Experimental Study of the Corona Performance of Aged Sand-Cast Substation Connectors

.G.-P.); Abstract: Substation connectors, like many other high-voltage products, are tested once manufactured. However, the corona behavior of aged specimens can di ﬀ er from that exhibited by newer ones, thus generating detrimental technical and environmental e ﬀ ects. Manufacturers need to know the long-term corona behavior of substation connectors to o ﬀ er their customers maximum safety and transparency about such critical products. This paper analyzes the ageing e ﬀ ect on the surface roughness and the visual corona extinction voltage of sand-cast aluminum connectors, which were artiﬁcially aged in a salt spray chamber for di ﬀ erent periods. The experimental results show an important variability of the surface roughness and corona extinction voltage (CEV) among connectors with the same ageing level due to the sand-casting manufacturing process. The results in this paper also show a slight increase in the surface roughness with the ageing period, although there is not a clear pattern between the applied ageing level and the experimental CEV value. It is concluded that the inherent variability among connectors due to the sand-cast process has more inﬂuence than the ageing e ﬀ ect itself.


Introduction
Corona is a partial discharge in air that occurs under a non-uniform electric field [1] close to the areas of highest electrical stress of high-voltage electrodes such as conductors or connectors when the electric field strength exceeds a threshold value [2]. It is already known that the generation of corona depends on several causes, including electrode geometry, spatial factors, and surface contamination. Therefore, important parameters influencing corona appearance are the geometrical radius of the surface [3,4], as well as surface condition and roughness [5]. Small radii geometries and configurations with sharp points or edges are more prone to corona occurrence due to the concentration of the electric field lines [6]. Partial discharges and corona generate undesirable effects, such as power losses, acoustic noise, radio interference voltage, very high frequency (VHF) and ultra high frequency (UHF) radiation [7] and chemical reactions. Corona activity also generates electromagnetic waves in the ultraviolet (UV) and visible spectrum, specifically in the blue range [8].
Conductors and bundle conductors for overhead power lines are operated at high electrical stress, although this is restricted to limit corona losses. The surface electric field strength is limited to 16 kV RMS /cm for lines rated above 110 kV [9,10], 17.4-20 kV RMS /cm for 400 kV AC power lines [11], although in ± 800 kV DC power lines it can be around 24 kV/cm [12]. In [13], it was proven that sand and dust can increase corona activity. It is known that the surface condition can impact corona activity This section describes the T-type electrical connectors analyzed in this work. T-type connectors connect two stranded aluminum conductors together, as shown in Figure 1. The analyzed connectors are from the catalogue of SBI Connectors (SBI Connectors, Sant Esteve Sesrovires, Spain), and are made of hypo-eutectic sand-cast A356 aluminum alloy. The substation connectors analyzed in this work were aged according to the specification of the ISO 9227-2017 standard, Corrosion tests in artificial atmospheres-Salt spray tests, which specifies the procedure, reagents and apparatus required to conduct neutral salt spray tests to assess the resistance-to-corrosion of metallic materials [25].
The salt solution was prepared in accordance with the ISO 9227-2017 standard using low conductivity distilled water (< 20 μS/cm at 25 ± 2 °C) to produce a concentration of sodium chloride in the range of 45-55 g/L, as specified in the ASTM B117-19 salt spray test standard [26]. It was carried out by dissolving approximately 5 parts by mass of NaCl into 95 parts of distilled water. Salt fog collection rates were measured to ensure they were within the standard range of 1-2 ml/hr/80cm 2 . The chamber temperature was adjusted at a constant value of 35 ± 2 °C.
A salt spray corrosion chamber (CCI manufacturer, CCA model, 400 L inner volume, fulfilling the requirements of the ASTM B117-19 standard) was used to conduct the ageing tests.
A total of 24 T-type substation connectors were analyzed, 12 of the J line and the remaining 12 of the S line, which were aged according to 4 ageing levels, as detailed in Table 1. Before being introduced into the chamber, the connectors were cleaned using a clean soft brush and distilled water. After the cleaning, they were rinsed with clean distilled water and dried. Next, the test specimens were vertically placed inside the salt spray chamber, as shown in Figure 2b, with The substation connectors analyzed in this work were aged according to the specification of the ISO 9227-2017 standard, Corrosion tests in artificial atmospheres-Salt spray tests, which specifies the procedure, reagents and apparatus required to conduct neutral salt spray tests to assess the resistance-to-corrosion of metallic materials [25].
The salt solution was prepared in accordance with the ISO 9227-2017 standard using low conductivity distilled water (<20 µS/cm at 25 ± 2 • C) to produce a concentration of sodium chloride in the range of 45-55 g/L, as specified in the ASTM B117-19 salt spray test standard [26]. It was carried out by dissolving approximately 5 parts by mass of NaCl into 95 parts of distilled water. Salt fog collection rates were measured to ensure they were within the standard range of 1-2 mL/hr/80 cm 2 . The chamber temperature was adjusted at a constant value of 35 ± 2 • C.
A salt spray corrosion chamber (CCI manufacturer, CCA model, 400 L inner volume, fulfilling the requirements of the ASTM B117-19 standard) was used to conduct the ageing tests.
A total of 24 T-type substation connectors were analyzed, 12 of the J line and the remaining 12 of the S line, which were aged according to 4 ageing levels, as detailed in Table 1. Before being introduced into the chamber, the connectors were cleaned using a clean soft brush and distilled water. After the cleaning, they were rinsed with clean distilled water and dried. Next, the test specimens were vertically placed inside the salt spray chamber, as shown in Figure 2b, with care taken to not to place them in the direct travel direction of the spray from the nozzles of the atomizer.  After conducting the ageing tests using the salt spray corrosion chamber, the connectors were removed from the chamber and cleaned with distilled water using a brush, in order to remove the superficial residues of spray solution from the surface of the connectors. Finally, the samples were dried immediately in a stream of air [25]. The application of a chemical cleaning to the connectors was avoided, as suggested in the ASTM B117-19 standard [26], in order to protect the structure of the After conducting the ageing tests using the salt spray corrosion chamber, the connectors were removed from the chamber and cleaned with distilled water using a brush, in order to remove the superficial residues of spray solution from the surface of the connectors. Finally, the samples were dried immediately in a stream of air [25]. The application of a chemical cleaning to the connectors was avoided, as suggested in the ASTM B117-19 standard [26], in order to protect the structure of the surface layer of the connectors. After this cleaning and drying procedure, the connectors were ready to be tested.

The Surface Roughness Measurements
The substations connectors analyzed in this work were manufactured by sand casting. They are made with A356 Al alloy with T6 heat treatment, whereas the accompanying all aluminum alloy conductors (AAAC) are made of 1350 Al alloy [27] and have a diameter of 15.8 mm. As the name indicates, sand casting uses a special kind of fine sand as the mold material. Sand casting is commonly applied in foundries for industrial application because it is a relatively cheap process. Sand casting is probably the most common process to produce substation connectors. However, as explained, sand-cast products present several disadvantages, such as high porosity compared to machined components, low dimensional accuracy due to the sand shrinking, and poor surface finish because of the sand texture. Therefore, sand-cast connectors tend to present defects such as porosity, shrinkage, surface and pouring metal defects. It is well known that surface finish can impact both surface roughness and corona performance. Sand-casting processes lead to surface roughness in the order of 5.3-50 µm [19].
Surface roughness measurements were carried out by means of a diamond tip stylus profiler with tip radius of 2 µm, a measuring force of 0.75 mN and a measuring range along the X axis of 17.5 mm (surface roughness tester Mitutoyo Surftest 211; Kawasaki, Japan) using a roughness calibration specimen with R a = 3 µm. The Mitutoyo Surftest 211 stylus profiler allows direct and fast measurements of the R a , R q and R z parameters in accordance with the EN-ISO 4287 standard [28]. During measurements, the stylus profiler tip raises and lowers from an initial reference level because of the valleys and peaks profile of the connector surface.
The following paragraphs describe the main parameters used in this work for characterizing surface roughness of the analyzed substation connectors.
Parameter R a [µm] is defined as the average value of the absolute roughness [27], where n is the number of measurements along the sampling length L, and z i is the height value of the roughness associated to the i-th point. Parameter R q [µm] is the root-mean-square value of the roughness, Finally, parameter R z [µm] is the mean maximum height of the measured profile, R pi and R vi being, respectively, the i-th highest peak, and the i-th deepest valley. Parameters R a , R q and R z were measured along four flat and straight paths marked in the bottom surface of the connector, which are highlighted in Figure 3. Afterwards, the mean values of the three parameters were obtained by averaging the values in all points of the four roughness profiles provided by the Mitutoyo Surftest 211 stylus profiler.
Energies 2020, 13, x FOR PEER REVIEW 5 of 13 surface layer of the connectors. After this cleaning and drying procedure, the connectors were ready to be tested.

The Surface Roughness Measurements
The substations connectors analyzed in this work were manufactured by sand casting. They are made with A356 Al alloy with T6 heat treatment, whereas the accompanying all aluminum alloy conductors (AAAC) are made of 1350 Al alloy [27] and have a diameter of 15.8 mm. As the name indicates, sand casting uses a special kind of fine sand as the mold material. Sand casting is commonly applied in foundries for industrial application because it is a relatively cheap process. Sand casting is probably the most common process to produce substation connectors. However, as explained, sandcast products present several disadvantages, such as high porosity compared to machined components, low dimensional accuracy due to the sand shrinking, and poor surface finish because of the sand texture. Therefore, sand-cast connectors tend to present defects such as porosity, shrinkage, surface and pouring metal defects. It is well known that surface finish can impact both surface roughness and corona performance. Sand-casting processes lead to surface roughness in the order of 5.3-50 μm [19].
Surface roughness measurements were carried out by means of a diamond tip stylus profiler with tip radius of 2 µ m, a measuring force of 0.75 mN and a measuring range along the X axis of 17.5 mm (surface roughness tester Mitutoyo Surftest 211; Kawasaki, Japan) using a roughness calibration specimen with Ra = 3 µ m. The Mitutoyo Surftest 211 stylus profiler allows direct and fast measurements of the Ra, Rq and Rz parameters in accordance with the EN-ISO 4287 standard [28]. During measurements, the stylus profiler tip raises and lowers from an initial reference level because of the valleys and peaks profile of the connector surface.
The following paragraphs describe the main parameters used in this work for characterizing surface roughness of the analyzed substation connectors.
Parameter Ra [μm] is defined as the average value of the absolute roughness [27], where n is the number of measurements along the sampling length L, and zi is the height value of the roughness associated to the i-th point.
Parameter Rq [μm] is the root-mean-square value of the roughness, (2) Finally, parameter Rz [μm] is the mean maximum height of the measured profile, Rpi and Rvi being, respectively, the i-th highest peak, and the i-th deepest valley. Parameters Ra, Rq and Rz were measured along four flat and straight paths marked in the bottom surface of the connector, which are highlighted in Figure 3. Afterwards, the mean values of the three parameters were obtained by averaging the values in all points of the four roughness profiles provided by the Mitutoyo Surftest 211 stylus profiler.

The Corona Tests in the High-Voltage Laboratory
This section describes the visual corona tests carried out in the AMBER high-voltage laboratory of the Universitat Politècnica de Catalunya (Terrassa, Spain). Similarly to what was undertaken in [20], experimental visual corona tests were carried out by means of a calibrated 130 kV high-voltage generator (BK-130 from Phenix technologies, Accident, Maryland USA), two calibrated 120 kV DC generators (4129-10 model from Phenix technologies, Accident, Maryland USA) with positive and negative outputs, respectively, and a digital camera (Canon EOS-70D resolution incorporating a 20.2 Mpixels CMOS APS-C sensor, 18-135 mm F3.5-5.6 IS STM lenses with super spectra coatings). Visual corona tests were performed in the laboratory in total darkness. Long exposure photographs were taken (60 seconds exposition, ISO-400 sensitivity, aperture f/5.6, tungsten color temperature, long exposure noise reduction filter activated in the camera menu) to improve the performance of the digital photographs in detecting the corona activity. In a previous study [29], it was demonstrated that digital photographs taken under the described conditions have similar sensitivity than that provided by a conventional partial discharge (PD) detector, with the advantage that visual corona photographs allow for the locating of the corona discharge points on the surface of the connectors.
Visual corona tests were performed according to the test procedure found in the IEEE 1829-2017 report [30]. The voltage was gradually increased starting from 0 kV at an approximate rate of 2 kV/s [31] until corona activity was identified on the test object, this being the visual corona inception voltage (CIV). Then, the voltage was increased by approximately 10% and maintained for 60 seconds. Next, the applied voltage was slowly reduced until corona extinction, this value being the corona extinction voltage (CEV). This process was repeated three times for each connector and type of applied voltage (AC, positive and negative DC), while completing a photographic report of each tested specimen.
Experimental corona tests were performed with the connectors placed at a reduced height above the metallic ground plane, as proposed in [20]. This approach allows performing the tests in a small-size high-voltage laboratory, thus drastically simplifying the time required to assemble the experimental setup, reducing the voltage applied and the cost of the high-voltage generators required, while ensuring corona occurrence. The bottom flat surface of the connectors was placed at a height of 225 mm above the ground plane in order to force the electric field to reach the CEV value when operating below 120 kV. The experimental layout to determine the CEV value is shown in Figure 4.

The Corona Tests in the High-Voltage Laboratory
This section describes the visual corona tests carried out in the AMBER high-voltage laboratory of the Universitat Politècnica de Catalunya (Terrassa, Spain). Similarly to what was undertaken in [20], experimental visual corona tests were carried out by means of a calibrated 130 kV high-voltage generator (BK-130 from Phenix technologies, Accident, Maryland USA), two calibrated 120 kV DC generators (4129-10 model from Phenix technologies, Accident, Maryland USA) with positive and negative outputs, respectively, and a digital camera (Canon EOS-70D resolution incorporating a 20.2 Mpixels CMOS APS-C sensor, 18-135 mm F3.5-5.6 IS STM lenses with super spectra coatings). Visual corona tests were performed in the laboratory in total darkness. Long exposure photographs were taken (60 seconds exposition, ISO-400 sensitivity, aperture f/5.6, tungsten color temperature, long exposure noise reduction filter activated in the camera menu) to improve the performance of the digital photographs in detecting the corona activity. In a previous study [29], it was demonstrated that digital photographs taken under the described conditions have similar sensitivity than that provided by a conventional partial discharge (PD) detector, with the advantage that visual corona photographs allow for the locating of the corona discharge points on the surface of the connectors.
Visual corona tests were performed according to the test procedure found in the IEEE 1829-2017 report [30]. The voltage was gradually increased starting from 0 kV at an approximate rate of 2 kV/s [31] until corona activity was identified on the test object, this being the visual corona inception voltage (CIV). Then, the voltage was increased by approximately 10% and maintained for 60 seconds. Next, the applied voltage was slowly reduced until corona extinction, this value being the corona extinction voltage (CEV). This process was repeated three times for each connector and type of applied voltage (AC, positive and negative DC), while completing a photographic report of each tested specimen.
Experimental corona tests were performed with the connectors placed at a reduced height above the metallic ground plane, as proposed in [20]. This approach allows performing the tests in a smallsize high-voltage laboratory, thus drastically simplifying the time required to assemble the experimental setup, reducing the voltage applied and the cost of the high-voltage generators required, while ensuring corona occurrence. The bottom flat surface of the connectors was placed at a height of 225 mm above the ground plane in order to force the electric field to reach the CEV value when operating below 120 kV. The experimental layout to determine the CEV value is shown in Figure 4.

Experimental Results
This section presents the experimental results performed in this work, including the surface roughness measurements and the results of the visual CEV tests. Tables 2 and 3 summarize, respectively, the maximum, minimum and mean values of the Ra, Rq and Rz parameters along the four paths of each connector shown in Figure 3. Table 2. T-type J-line connectors. Results of the surface roughness measurements.

Ra [μm]
Rq  Table 3. T-type S-line connectors. Results of the surface roughness measurements.

Experimental Results
This section presents the experimental results performed in this work, including the surface roughness measurements and the results of the visual CEV tests. Tables 2 and 3 summarize, respectively, the maximum, minimum and mean values of the R a , R q and R z parameters along the four paths of each connector shown in Figure 3.   Tables 2 and 3 in a graphic format, as well as the tendency lines obtained by applying a linear regression.  The results in Figures 5 and 6 show an important variability of the surface roughness parameters Ra, Rq and Rz among connectors with the same ageing level, although new connectors (A0) have lower roughness-parameter dispersion among different connectors, specifically for their maximum values. However, the tendency lines show a slight increase in the surface roughness with ageing level.

Visual CEV Tests
It is a recognized fact that CEV values depend on the local atmospheric conditions, i.e., pressure, temperature, and absolute humidity. Specifically, ambient pressure has a deep impact on CEV values [8,32]. Thus, when testing under variable atmospheric conditions, the obtained CEV values must be corrected to those of the standard reference atmosphere (P =101.3 kPa, T = 20 °C, and H = 11 g/m 3 ), as suggested in the IEEE Std 4-2013 standard [31], which was carried out in this work. Tables 4 and 5 and Figure 7 summarize the results of the visual corona tests.  The results in Figures 5 and 6 show an important variability of the surface roughness parameters R a , R q and R z among connectors with the same ageing level, although new connectors (A0) have lower roughness-parameter dispersion among different connectors, specifically for their maximum values. However, the tendency lines show a slight increase in the surface roughness with ageing level.

Visual CEV Tests
It is a recognized fact that CEV values depend on the local atmospheric conditions, i.e., pressure, temperature, and absolute humidity. Specifically, ambient pressure has a deep impact on CEV values [8,32]. Thus, when testing under variable atmospheric conditions, the obtained CEV values must be corrected to those of the standard reference atmosphere (P = 101.3 kPa, T = 20 • C, and H = 11 g/m 3 ), as suggested in the IEEE Std 4-2013 standard [31], which was carried out in this work. Tables 4 and 5 and Figure 7 summarize the results of the visual corona tests. Percentages: maximum CEV deviation among the 3 sample replicas with respect their average value Percentages: maximum CEV deviation among the 3 sample replicas with respect their average value Substation connectors are designed to withstand a voltage level at least 10% above the nominal voltage [33]. They are also designed so that the electric field strength at any point of their surface is always below 30%-40% the minimum electric field strength required to generate corona. Therefore, the CEV deviations of the different connector replicas shown in Tables 4 and 5 are not critical to generate unwanted corona effect at the rated operating voltage. Figure 7 show the results presented in Tables 4 and 5.
The results summarized in Tables 2 and 3 and Figure 7a show that there is not a clear pattern between the applied ageing level, and the visual CEV value. There is a possible explanation of this behavior. On one hand, the dimensions of the geometric radii of the connector are much greater (in the order of several mm) than those of the surface roughness (in the order of several µm). On the other hand, the roughness is not very affected by ageing due to the specific characteristics of the sand-casting process, which causes a rather poor surface finish. In addition, the geometry of substation connectors is optimized against corona, so they have larger dimensions and a more complex geometry than those of the line conductors, thus hindering corona occurrence. Therefore, the variability among connectors due to the sand casting process seems to have more influence than the ageing effect itself.

12
124.79 105.2 100.1 Percentages: maximum CEV deviation among the 3 sample replicas with respect their average value Substation connectors are designed to withstand a voltage level at least 10% above the nominal voltage [33]. They are also designed so that the electric field strength at any point of their surface is always below 30%-40% the minimum electric field strength required to generate corona. Therefore, the CEV deviations of the different connector replicas shown in Tables 4 and 5 are not critical to  generate unwanted corona effect at the rated operating voltage.  Figures 7 show the results presented in Tables 4 and 5. The results summarized in Tables 2 and 3 and Figure 7a show that there is not a clear pattern between the applied ageing level, and the visual CEV value. There is a possible explanation of this behavior. On one hand, the dimensions of the geometric radii of the connector are much greater (in the order of several mm) than those of the surface roughness (in the order of several μm). On the other hand, the roughness is not very affected by ageing due to the specific characteristics of the sandcasting process, which causes a rather poor surface finish. In addition, the geometry of substation connectors is optimized against corona, so they have larger dimensions and a more complex geometry than those of the line conductors, thus hindering corona occurrence. Therefore, the variability among connectors due to the sand casting process seems to have more influence than the ageing effect itself.

Conclusions
Substation connectors' manufacturers are very interested in knowing the long-term corona behavior of their products since corona is an undesired phenomenon that produces undesired and harmful effects. This allows for manufacturers to offer customers maximum safety and transparency regarding the behavior of the connectors.
This paper studied the ageing effect on the surface roughness and the visual corona extinction voltage of sand-cast aluminum connectors. Different aluminum connectors were artificially aged in a salt spray chamber during different periods, i.e., 0 h, 500 h, 750 h, and 1000 h, thus obtaining four ageing levels.
The experimental results presented in this paper show that the surface roughness tends to slightly increase with the ageing level of the connectors. However, surface roughness measurements show an important variability in the surface roughness parameters among connectors with the same ageing level, although new connectors (ageing 0) present slightly less variability of the roughness parameters Ra, Rq and Rz, in particular for their maximum values. However, visual CEV tests revealed that there is not a clear pattern between the applied ageing level and the experimental CEV value.

Conclusions
Substation connectors' manufacturers are very interested in knowing the long-term corona behavior of their products since corona is an undesired phenomenon that produces undesired and harmful effects. This allows for manufacturers to offer customers maximum safety and transparency regarding the behavior of the connectors. This paper studied the ageing effect on the surface roughness and the visual corona extinction voltage of sand-cast aluminum connectors. Different aluminum connectors were artificially aged in a salt spray chamber during different periods, i.e., 0 h, 500 h, 750 h, and 1000 h, thus obtaining four ageing levels.
The experimental results presented in this paper show that the surface roughness tends to slightly increase with the ageing level of the connectors. However, surface roughness measurements show an important variability in the surface roughness parameters among connectors with the same ageing level, although new connectors (ageing 0) present slightly less variability of the roughness parameters R a , R q and R z , in particular for their maximum values. However, visual CEV tests revealed that there is not a clear pattern between the applied ageing level and the experimental CEV value. This is attributed to the fact that, on one hand, the geometric radii of the connectors are in the order of several millimeters, whereas the surface roughness is in the order of several micrometers, thus having little impact on corona inception conditions. On the other hand, it was verified that surface roughness is not very affected by ageing due to the particular characteristics of sand-cast connectors, which exhibit a rather poor surface finish. Finally, substation connectors are self-protected against corona appearance due to the specific geometric design (large dimensions and complex geometry compared with those of the line conductors) that is optimized against corona under rated operating conditions. It is concluded that the inherent variability among connectors, mostly due to the sand-casting process, seems to have more influence than the ageing effect itself.