect of periodically grooved dielectric on extending the flashover time lag in a vacuum, and ascribed the effect to the suppression of SEEA process [20]. Chang’s computational and simulative analysis showed that grooved surface can explicitly suppress multipactor in the development stage in a vacuum, and the analysis was verified by further experiments [21].

To sum up, most researches focus on the influence of the surface characteristics on the surface flashover in a vacuum, and less attention is paid to the SF₆ environment. Meanwhile, there are relatively few studies on the flashover problem of insulating materials under nanosecond pulses. In the paper, the insulators made of FRP material by a vacuum technology with different surface roughness and different grooves are experimentally investigated under nanosecond pulses in SF₆. Moreover, the results are analyzed by flashover theory and compared to the conditions in a vacuum.

2. Experimental Setup

2.1. Epoxy FRP Insulator Samples

The fiber used in the epoxy FRP material in this paper is a kind of alkali-free glass fiber roving, with the characteristics of high strength, high modulus, low density, and good water resistance. The modified E-51 epoxy resin, with excellent electrical insulation, high mechanical strength, and chemical corrosion resistance characteristics, is adopted. In the molding processing, the oil staining on the mold surface should, firstly, be cleaned with gasoline and acetone sequentially, and then coated with mold release agent and mold release film, and dried for later use. Afterwards, the glass fiber cloths are stacked in the mold, which is then sealed and pumped to a vacuum. The heated resin with low viscosity is then introduced to the mold via flow channels and impregnated into the stacked glass fiber cloths. When the mold is impregnated with resin, it is kept in an oven for curing reaction. After the solidification, the FRP material is separated from the mold and further machined to different surfaces.

There are four kinds of FRP samples used in this paper, as shown by Figure 1: an original sample with smooth surface, a sample with rough surface (grinding with 100 mesh coarse sandpaper), and two samples with convex grooves. The samples are cylindrical with a height of 50 mm and a diameter of 40 mm. The groove depth and thickness of the FRP insulators with convex grooves are 15 mm and 8 mm, respectively. The number of grooves is 3, and the groove spacing is 2 mm. The difference between them is the material of the convex parts. One is machined from a larger FRP cylinder and the other one is a combination of FRP cylinder with concave grooves and nylon rings. The nylon rings and the fiberglass insulators are bonded with epoxy adhesive to make the nylon rings’ interface seamless.

2.2. Experiment Platform and Procedure

In order to study the surface flashover characteristics of FRP insulators in 0.1 MPa SF₆, an experimental platform with the equivalent circuit, as shown in the upper figure of Figure 2, is established, which is similar to the experimental platform in [22,23], with the exception of the compact Marx generator. The output pulse voltage amplitude ranges from 70 kV to more than 500 kV. The 10–90% time increase in the pulse voltage is approximately 20 ns. The oscillation of the trailing edge is caused by the distributed capacitance at the output end of Marx generator, i.e., there are distributed capacitances between the output cable and the Marx generator housing, and between the input cable and the experimental cavity.

The cylindrical sample is sandwiched in the upper and lower parallel column electrodes. During the experiment, the cavity is firstly evacuated to a vacuum, and then SF₆ gas is filled to the pressure of 0.1 MPa. This environment is chosen as the experimental environment to simulate one of the working environments of epoxy resin FRP materials in the EMP simulator. The flashover phenomenon can be observed through the measured waveforms and the visible quartz glass window.

The Marx generator is charged to a high voltage so that the surface flashover of the FRP insulator can occur at the front edge of the pulse. Each experimental sample is subjected to 20 experiments.

3. Experimental Results

The typical flashover voltage waveforms are shown in Figure 3. The flashover voltage obtained from D₁ is used for monitoring the status of the Marx generator, while that from D₂ directly shows the flashover process.

The experimental results of the four samples are shown in Figure 4. The flashover voltage of the FRP insulator with a smooth surface is basically the same as that with rough surface, indicating that the surface roughness has no significant influence on the flashover voltage under the experimental condition. The surface flashover voltages of the two types of FRP insulators with convex grooves have increased compared with the FRP insulators with no grooves This phenomenon is different from the comparative experimental results of FRP insulators with and without concave grooves [24], in which their flashover voltages are nearly the same. The flashover voltage (~18% improved) of the cylindrical FRP insulators embedded with nylon rings is higher than that (~11% improved) with FRP convex grooves, while the probable cause of the difference is the groove material with inequable permittivity.

A bulk breakdown phenomenon should be noted. In the experiments of some FRP insulators, the peak magnitude of the applied flashover voltage drops suddenly and sometimes dramatically after certain shots, and carbonized breakdown channels are found in the insulators. There are mainly two types of carbonized channels as Figure 5 shows. One is formed in the root of the convex groove which can only slightly reduce the flashover voltage. The other one is the bulk breakdown channel connecting the anode to the cathode, and, in this case, the samples’ flashover voltage (in fact, its breakdown voltage) is greatly reduced. The latter carbonized channels are relatively rare.

4. Analysis and Discussion

4.1. Influence of the Surface Roughness

The surface profile of the FRP insulators is measured by a high-resolution laser profilometer, and its height resolution can be as high as 2 nm. The scanning objects are the flank areas of the cylinder samples. Because the flank area is not a flat plane, the scanning area is chosen as a rectangle with a high length to width ratio. The length along the axial direction of the cylinder is 10 mm, while the width is only 0.1 μm. The scanning steps are 10 μm and 20 nm, respectively.

The measurement results of the surface profile are shown in Figure 6. Because it is hard to set the axial direction of the samples strictly parallel to the scanning direction, the baselines in Figure 6 are not parallel to the x-axis. However, the following analysis is not affected. Figure 6 shows that for samples with smooth surface, the actual surface is not flat on a micro-level, and the height of the protrusions and burrs is around 100 μm. For samples with rough surface, the density of protrusions and burrs is larger and most of the protrusions’ height is close to 200 μm. Therefore, the surface’s microstructures of the two type samples differ a lot.

Surface roughness has a great impact on the surface flashover characteristics of insulating materials. At present, the researchers’ understanding of the influence of surface roughness on flashover voltage can be classified into the following two aspects [12,13,14,15,16,17]. The first aspect corresponds to the improvement in flashover voltage. The material with uneven surface makes the creep distance of the surface increase, and thus directly hinders the secondary electron multiplication process and suppresses the formation of the through flashover. From the point of surface traps, the surface rough can bring about deep level traps and can restrain charge movement. The charges are attracted in the trap center and it is difficult for them to participate in the development of flashover. The flashover voltage can increase accordingly. From the other aspect, the increase in the surface roughness results in an increase in the local surface electric field distortion, which makes the flashover easier to occur and results in a reduction in the flashover voltage. However, the experimental results in this paper cannot be classified into neither of the above two aspects because the flashover voltages of samples with smooth and rough surface are almost the same.

Another key characteristic of the surface with respect to flashover phenomenon is the SEEY, particularly for a vacuum insulating environment. The change in the surface roughness always leads to a different SEEY, which effectively determines the flashover voltage in a vacuum. The SEEY of smooth and rough FRP insulators are shown in Figure 7, and the incident path of the electrons is perpendicular to the surface. The curve of nylon material is used for later analysis.

For the electrons with the same injection energy, the SEEY of the smooth surface is much larger than that of the rough surface over most part of the injection energy range. With an increase in the injection energy, the difference becomes less and less significant. The peak value of the coefficient occurs when the injection energy is around 320 eV for both kinds of materials. The measurement results are agree with those for other polymer materials.

According to the secondary electron emission avalanche (SEEA) theory [25], because the SEEY of the rough FRP material is smaller than that of the smooth FRP material, the number of secondary electrons of smooth FRP material emitted from the surface is larger in the electron multipactor process, as is the accumulation of positive charges on the surface under the applied voltage. In the following development stages, as the emitted electrons collide with the surface of the FRP material, the desorption process of SF₆ gas molecules adsorbed on the surface of FRP insulator and their ionization contribute to the electron multipactor process along with the secondary electrons emitted from the surface. In addition, the surface electric field distortion near the cathode caused by the accumulated positive charge will enhance the electrical field around the cathode. Hence, for the smooth FRP material with a higher coefficient, and thus more emitted electrons, its flashover voltage will be lower than that of rough FRP material in a vacuum, which is not consistent with the experimental results in SF₆ under a nanosecond pulse voltage, as shown by Figure 4.

The applied pulse voltage and the insulating environment might offer a possible interpretation for the experimental phenomenon.

(1)

Compared to the long-duration applied voltage, i.e., AC, DC, or impulse voltage, the density of surface charge accumulated under nanosecond pulse is relatively lower due to the insufficient time for charges to migrate and be attracted by the surface traps, and the effect on the successive flashover process is not significant [26]. Furthermore, under long-duration voltages, the field distortion, due to a large amount of accumulated surface charges, can induce local discharge and benefits for the generation of seed electrons. However, under nanosecond pulses, the migration of charges (mainly electrons) is directly governed by the electron multipactor process, and there is no suitable conditions for other time-consuming processes. Therefore, the influence of surface charge accumulation and other accompanying surface characteristics may not be critical when discussing flashover phenomenon under nanosecond pulses.

(2)

Whether in the gas or on the dielectric surface, one of the preconditions for electron multiplication is that the electron’s incident energy must be above the threshold. In the SF₆ gas environment, when the incident electron energy is greater than 15.67 eV (first ionization energy), SF₆ gas molecules can be ionized effectively into electrons and SF₅ positive ions. The ionization processes in SF₆ and their corresponding incident energy are shown in Figure 8, while most peak values are reached when the energy of the incident electron is near 100 eV [27]. However, for most polymer materials, secondary electrons are emitted from the surface, and the energy threshold is above 100 eV when SEEY is greater than 2 for smooth or rough FRP, as shown in Figure 7. The threshold energy is much larger than the first ionization energy of SF₆ molecules. In the electron multiplication process, the gas rather than the dielectric surface provides the source of electrons. Therefore, it can be inferred that the diminution of the SEEY due to the rough surface of FRP material does not significantly improve the surface flashover performance in SF₆ gas.

In summary, the generation of secondary electrons on the surface of materials is the main mode for the electron multipactor process in a vacuum, which explains that the surface flashover voltage is much lower than the bulk breakdown voltage and that the surface roughness can greatly change the flashover characteristic of the dielectric in a vacuum. The situation in SF₆ gas is different, as the contribution of gas molecules’ ionization to the development of avalanche is greater than that on the surface. In addition, under nanosecond pulses, the flashover, mainly an electron process, lasts less than tens of nanoseconds. Hence, the surface roughness has little effect on the flashover characteristics of FRP insulators under nanosecond pulses in SF₆.

4.2. Influence of the Convex Grooves

The concave grooves of different shapes and sizes have no significant influence on the flashover voltage along the surface of FRP insulators [24], which further validates the conclusion of last section. This section discusses the samples’ characteristics concerning convex grooves, showing that the flashover voltage of samples is higher than samples with a smooth surface.

The convex grooves change the electrical field distribution on the samples’ surface. The contrast contour diagram of electric field distribution of FRP insulators with and without convex grooves is shown in Figure 9a. The relative permittivity of the E-51 epoxy resin is approximately 4, and the added fiberglass can enhance the permittivity; thus, the relative permittivity of FRP material is set to 5. The relative permittivity of nylon 1010 is between 2.5 and 3.6, and the value is set to 3.5. The electric field strength is extracted and analyzed along the surface path of sample surface. The data extraction path of insulator with grooves is on the outer surface of the grooves and is parallel to the path of insulator without grooves.

It can be seen from Figure 9b that the electric field strength of the FRP insulator without grooves is much higher than that of the FRP insulator with grooves on the outer surface. Generally, the development of streamer in a flashover needs sufficient time and proper condition on the path. After the high electric field region (usually the cathode triple junction (CTJ)) initiates flashover, the streamer propagates along the flashover path towards the other electrode. If the applied voltage lasts too short or the background field in the flashover path is too low, the streamer becomes extinct [28]; thus, the penetrating flashover cannot occur. In addition to the declination of the electrical field along the surface, the convex grooves can also change the orientation relation of the electrical field and the flashover channel, from parallel relation to partly parallel and partly perpendicular with each other. Therefore, the flashover voltage along the surface of cylindrical FRP insulator increases after the convex groove is introduced in the experiment, which also proves that the millimeter convex groove can improve the flashover voltage along the surface of cylindrical FRP insulator under a nanosecond pulse.

The bulk breakdown in the root of the convex groove made of FRP reduces the improvement effect; thus, the material of the convex groove is changed to nylon 1010. The nylon rings are glued to the FRP samples with concave grooves to avoid bulk breakdown, and the overall size of the samples with FRP grooves and nylon grooves is identical. In Figure 7, the SEEY of nylon material is lower than that of smooth FRP over the test range of injection electrons. On the basis of SEEA, the flashover voltage of samples with nylon grooves should be greater than the samples with FRP grooves, which is different from the experimental results. The experimental results further verify the conclusion that the SEEY is not a key factor in the experimental condition.

The electrical fields along the flashover path of samples with different grooves are analyzed in Figure 9. It can be seen from Figure 9 that the electric field of the FRP insulator with nylon grooves is slightly higher than that of the FRP insulator with convex grooves on the outer surface of grooves. According to the previous analysis, the propagating streamer on the nylon surface might become extinct at a lower voltage compared to the FRP surface. The simulation results are consistent with the flashover voltages measured in the experiment.

From the above analysis, it can be concluded that the beneficial effects of convex grooves on flashover voltage are achieved by changing the electrical field intensity and the direction on the samples’ surface. The optimization of electrical field distribution to improve the flashover characteristic of insulation structure is conventional and feasible.

4.3. Explanations for Bulk Breakdown

The FRP material is a kind of composite made from fiberglass, epoxy resin, and other additives. In the forming process, even if vacuum technology is applied, bubbles, micro-voids, and impurities can inevitably be introduced in the material. The SEM scanning pictures of the cutting-off FRP material are shown in Figure 10, where Figure 10a shows the cross-section picture from the end view and Figure 10b shows the side view.

It is clearly shown that the long and thin glass fibers with circular cross section are irregularly distributed in the resin, and the diameter of the fiberglass is approximately 20 μm. At the end of the fractured fiberglass, between the very close fiberglass or along the fiberglass, there might be bubbles, micro-voids, and impurities. Especially on the fracture surface, the external force in the machining process can break the crisp cured resin into smaller particles and leave deep or shallow micro-pores.

Bubbles, micro-voids, and impurities can affect the electric field distribution inside the FRP insulator under nanosecond pulses due to their lower permittivity compared with the resin and fiberglass, and their withstanding voltage can be higher than that of the resin with the same size. Local discharge can occur when high-voltage nanosecond pulses are applied. Over time, the local discharges can destroy the around resin dielectrics and the influence can extend along the direction of electrical field. If there are too many local discharges, a conductor can be inserted in the FRP material. The remaining part of the material cannot withstand the whole applied voltage, and then breakdown can occur. If the bubbles, micro-voids, or impurities are continuous, especially along the direction of electrical field, the situation can be even worse. Hence, the insulation performance of FRP material mostly depends on the forming process with vacuum technology. In our engineering practice to build high-voltage EMP simulator, parts of the insulation failure of FRP insulators occurred in the inner body but not on the surface. The flashover insulation failure can be recovered by cleaning or sometimes reprocessing the flashover location. However, after bulk breakdown, the insulator can completely lose its performance, and a long and black carbonization channel can be seen through the transparent resin.

5. Conclusions

In this paper, the flashover characteristics of the FRP insulator with different surfaces are experimentally studied in SF₆ gas insulation environment using nanosecond pulses. With the results analyzed and discussed, three major conclusions are summarized as follows. (1) Different from the vacuum insulation environment, the surface roughness can change the surface micro-structure and the SEEY of the FRP insulator’s surface, but has little positive influence on the flashover voltage under nanosecond pulses in SF₆. It is thus speculated that the SEEY of the material is not a critical parameter when evaluating the insulation characteristics in SF₆ environment under nanosecond pulse. (2) The convex grooves can enhance the flashover voltage of FRP insulators by means of changing the intensity and direction of the electrical field along the flashover path. It also implies that the optimization of electrical field distribution is a conventional but effective method to improve the surface flashover under nanosecond pulses. (3) Due to bubbles, micro-voids, and impurities introduced in the forming process, the bulk breakdown in the body of FRP insulators is one of the general failure modes. Going forward, attention should be paid to this phenomenon when applied in high-voltage insulation.