Decomposition Characteristics of SF6 under Flashover Discharge on the Epoxy Resin Surface

In this paper, the flashover discharging experiment was carried out on epoxy resin surface in an SF6 atmosphere under pin-plate electrodes, with the electrodes distance from 5 mm to 9 mm. The concentration of seven characteristic gases was detected, indicating that the concentration of SOF2 and CF4 was the two highest, followed by SO2, CO2, SO2F2, CS2, and H2S. Based on the changes in the concentration of the characteristic gases, a preliminary rule was proposed to predict the occurrence of flashover discharge on epoxy resin: When the concentration of SOF2 reaches twice of CF4 concentration, and the total concentration of both SOF2 and CF4 is much higher than that of H2S, a possible flashover discharge on the epoxy resin surface in SF6-infused electrical equipment occurs. Through the simulation of decomposition of epoxy resin, it has been revealed that H2O has different generation paths that can facilitate the formation of SOF2, finally influencing the concentration variation of the seven characteristic gases.


Introduction
Gas-insulated electrical equipment has been widely used in power system because of their advantages such as compact size, high reliability, and a long maintenance period [1,2]. It was found that a considerable proportion of the internal faults of GIS (gas insulated switchgear) are closely related to the insulating spacer used to support the copper conductor, among which the surface discharge of insulating spacer is hard to be detected due to its randomness and instantaneity [3,4]. So far, attention has been paid to the fault detection technology for insulating spacer. The monitoring system on the decomposed gases of SF 6 develops rapidly, which provides an effective way to judge the operation situation of SF 6 -infused electrical equipment. SF 6 can decompose into a series of subfluorides and sulfides when high energy imposes on SF 6 molecules on the condition of partial discharge, overheating, corona, electrical breakdown, or even ultraviolet radiation [5][6][7]. Conventionally, subfluorides can recombine with fluorinion to form SF 6 molecules again, owing to the intrinsic electronegativity property of SF 6 . However, in most cases, air and water vapor mixed with SF 6 exists. These impurities react with subfluorides and produce various gases like SOF 4 , SOF 2 , SO 2 F 2 , SF 4 , SO 2 , CF 4 , CO 2 , H 2 S, and so on [8][9][10], which prevent the recombination of SF 6 and subsequently reduce its insulation strength. Based on the concentration variation, a preliminary fault diagnosis method was provided to judge running situation of electrical equipment [8][9][10]. However, given the complicated internal situation of electrical equipment, there are many influential factors that tend to make concentration of decomposition components from SF 6 Materials 2019, 12 fluctuate drastically, such as temperature, humidity, and running voltage. Flashover discharge on the surface of epoxy resin gives rise to partial decomposition of epoxy resin, leading to potential chemical reaction between epoxy resin and SF 6 decomposition byproducts. However, this critical influential factor on SF 6 byproduct concentration attracts less attention today. The decomposition of SF 6 has been extensively studied. Sausers et al. [11] conducted experiments on the effect of oxygen on SF 6 decomposition byproducts and found that oxygen can promote the production of SO 2 F 2 and SOF 4 , which has little effect on the generation rate of SOF 2 . Derdouri et al. [12] designed a study on the SF 6 decomposition products when water vapor exists and found that SOF 2 and SO 2 F 2 generate stably in existence of water vapor. Later, it was found that SF 6 decomposition byproducts, including SOF 4 , SOF 2 , SO 2 F 2 , SO 2 , HF, would appear as long as water vapor and oxygen existed [13]. Additionally, a study reported that several SF 6 decomposition byproducts under four different kinds of partial discharge models, showing that CF 4 , CO 2 , SO 2 F 2 , SOF 4 , SOF 2 , SO 2 , and H 2 S were main decomposition products of SF 6 , among which the content of SOF 2 and SO 2 F 2 were the highest [14]. Furthermore, Belmadani et al. [15] explored the characteristics of SF 6 decomposition products under AC arc and pointed out that the total content of SOF 2 and SO 2 would increase with current intensity. It has been generally accepted that all SF 6 decomposition products concentration would increase with discharge energy, regardless of the kind of discharge [16][17][18]. The influences of solid insulation material on the decomposition of SF 6 were also studied. For example, it was reported that CF 4 would appear [12], since organic materials contain abundant C-H bonds that can easily react with fluorinion under high-energy arc discharge. Afterward, a small amount of CS 2 was detected when flashover discharges happened on epoxy resin surface in 110 kV GIS, thus making it a characteristic gas in a fault detection system [19]. There is no dispute that SOF 2 , SO 2 F 2 , and SO 2 will occur easily in partial discharge situation, and that H 2 S would be generated when discharge energy is high enough. Moreover, CF 4 , CO 2 , and CS 2 would be generated when organic solid insulation materials decomposed partially under high voltage [12,15].
So far, no report has focused on how epoxy resin affects the decomposition of SF 6 under flashover discharge yet. Additionally, it is too complicated to elaborate the chemical mechanism behind it. Although thermal and mechanical of epoxy resin can be enhanced through incorporating some filler into the polymer matrix, it still can decompose under arc discharge [20][21][22]. In the current study, pin-plate electrodes on the epoxy resin insulation spacer surface were designed; flashover discharge at different distances of pin-plate electrodes was carried out; and the concentration of decomposition byproducts of SF 6 , including SOF 2 , SO 2 F 2 , SO 2 , H 2 S, CF 4 , CO 2 , and CS 2 , was examined in order to find the relationship between decomposition of SF 6 under flashover discharge on the epoxy resin surface. Besides, by simulating decomposition process of epoxy resin, the generation path of each gas was revealed.

Materials
In the present study, the bisphenol-A epoxy resin E51 (DGEBA) was purchased from Wuxi Lan-Star Petrochemical Co., Ltd. (Wuxi, China). Epoxy resin was cured by an amine curing agent consisting of an adduct of diethylenetriamine and butyl glycidyl ether, known as 593 curing agent, supplied by Wuhan Shen Chemical Reagents and Equipments Co., Ltd. (Wuhan, China). High-precision SF 6 gas with grade of purity of 99.999% was supplied by Wuhan Newradar Gas Co., Ltd. (Wuhan, China). Therefore, the original percentage content of O 2 and H 2 O in the gas was low enough to be omitted.

Experiment Platform
Among a multitude of documented electrical faults, flashover in the electrical equipment happens frequently. Today flashover fault can be detected easily, but it is still hard to explain the complicated mechanism behind it [5][6][7]. Flashover discharge emphasizes on electrical breakdown of gap happening Materials 2019, 12, 1408 3 of 15 on solid material surface, such as flashover discharge on basin spacer surface in gas-insulated electrical equipment, where flashover voltage or flashover current was studied. Therefore, in our experiment, pin-plate electrodes on the epoxy resin surface was devised to investigate influence of surface breakdown of epoxy resin on SF 6 decomposition byproducts category and their concentration.
To create a severely uneven electrical field, pin-plate electrodes were used in this study, as shown in Figure 1. Given that the conducting rod in high-voltage cable is made of copper with better electric conductivity, the positive electrode in our study was a long-customized needle made of copper, the curvature radius of the needle electrode is 0.5 mm and beveled at 30 • on the sample surface. The plate electrode is a semicircle copper sheet with a diameter of 10 mm and is stuck to the surface of epoxy resin with a small amount of conductive silver paste to ensure that the electrical arc was close enough to the surface when the gap breaks down. The epoxy spacer sample is a square piece with a size of 50 mm × 50 mm × 2 mm. The pin-plate electrodes are positioned in a sealed cylinder made of polymethyl methacrylate (PMMA) as shown in Figure 2. The AC high voltage power can provide a maximum value of 50 kV. Concentration of decomposed components of SF 6 was measured on the gas chromatograph-mass spectrometer (GC-MS) and the type of GC-MS is Shimadzu QP-2010Ultra (Shimadzu Co., LTD, Kyoto, Japan).
To protect the voltage divider, the value of R2 in Figure 2 was usually below 200 Ω, and the value of R1 was usually above 3000 Ω to protect the transformer when voltage divider was totally destroyed. Because epoxy resin has a high electrical resistivity with an order of magnitude above 10 13 Ω cm, it is fully accepted that the break-down voltage on the sample surface equals to the highest shown value on the operating board when the gap was not broken down.
Because roughness of sample surface has a significant influence on flashover discharge property of epoxy resin, roughness of the samples was assessed on the Surface Profilometer 2300 supplied by Wale Electromechanical Technology Co., Ltd. (Xi'an, China). Roughness of all the epoxy resin samples had an average value of 0.35 µm in the range from 0.27 µm to 0.48 µm. Conventionally, roughness of solid material surface should be kept below 1 µm in order to ensure the AC (Alternating Current) arc creeping closely on the sample surface. in our experiment, pin-plate electrodes on the epoxy resin surface was devised to investigate influence of surface breakdown of epoxy resin on SF6 decomposition byproducts category and their concentration.
To create a severely uneven electrical field, pin-plate electrodes were used in this study, as shown in Figure 1. Given that the conducting rod in high-voltage cable is made of copper with better electric conductivity, the positive electrode in our study was a long-customized needle made of copper, the curvature radius of the needle electrode is 0.5 mm and beveled at 30° on the sample surface. The plate electrode is a semicircle copper sheet with a diameter of 10 mm and is stuck to the surface of epoxy resin with a small amount of conductive silver paste to ensure that the electrical arc was close enough to the surface when the gap breaks down. The epoxy spacer sample is a square piece with a size of 50 mm × 50 mm × 2 mm. The pin-plate electrodes are positioned in a sealed cylinder made of polymethyl methacrylate (PMMA) as shown in Figure 2. The AC high voltage power can provide a maximum value of 50 kV. Concentration of decomposed components of SF6 was measured on the gas chromatograph-mass spectrometer (GC-MS) and the type of GC-MS is Shimadzu QP-2010Ultra (Shimadzu Co., LTD, Kyoto, Japan).
To protect the voltage divider, the value of R2 in Figure 2 was usually below 200 Ω, and the value of R1 was usually above 3000 Ω to protect the transformer when voltage divider was totally destroyed. Because epoxy resin has a high electrical resistivity with an order of magnitude above 10 13 Ω cm, it is fully accepted that the break-down voltage on the sample surface equals to the highest shown value on the operating board when the gap was not broken down.
Because roughness of sample surface has a significant influence on flashover discharge property of epoxy resin, roughness of the samples was assessed on the Surface Profilometer 2300 supplied by Wale Electromechanical Technology Co., Ltd. (Xi'an, China). Roughness of all the epoxy resin samples had an average value of 0.35 μm in the range from 0.27 μm to 0.48 μm. Conventionally, roughness of solid material surface should be kept below 1 μm in order to ensure the AC (Alternating Current) arc creeping closely on the sample surface.

Experiment Procedures
The specific experimental procedures are listed as follows: First, all experiment materials were cleaned with anhydrous ethanol to get rid of dust and other impurities that may adhere to the experiment materials. After being volatilized with anhydrous ethanol, the epoxy spacer and electrodes were placed in the fixed position in the container. The distance between the pin-plate electrodes was set at 5, 6, 7, 8 and 9 mm, respectively. Then, the container was sealed.
Second, the chamber was evacuated to 0.01 MPa and filled with fresh SF6 until the pressure reached 0.2 MPa. It needed to take three times to remove most of water vapor and air. Subsequently, 100 mL of gas was extracted as the initial reference.
Finally, voltage was imposed slowly at the rate of 2 kV/30 s until the gap between the electrodes broke down, then the discharge voltage was recorded. After the decomposed gases diffused for 3 min in the chamber, 100 mL of gas was extracted from the chamber and then injected into the GC-MS. Twenty-one flashover discharge times were conducted for each gap distance in order to find the basic growth law of SF6 decomposition byproducts. As we all know, the surface flashover does significant harm to the dielectric property of epoxy resin. As the discharge number increased, the flashover voltage decreased slowly. The small decomposition, or the carbonization of the epoxy resin surface can cause obvious characteristic gases concentration change [23]. Therefore, 21 times of flashover discharges are enough to make significant harm to dielectric properties of epoxy resin and create more precise concentration variation. Figure 3 shows initial flashover discharge voltages vs. discharge times at different pin-plate electrodes distances. It is apparent that discharge voltage decreased rapidly in the first three times flashover discharges. Then, discharge voltage fluctuated slowly in a downward trend on the whole. After the breakdown of the gap between the electrodes, roughness of the epoxy resin sample surface increased with an obvious carbonization trace, which would greatly reduce flashover discharge voltage under the uneven electrical field. At the same time, water vapor generating during the decomposition of epoxy resin could also enhance the conductivity of carbonized trace on the surface [24]. As a result, the flashover discharge voltage decreased in the first three times. However, the roughness of the sample surface did not markedly change after being discharged four to six times, and water vapor maintained at a relatively stable level because it was exhausted quickly after it was generated. Hence, the flashover discharge voltage fluctuated in a downward trend.

Experiment Procedures
The specific experimental procedures are listed as follows: First, all experiment materials were cleaned with anhydrous ethanol to get rid of dust and other impurities that may adhere to the experiment materials. After being volatilized with anhydrous ethanol, the epoxy spacer and electrodes were placed in the fixed position in the container. The distance between the pin-plate electrodes was set at 5, 6, 7, 8 and 9 mm, respectively. Then, the container was sealed.
Second, the chamber was evacuated to 0.01 MPa and filled with fresh SF 6 until the pressure reached 0.2 MPa. It needed to take three times to remove most of water vapor and air. Subsequently, 100 mL of gas was extracted as the initial reference.
Finally, voltage was imposed slowly at the rate of 2 kV/30 s until the gap between the electrodes broke down, then the discharge voltage was recorded. After the decomposed gases diffused for 3 min in the chamber, 100 mL of gas was extracted from the chamber and then injected into the GC-MS. Twenty-one flashover discharge times were conducted for each gap distance in order to find the basic growth law of SF 6 decomposition byproducts. As we all know, the surface flashover does significant harm to the dielectric property of epoxy resin. As the discharge number increased, the flashover voltage decreased slowly. The small decomposition, or the carbonization of the epoxy resin surface can cause obvious characteristic gases concentration change [23]. Therefore, 21 times of flashover discharges are enough to make significant harm to dielectric properties of epoxy resin and create more precise concentration variation. Figure 3 shows initial flashover discharge voltages vs. discharge times at different pin-plate electrodes distances. It is apparent that discharge voltage decreased rapidly in the first three times flashover discharges. Then, discharge voltage fluctuated slowly in a downward trend on the whole. After the breakdown of the gap between the electrodes, roughness of the epoxy resin sample surface increased with an obvious carbonization trace, which would greatly reduce flashover discharge voltage under the uneven electrical field. At the same time, water vapor generating during the decomposition of epoxy resin could also enhance the conductivity of carbonized trace on the surface [24]. As a result, the flashover discharge voltage decreased in the first three times. However, the roughness of the sample surface did not markedly change after being discharged four to six times, and water vapor maintained at a relatively stable level because it was exhausted quickly after it was generated. Hence, the flashover discharge voltage fluctuated in a downward trend.  A longer distance between the electrodes could cause a higher discharge voltage in the first three times, but longer distances cannot guarantee a higher discharge voltage as the discharge continued. In Figure 3, by keeping the distance of pin-plate electrodes at 9 mm, the flashover voltage increased slowly from 10 to 15 times flashover discharges, which can be attributed to the randomness of flashover discharge and surface structure change. A longer gap distance led to greater dispersion of discharge voltage, creating more arc discharge channels among which the actual surface creepage distance was longer than 9 mm ( Figure 4). In Figure 4a, the flashover distance was obviously larger than 9 mm, while the creepage distances in Figure 4b,c were a little shorter. Mostly, the creepage distances were all above 9 mm in which the surface structure played an important role in the development direction of electrical tree. Compared with discharging at shorter gap distances, the electrical tree may develop in more directions at the gap distance of 9 mm, leading to more fluctuating flashover discharge voltages. The surface roughness of the epoxy resin samples before flashover discharge had an average value of 0.35 μm. After 21 times discharges, there generated several visible carbonized traces, so the roughness around the traces on the samples were measured to assess the decomposition extent of epoxy resin at different gap distances. The average roughness of the epoxy resin samples after 21 times discharges was shown in Table 1. It was reported that with the accumulation of aging energy on the material surface, the particles formed on the material surface were increased both in number and size, leading to the growth of A longer distance between the electrodes could cause a higher discharge voltage in the first three times, but longer distances cannot guarantee a higher discharge voltage as the discharge continued. In Figure 3, by keeping the distance of pin-plate electrodes at 9 mm, the flashover voltage increased slowly from 10 to 15 times flashover discharges, which can be attributed to the randomness of flashover discharge and surface structure change. A longer gap distance led to greater dispersion of discharge voltage, creating more arc discharge channels among which the actual surface creepage distance was longer than 9 mm (Figure 4). In Figure 4a, the flashover distance was obviously larger than 9 mm, while the creepage distances in Figure 4b,c were a little shorter. Mostly, the creepage distances were all above 9 mm in which the surface structure played an important role in the development direction of electrical tree. Compared with discharging at shorter gap distances, the electrical tree may develop in more directions at the gap distance of 9 mm, leading to more fluctuating flashover discharge voltages.  A longer distance between the electrodes could cause a higher discharge voltage in the first three times, but longer distances cannot guarantee a higher discharge voltage as the discharge continued. In Figure 3, by keeping the distance of pin-plate electrodes at 9 mm, the flashover voltage increased slowly from 10 to 15 times flashover discharges, which can be attributed to the randomness of flashover discharge and surface structure change. A longer gap distance led to greater dispersion of discharge voltage, creating more arc discharge channels among which the actual surface creepage distance was longer than 9 mm (Figure 4). In Figure 4a, the flashover distance was obviously larger than 9 mm, while the creepage distances in Figure 4b,c were a little shorter. Mostly, the creepage distances were all above 9 mm in which the surface structure played an important role in the development direction of electrical tree. Compared with discharging at shorter gap distances, the electrical tree may develop in more directions at the gap distance of 9 mm, leading to more fluctuating flashover discharge voltages. The surface roughness of the epoxy resin samples before flashover discharge had an average value of 0.35 μm. After 21 times discharges, there generated several visible carbonized traces, so the roughness around the traces on the samples were measured to assess the decomposition extent of epoxy resin at different gap distances. The average roughness of the epoxy resin samples after 21 times discharges was shown in Table 1. It was reported that with the accumulation of aging energy on the material surface, the particles formed on the material surface were increased both in number and size, leading to the growth of The surface roughness of the epoxy resin samples before flashover discharge had an average value of 0.35 µm. After 21 times discharges, there generated several visible carbonized traces, so the roughness around the traces on the samples were measured to assess the decomposition extent of epoxy resin at different gap distances. The average roughness of the epoxy resin samples after 21 times discharges was shown in Table 1. It was reported that with the accumulation of aging energy on the material surface, the particles formed on the material surface were increased both in number and size, leading to the growth of surface roughness; and, the break of the molecular chains of epoxy resin on the surface resulted in oxidation and carbonization [25]. Increased roughness can also result in larger fluctuation of break-down voltages. Taken together, average roughness should be kept in a reasonable range for ensuring the arc creeping to be closely enough to the surface. Figure 5 shows the concentration of the seven characteristic gases (CF 4 , CO 2 , SO 2 F 2 , SOF 2 , H 2 S, SO 2 , and CS 2 ) vs. discharge times at different pin-plate electrodes distances. The results showed a tendency that all gases concentration increased steadily with the increase of flashover discharge times. It is obvious that the concentration of SOF 2 and CF 4 was the highest while the concentration of other gases was relatively lower, especially SO 2 F 2 , CS 2 , and H 2 S. Taken together, the concentration of the seven characteristic gases follows the following order: surface roughness; and, the break of the molecular chains of epoxy resin on the surface resulted in oxidation and carbonization [25]. Increased roughness can also result in larger fluctuation of breakdown voltages. Taken together, average roughness should be kept in a reasonable range for ensuring the arc creeping to be closely enough to the surface. Figure 5 shows the concentration of the seven characteristic gases (CF4, CO2, SO2F2, SOF2, H2S, SO2, and CS2) vs. discharge times at different pin-plate electrodes distances. The results showed a tendency that all gases concentration increased steadily with the increase of flashover discharge times. It is obvious that the concentration of SOF2 and CF4 was the highest while the concentration of other gases was relatively lower, especially SO2F2, CS2, and H2S. Taken together, the concentration of the seven characteristic gases follows the following order: SOF2 > CF4> SO2 > CO2 > SO2F2 >CS2 > H2S.  At the pin-plate electrodes distance of 5 mm, after flashover discharge six times, the concentration of SOF 2 rose above 50 ppm, a value that can be detected easily. At the pin-plate electrodes distance of 9 mm, after flashover discharge six times, the concentration of SOF 2 rose closely to 300 ppm. Thus, longer pin-plate electrodes distance may facilitate the generation of characteristic gas. After flashover discharge 15 times, the SOF 2 concentration was almost twice the CF 4 concentration, whereas other gas byproducts increased at slow rates.

Concentration Variation of Seven Characteristic Gases
3.2.1. Analysis of SOF 2 and SO 2 F 2 The concentration of SOF 2 and SO 2 F 2 vs. flashover discharge times at different discharge distance were shown in Figure 6. The data indicate that SOF 2 and SO 2 F 2 generated steadily as the discharge times increased, and the concentration of SOF 2 was nearly 100 times as high as that of SO 2 F 2 . After 6 times flashover discharge, SOF 2 reached almost 300 ppm, while SO 2 F 2 was still below 3 ppm. The concentration of SO 2 F 2 did not reach 10 ppm before 21 discharge times, however, the SOF 2 concentration exceeded 1000 ppm at this discharge time point. It was also found that as the discharge distance increased, the content of both SOF 2 and SO 2 F 2 elevated.
Dominant reaction pathways for SOF 2 and SO 2 F 2 were described in Equations (1)-(5). As reported previously, when partial discharge happened in SF 6 atmosphere involving no solid insulation materials, SOF 2 and SO 2 F 2 concentration increased at similar rates [13,26]. However, in our experiment, the SO 2 F 2 concentration remained far below the SOF 2 concentration. Therefore, flashover discharge happening on epoxy resin may cause a different chemical reaction pathway to facilitate SOF 2 generation. At the pin-plate electrodes distance of 5 mm, after flashover discharge six times, the concentration of SOF2 rose above 50 ppm, a value that can be detected easily. At the pin-plate electrodes distance of 9 mm, after flashover discharge six times, the concentration of SOF2 rose closely to 300 ppm. Thus, longer pin-plate electrodes distance may facilitate the generation of characteristic gas. After flashover discharge 15 times, the SOF2 concentration was almost twice the CF4 concentration, whereas other gas byproducts increased at slow rates.

Analysis of SOF2 and SO2F2
The concentration of SOF2 and SO2F2 vs. flashover discharge times at different discharge distance were shown in Figure 6. The data indicate that SOF2 and SO2F2 generated steadily as the discharge times increased, and the concentration of SOF2 was nearly 100 times as high as that of SO2F2. After 6 times flashover discharge, SOF2 reached almost 300 ppm, while SO2F2 was still below 3 ppm. The concentration of SO2F2 did not reach 10 ppm before 21 discharge times, however, the SOF2 concentration exceeded 1000 ppm at this discharge time point. It was also found that as the discharge distance increased, the content of both SOF2 and SO2F2 elevated.
Dominant reaction pathways for SOF2 and SO2F2 were described in Equations (1)-(5). As reported previously, when partial discharge happened in SF6 atmosphere involving no solid insulation materials, SOF2 and SO2F2 concentration increased at similar rates [13,26]. However, in our experiment, the SO2F2 concentration remained far below the SOF2 concentration. Therefore, flashover discharge happening on epoxy resin may cause a different chemical reaction pathway to facilitate SOF2 generation. Actually, under high-energy electron collision, electron-impact-induced dissociation of SF6 can lead to fluoride sulfide generation such as SF5, SF4, SF3, SF2, SF, and so on. It was reported that the amount of SF4 and SF2 was larger than that of other gas byproducts [26,27]. When SF4 comes into contact with H2O, SOF2 will soon form, as illustrated in Equation (2). SF2 can also be oxidized easily into SOF2 because S atoms of +2 valence can be turned to S atoms of +4 valence during reaction with Actually, under high-energy electron collision, electron-impact-induced dissociation of SF 6 can lead to fluoride sulfide generation such as SF 5 , SF 4 , SF 3 , SF 2 , SF, and so on. It was reported that the amount of SF 4 and SF 2 was larger than that of other gas byproducts [26,27]. When SF 4 comes into contact with H 2 O, SOF 2 will soon form, as illustrated in Equation (2). SF 2 can also be oxidized easily into SOF 2 because S atoms of +2 valence can be turned to S atoms of +4 valence during reaction with high-active O atoms from arc discharge as shown in Equation (1). Therefore, it is reasonable to mark it as an indicator for flashover discharge fault on solid insulation surface.
Generation of SO 2 F 2 is more complicated than that of SOF 2 . From the chemical perspective, the S atom in SOF 2 molecule is +4 valence, that means the unsaturated chemical valence needs further oxidation to turn into SO 2 F 2 that has an S atom of +6 valence. On the other hand, SF 4 can be oxidized into SOF 4 , then SOF 4 contacts with H 2 O to generate SO 2 F 2 molecules, which is the main generation resource of SO 2 F 2 , as indicated in Equations (4) and (5) [28].
To sum up, SOF 2 was a generated from the reaction between SF 4 and H 2 O, while SO 2 F 2 mainly comes from the reaction between SOF 4 and H 2 O. SOF 4 needs further oxidization of SF 4 by activating O atoms. So the generation of SOF 2 is easier than that of SO 2 F 2 . In our study, another characteristic gas CO 2 was generated and its amount increased steadily with the increase of flashover discharge times. The generation of CO 2 can consume a large number of active O atoms, so the SO 2 F 2 generation became more difficult under flashover discharge happening on the epoxy resin surface, compared with partial discharge in the SF 6 atmosphere involving no epoxy resin. As a result, the production of SO 2 F 2 increased at a quite low rate with flashover discharge times increasing.
Without reckoning the effect of impurities such as H 2 O or O 2 in the original SF 6 , epoxy resin decomposition may produce a small amount of water under flashover discharge, which acts as a main factor for Equation (2). Furthermore, H 2 O would decompose and release O 2 under imposition of high-current, so Equation (1) also contributes to the generation of SOF 2 . Hence, the concentration of SOF 2 rose rapidly as discharge continued. Figure 7 shows the concentration variation of SO 2 vs. discharge times at different pin-plate electrodes distances. At the beginning, the amount of SO 2 was quite small, around 5 ppm after flashover discharge six times. Later, SO 2 was generated steadily as discharge continued, and its concentration reached to about 30 ppm and 55 ppm after discharging 12 times and 21 times, respectively. In addition, the concentration of SO 2 also elevated with the discharge distance increasing.

Analysis of SO 2
S + 2O→SO 2 (6) SOF 2 + H 2 O→SO 2 + 2HF (7) SF + OH→SO + HF (8) SO + O→SO 2 (9) To better explain the concentration variation of SO 2 , the generation path of SO 2 is depicted in Equations (6)- (9) [29]. It can be seen that the production of SO 2 needs oxygen or water. It is extensively accepted that F atom has stronger reducibility than O atom, but the structure of SO 2 (O=S=O) is more symmetrical than that of SOF 2 . Therefore, it is also easy for the generation of SO 2 .
During the reactions, the amount of H 2 O was limited, and most of H 2 O reacted with subfluorides to form SOF 2 , so there is less amount of H 2 O for the generation of SO 2 in Equation (7). Besides, there were very few S atoms provided for the Equation (6) reaction, since S atom is not the main decomposition byproduct of SF 6 . As a result, the content of SO 2 was quite low at the beginning. As discharge continued, S and SF accumulated and had more chances to react with O 2 and H 2 O. Thus, after about discharging nine times, the SO 2 concentration reached a high level and increased steadily.
In conclusion, SO 2 began to generate after a certain amount of SF 6 decomposed into S or SF, indicating that SF 6 has already lost its original dielectric property. In order to take SO 2 into consideration for the flashover discharge fault diagnosis method, it is better to take the total concentration of SOF 2 and SO 2 as an indicator for flashover discharge fault.  Figure 8 shows the concentration variation of CO2 and CF4 vs. discharge times under different distances. Both of them were produced once the flashover discharge occurred. After discharging six times, the concentration of CO2 reached about 10 ppm, while CF4 exceeded 50 ppm. After discharging 21 times, CF4 concentration reached over 400 ppm, yet CO2 concentration was still below 90 ppm, which means that the amount of CF4 was nearly five times as large as that of CO2.

Analysis of CO2 and CF4
CF4 is an important gas byproduct involved in solid insulation materials decomposition defects. When flashover discharge occurred on epoxy resin, methyl fragments (CHx) were produced. Through substitution reaction with free F atoms, CHx would be turned into CF4. Flashover discharge provided enough energy for the formation of CF4, whereas CO2 mainly came from intrinsic decomposition of epoxy resin. So the concentration of CF4 was much higher than that of CO2.
In conclusion, CF4 has a close relationship with epoxy resin decomposition, which can indicate the occurrence of flashover discharge fault on epoxy resin dielectrics in SF6 insulation equipment. Therefore, the concentration of CF4 should be monitored intensely.  Figure 9 shows the concentration changes of CS2. It can be seen that the content of CS2 did not reach 3 ppm until discharging 10-15 times. As the discharge times increased, the CS2 amount grew slowly. After discharging 21 times, the maximum concentration of CS2 was still less than 6 ppm.  Figure 8 shows the concentration variation of CO 2 and CF 4 vs. discharge times under different distances. Both of them were produced once the flashover discharge occurred. After discharging six times, the concentration of CO 2 reached about 10 ppm, while CF 4 exceeded 50 ppm. After discharging 21 times, CF 4 concentration reached over 400 ppm, yet CO 2 concentration was still below 90 ppm, which means that the amount of CF 4 was nearly five times as large as that of CO 2 . CF 4 is an important gas byproduct involved in solid insulation materials decomposition defects. When flashover discharge occurred on epoxy resin, methyl fragments (CHx) were produced. Through substitution reaction with free F atoms, CHx would be turned into CF 4 . Flashover discharge provided enough energy for the formation of CF 4 , whereas CO 2 mainly came from intrinsic decomposition of epoxy resin. So the concentration of CF 4 was much higher than that of CO 2 .

Analysis of CO 2 and CF 4
In conclusion, CF 4 has a close relationship with epoxy resin decomposition, which can indicate the occurrence of flashover discharge fault on epoxy resin dielectrics in SF 6 insulation equipment. Therefore, the concentration of CF 4 should be monitored intensely.  Figure 8 shows the concentration variation of CO2 and CF4 vs. discharge times under different distances. Both of them were produced once the flashover discharge occurred. After discharging six times, the concentration of CO2 reached about 10 ppm, while CF4 exceeded 50 ppm. After discharging 21 times, CF4 concentration reached over 400 ppm, yet CO2 concentration was still below 90 ppm, which means that the amount of CF4 was nearly five times as large as that of CO2.

Analysis of CO2 and CF4
CF4 is an important gas byproduct involved in solid insulation materials decomposition defects. When flashover discharge occurred on epoxy resin, methyl fragments (CHx) were produced. Through substitution reaction with free F atoms, CHx would be turned into CF4. Flashover discharge provided enough energy for the formation of CF4, whereas CO2 mainly came from intrinsic decomposition of epoxy resin. So the concentration of CF4 was much higher than that of CO2.
In conclusion, CF4 has a close relationship with epoxy resin decomposition, which can indicate the occurrence of flashover discharge fault on epoxy resin dielectrics in SF6 insulation equipment. Therefore, the concentration of CF4 should be monitored intensely.  Figure 9 shows the concentration changes of CS2. It can be seen that the content of CS2 did not reach 3 ppm until discharging 10-15 times. As the discharge times increased, the CS2 amount grew slowly. After discharging 21 times, the maximum concentration of CS2 was still less than 6 ppm. 3.2.4. Analysis of CS 2 Figure 9 shows the concentration changes of CS 2 . It can be seen that the content of CS 2 did not reach 3 ppm until discharging 10-15 times. As the discharge times increased, the CS 2 amount grew slowly. After discharging 21 times, the maximum concentration of CS 2 was still less than 6 ppm. The generation of CS2 is difficult because its generation needs active C atoms and S atoms [14,27,28]. On one hand, the amount of C atoms from the decomposition of epoxy resin was limited. The formation of SO2 also consumed some S atoms, so the remaining amount of S atoms from SF6 ionization was too low to facilitate the formation of CS2. C atoms mainly came from ionization of epoxy resin decomposition, and a large number of C atoms were required by the formation of CO2 and CF4, so the existing amount of CS2 was quite low. Besides, the 9 mm gap distance also contributed to the formation of CS2 to some extent.

Analysis of CS2
As one of the typical byproducts of flashover discharging on epoxy resin dielectrics, CS2 is too difficult to be generated in comparison with CF4. Therefore, the appearance of CS2 could indicate the severest stage of flashover discharge. Figure 10 shows the concentration variation of H2S vs. discharge times at different pin-plate electrodes distances. The results show that the concentration of H2S was less than 0.3 ppm after discharging nine times, and its concentration only reached 1 ppm after discharging 15 times. After discharging 21 times, its concentration was about 2.5 ppm. A conclusion can be drawn that, like CS2, the generation of H2S is as difficult as that of CS2 among the seven characteristic gases.

Analysis of H2S
The generation of H2S requires active S atoms and free H atoms [14,28,29]. The S atom is not the main decomposition byproduct of SF6 [26], and the rising concentration of SO2 could also reduce the reaction between S atoms and free H atoms. Therefore, the amount of H2S was extremely small. Although the longer distance between the electrodes can facilitate the generation of H2S, its amount still stayed at a very low level.
In conclusion, the generation of H2S can also be used as an indicator to judge the serious stage of flashover discharge faults on epoxy resin. The generation of CS 2 is difficult because its generation needs active C atoms and S atoms [14,27,28]. On one hand, the amount of C atoms from the decomposition of epoxy resin was limited. The formation of SO 2 also consumed some S atoms, so the remaining amount of S atoms from SF 6 ionization was too low to facilitate the formation of CS 2 . C atoms mainly came from ionization of epoxy resin decomposition, and a large number of C atoms were required by the formation of CO 2 and CF 4 , so the existing amount of CS 2 was quite low. Besides, the 9 mm gap distance also contributed to the formation of CS 2 to some extent.
As one of the typical byproducts of flashover discharging on epoxy resin dielectrics, CS 2 is too difficult to be generated in comparison with CF 4 . Therefore, the appearance of CS 2 could indicate the severest stage of flashover discharge. Figure 10 shows the concentration variation of H 2 S vs. discharge times at different pin-plate electrodes distances. The results show that the concentration of H 2 S was less than 0.3 ppm after discharging nine times, and its concentration only reached 1 ppm after discharging 15 times. After discharging 21 times, its concentration was about 2.5 ppm. A conclusion can be drawn that, like CS2, the generation of H 2 S is as difficult as that of CS 2 among the seven characteristic gases.

Analysis of H 2 S
The generation of H 2 S requires active S atoms and free H atoms [14,28,29]. The S atom is not the main decomposition byproduct of SF 6 [26], and the rising concentration of SO 2 could also reduce the reaction between S atoms and free H atoms. Therefore, the amount of H 2 S was extremely small. Although the longer distance between the electrodes can facilitate the generation of H 2 S, its amount still stayed at a very low level.

Basic Rules for Judging the Happening of Flashover Discharge
Based on the concentration variation of seven different characteristic gases above, it can be seen that SOF2 and CF4 had the highest concentration far beyond other gas byproducts, both of which can In conclusion, the generation of H 2 S can also be used as an indicator to judge the serious stage of flashover discharge faults on epoxy resin.

Basic Rules for Judging the Happening of Flashover Discharge
Based on the concentration variation of seven different characteristic gases above, it can be seen that SOF 2 and CF 4 had the highest concentration far beyond other gas byproducts, both of which can therefore represent the main decomposition component resulting from flashover discharge happening on the epoxy resin surface. Compared with the previous studies focusing on partial discharge in SF 6 insulation equipment [1,14,17,24,29], flashover discharge on the epoxy resin surface could cause a drastically elevation in the concentration of SOF 2 and CF 4 , but generated a relatively lower concentration for H 2 S. According to these two statistics' data, a preliminary rule can be summarized as follows: c SOF 2 > 2c CF 4 and c SOF 2 + c CF 4 > 400c H 2 S (10) among which, c represents concentration for the selected gases. That is to say, when concentration of SOF 2 reaches twice of that of CF 4 , and the sum concentration of both SOF 2 and CF 4 is much higher than that of H 2 S, a possible flashover discharge on the epoxy resin surface in SF 6 -infused electrical equipment occurs.

Simulation for Decomposition of Epoxy Resin under High Energy
To verify whether H 2 O mainly comes from decomposition of epoxy resin, simulation methods were employed to further investigate decomposition of epoxy resin under high energy. For the sake of simplification, the energy for epoxy resin decomposition was supplied via temperature setting. The whole simulation was conducted with the ReaxFF program and NVT ensemble. NVT represents the three parameters (number (N), volume (V), and temperature (T)) that are fixed parameters, and the other two parameters (pressure (P) and energy (E)) are variables.
Firstly, a ball-stick model of the curing agent molecule was given in Figure 11. After curing, a bisphenol-A epoxy resin (DGEBA) molecule with polymerization degree of 0 was built, as shown in Figure 12. Then, a periodic unit cell of epoxy resin consisting of 20 epoxy resin molecules was constructed with an initial density of 0.5 g/cm 3 in a 23.2 Å × 23.2 Å × 23.2 Å cubic box, as shown in Figure 13. The structure was treated by annealing process at the cooling rate of 10 K/500 ps and Energy Minimization and Geometry Optimization were carried out before the decomposition simulation began. The final density of the epoxy resin model was 1.13 g/cm 3 .    The reaction temperature was set at 1300 K to ensure that energy was high enough for epoxy resin decomposition within a total simulation time of 1000 ps [30]. It is already known that the main small molecular products are CH2O, H2O, CO, CO2, CH4, and H2 [30]. Figure 14 shows the number change of fragments from the periodic unit cell of epoxy resin as a function of time. It can be seen that the number of H2O molecules was the highest among the 6 byproducts, followed by H2, CO2, CH2O, CO, and CH4. The dominant reaction pathways (RPWs) of H2O were shown in Figure 15. CH4 was mainly derived from demethylation of methyl-group-containing fragments, and H2 was mainly produced by free hydrogen derived from C-H bond. The reaction temperature was set at 1300 K to ensure that energy was high enough for epoxy resin decomposition within a total simulation time of 1000 ps [30]. It is already known that the main small molecular products are CH 2 O, H 2 O, CO, CO 2 , CH 4 , and H 2 [30]. Figure 14 shows the number change of fragments from the periodic unit cell of epoxy resin as a function of time. It can be seen that the number of H 2 O molecules was the highest among the 6 byproducts, followed by H 2 , CO 2 , CH 2 O, CO, and CH 4 . The dominant reaction pathways (RPWs) of H 2 O were shown in Figure 15. CH 4 was mainly derived from demethylation of methyl-group-containing fragments, and H 2 was mainly produced by free hydrogen derived from C-H bond.  In conclusion, epoxy resin could produce H2O, CH4, CO2 and H2 that would mainly affect the concentration of seven characteristic gases. H2O could facilitate the formation of SOF2 according to Equations (1)-(3), and CH4 could facilitate the formation of CF4 through substitution reaction under high-energy arc discharge. CF4 decomposed from SF6 can react with H2O to form SOF2, leading to significantly higher concentration of SOF2 than other decomposition byproducts. Although the theoretical amount of CH4 from decomposition of epoxy resin unit cell was relatively lower, F atoms  In conclusion, epoxy resin could produce H2O, CH4, CO2 and H2 that would mainly affect the concentration of seven characteristic gases. H2O could facilitate the formation of SOF2 according to Equations (1)-(3), and CH4 could facilitate the formation of CF4 through substitution reaction under high-energy arc discharge. CF4 decomposed from SF6 can react with H2O to form SOF2, leading to significantly higher concentration of SOF2 than other decomposition byproducts. Although the theoretical amount of CH4 from decomposition of epoxy resin unit cell was relatively lower, F atoms during ionization of SF6 can help intensify the chemical reaction changing CH4 to CF4 under arc In conclusion, epoxy resin could produce H 2 O, CH 4 , CO 2 and H 2 that would mainly affect the concentration of seven characteristic gases. H 2 O could facilitate the formation of SOF 2 according to Equations (1)-(3), and CH 4 could facilitate the formation of CF 4 through substitution reaction under high-energy arc discharge. CF 4 decomposed from SF 6 can react with H 2 O to form SOF 2 , leading to significantly higher concentration of SOF 2 than other decomposition byproducts. Although the theoretical amount of CH 4 from decomposition of epoxy resin unit cell was relatively lower, F atoms during ionization of SF 6 can help intensify the chemical reaction changing CH 4 to CF 4 under arc discharge; in addition, the C-F bond in the surface structure of fluorinated epoxy resin may also break to form CF 4 directly. Taken together, CF 4 had the second highest concentration during flashover discharge happening on the surface of epoxy resin.

Conclusions
In the present study, flashover discharge experiments were carried out on the epoxy resin surface in SF 6 different pin-plate electrodes distances. The concentration of seven characteristic gases including SOF 2 , SO 2 , CF 4 , CO 2 SO 2 F 2 , CS 2 , and H 2 S were measured. The following conclusions can be drawn: Initial flashover discharge voltage decreased sharply in the first three times of flashover discharge, and the discharge voltage fluctuated in a downward tendency.
The concentration of the seven characteristic gases elevated with the flashover discharge times increasing, and the final concentration from high to low follows the order: SOF 2 > CF 4 > SO 2 > CO 2 > SO 2 F 2 > CS 2 > H 2 S. Among these gases, the concentration of SOF 2 and CF 4 was obviously higher than that of other gases. In addition, the concentration of SO 2 F 2 , CS 2 , and H 2 S kept below 10 ppm after discharging 21 times. Based on the concentration variation of seven characteristic gases, a preliminary rule can be proposed: When the concentration of SOF 2 reaches two times of that of CF 4 , and the sum concentration of both SOF 2 and CF 4 is 400 times higher than that of H 2 S, a possible flashover discharge on the epoxy resin surface in SF 6 -infused electrical equipment occurs.
The simulation for decomposition of epoxy resin on the ReaxFF program showed that epoxy resin could produce H 2 O, CH 4 , CO 2 , and H 2 that would drastically affect the concentration of the seven characteristic gases. More specifically, H 2 O could facilitate the formation of SOF 2 , and CH 4 could react with subfluorides to form CF 4 under high-energy arc discharge.