[Trade Journal]
Publication: Electric Journal
Pittsburgh, PA, United States
vol. 16, no. 1, p. 8-16, col. 1-2
Application of Theory and Practice to the Design of Transmission Line Insulators
G. I. GILCHREST and T. A. KLINEFELTER
The theoretical elements of the problem which laid the foundations for the following developments were outlined in previous issues of the JOURNAL.(1) The present article(2) deals, for the most part, with laboratory tests of various new designs and a comparison of these designs with those now in commercial use.
USUALLY any design problem of engineering may be quite easily separated into two rather distinct phases. The one phase is termed "theoretical" and infers that the service experience, processes of manufacture, cost of materials, cost of manufacture, etc., are placed secondary in importance in the search of an ideal design. The other phase is termed "practical" and infers that the design has been evolved mostly from a consideration of service experience. It is generally conceded that a design evolved by either method may have certain advantages. The object of the following investigations has been, to link together these two phases in a specific application, namely; the design of pin-type transmission-line insulators.
Having dimensions of ring and rod chosen such as to give maximum breakdown voltage over the surface for the mean diameter of torus ring. Two logical questions at once arise :—
First, are the insulator designs installed in service at the present time satisfactory?
Second, can one type of design be developed that will be satisfactory in all localities?
The first question is answered by a resume of current engineering literature, which offers convincing evidence that there is a field for improvement. A comparison of the flashover voltage versus overall dimensions and weight of the present insulator designs would seem to warrant an attempt toward uniformity. Furthermore, the divergence of certain characteristics of some designs from the average curves indicates that some of the designs must be far from efficient.
The causes of such chaos in insulator designs are quite obvious. First of all, the progress in transmission engineering has been rapid. The expanding transmission companies demanded insulator designs which would offer a good factor of safety. There was no previous operating experience to use as a basis in new developments and consequently it was often necessary for the transmission engineer to propose his own design. Moreover, the majority of our present insulator types were designed when the electrical and mechanical characteristics of porcelain were less understood than at present. As a result, various features were accentuated as most important. That is, at one period a long leakage path was required, regardless of voltage distribution per shell from capacity current or leakage current; then again a high puncture voltage, then a high mechanical strength, and so on. Naturally, many mistakes were made and a large proportion of the older insulator designs have failed in service application.
CAUSES OF INSULATOR FAILURES
The knowledge that certain insulator types have failed in service is of little value in the redesign of insulators unless actual conditions of service and cause of failure are known. Also, the cause of failure of a particular type in one locality should be compared to the cause of failure in other localities. Hence, before attempting to develop a new type of design, a study was made of available data on insulator deterioration and the opinions of operating engineers in various parts of the country were obtained.
From these discussions, from published data on insulator deterioration, and from observations of insulators that had failed in operation, it would seem that the following are the main causes of failures :—
I—Improper distribution of dielectric field.
2—Improper distribution of surface leakage.
3—Porosity.
4—Mechanical breakage. (a) From handling. (b) Mischievous shooting and stone throwing. (c) Insufficient strength as a support. (d) Brittle material.
5—Lightning.
6—Birds and animals short-circuiting the line.
7—Unequal expansion of metal, cement and porcelain.
8—Internal stresses in the material.
9—Defective batches.
Items 3, 4(d) and 9 are the problems of the ceramic engineer, rather than of the designing engineer. These items have doubtless been of great importance in the past, but more scientific and painstaking factory control must minimize them in the future.
It was believed that the data previously outlined, in conjunction with the available data of other investigators(3) of both an analytical and experimental nature, afforded sufficient basis from which to formulate preliminary designs. Hence, several theoretical insulator designs were produced out of a usual commercial porcelain body. The dielectric field of these was then investigated under a voltage of approximately the same value that would be impressed in service. Thereafter practical considerations, such as deterioration of the various commercial units in service, manufacturing limitations, etc. were taken into account, with the resultant design described later.
METHOD OF DETERMINING FORM OF DIELECTRIC FIELD
The dielectric field was determined by the following procedure:—The insulator was fastened in a position such that the plane of the field to be determined extended horizontally. A piece of fullerboard was fitted over a half section of the insulator in this plane. In all cases the cross-arm supporting the insulator was grounded as in service where steel construction is used. Finely divided asbestos was then sifted evenly onto the sheet of fullerboard, voltage at 60 cycles of the desired value was applied, and the sheet was gently tapped until the particles had adjusted themselves. Permanent records were obtained by placing a sheet of photographic printing paper over the fullerboard, obtaining the field as above, and exposing the paper after the particles had become arranged. That the stronger portion of the field around an insulator was not disturbed materially by the presence of the fullerboard or the asbestos particles was proven by suspending a piece of finely drawn glass in parts of the field by means of a silk fibre supported by small insulated rods. As nearly as could be checked, the glass indicated the same direction of the field as the asbestos particles.
THEORETICAL INSULATOR DESIGNS
The dielectric fields of five theoretical designs were determined. Wherever a customary transmission cross-arm and line wire are used, there are two principal planes of the dielectric field which show the greatest difference, i.e., the plane of the cross-arm and the plane of the line wire. These two planes are 90 degrees apart and in passing from one to the other the transition is gradual. During the investigation, records were taken of the dielectric field for these two principal planes and of a plane midway between the two. The plane in which the particular field was taken is indicated by the reduced top projection at the upper left portion of each figure.
DIELECTRIC FIELD FORMS OF THEORETICAL DESIGNS
The direction of the electrostatic field about various theoretical designs is shown in Figs. I to 5, while Table I gives the length of path over the insulator surface between electrodes and the 60 cycle flashover voltage.
From a consideration of Table I it is evident that a flashover value of between 20 and 23 kilovolts is per inch (8 and 9 kilovolts per centimeter) of surface may reasonably be expected if the unit is designed with contours of the surfaces approximating the flow lines of the dielectric field. Of course, the flashover on the unit without rain sheds, Fig. 2, is somewhat lower, being 14.4 kilovolts per inch. The lower flashover on this unit is due to two conditions, i.e., the porcelain surface does not follow the dielectric field in all planes and the small tie wire produces corona and subsequent static discharges at a relatively low voltage. Placing a static shield on the top of this unit increased the flashover voltage 18 percent.
With the field form between cap and pin as given in Fig. 4, and the voltage values given in Table I, theoretical insulator designs could be determined for such electrodes. Such designs should follow the surfaces indicated on Fig. 4, as (a) and (b). The highest flashover voltage for a given surface distance between electrodes would thereby be obtained. Moreover, the flashover voltage of such a unit could be closely approximated if the electrodes have sufficient radius of curvature at points of contact with the insulating material and a good seal is made between the metal and the insulating material.
MODIFICATIONS OF THEORETICAL DESIGN TO MEET OPERATING AND MANUFACTURING CONDITIONS
Insulators based on such theoretical data would be excellent from the electrical and mechanical standpoints if they were to operate in clean, dry air. However, the commercial insulator must maintain the transmission system during the heaviest of snow and rain storms. Moreover, it 'must have sufficient leakage distance to prevent flashover or even high power loss from surface leakage when the surface becomes dirty and wet.
The production of one-piece insulators for high-voltage service, although possible, would be costly. Also, the puncturing voltage of a one piece unit would be low for a given thickness, since the stress in an insulating material between metal electrodes of different potential varies as a logarithmic function. The separation of the unit into parts that are cemented together, more uniformly distributes the stress of the dielectric if the unit is properly designed. It also decreases the probability of complete failure of the insulator and facilitates factory production, lessening the cost of the commercial unit.
The use of a special cap would be desirable from a dielectric standpoint. However, the voltage characteristics under rain are the same whether the usual line and tie wire or a special cap is used. Moreover, the cost and ease of replacement, cost of construction, etc., favor the line and tie wire construction.
PROPOSED COMMERCIAL INSULATOR DESIGN
With the above limitations of the theoretical designs and the causes of insulator failures in mind, the type of unit indicated in Fig. 6 was evolved.
Summed up briefly this type of design embodies the following features:
1—The surfaces a conform to the flow lines of the electrostatic field.
2—The surfaces b of the rain sheds conform to the equipotential surfaces.
3—The lines of mechanical stress are parallel to the electrostatic flow lines.
4—The leakage resistance per shell is about equal, being increased gradually from the head to the center shell. 5—It has approximately equal capacity per shell.
COMPARISON WITH OLDER DESIGNS
It is not possible to much more than indicate in the following discussion the methods employed to compare the proposed type of design given in Figs. 6 and 7 with the older commercial insulators. Samples of various commercial designs were produced and were subjected to rather thorough laboratory tests at the same time tests were made on insulators of the proposed design. It should be noted that the insulators of the new type used in the comparative tests do not exactly correspond to the proportions of Fig. 6. In order to lessen the cost of investigation, insulator sheds of several diameters were obtained from one set of molds by trimming the individual shells before burning. This also accounts for the unfinished appearance of the edges of the sheds, etc., in some of the experimental designs.
In the following comparison it is not assumed that the evolved design should be final in each detail. The main goal toward which work is being directed is uniformity of all the elements entering into the designs with the idea in view of arriving at a type of design which will be equally successful in resisting failure in service whatever the requirements are in that particular section. In the following comparisons the items causing failure in service are discussed in the order given at the beginning of the paper.
1—Dielectric Field Distribution—The shortest air path under electrostatic stress should be at least long enough to prevent overstressing of the air at any point. In the previous theoretical discussions it was proved that wherever porcelain and air are in series in a dielectric field the voltage gradient per unit distance through the porcelain will be 1/4 to 1/5 the voltage gradient through the air. It is obvious that any thin section of air between porcelain sheds of a customary line insulator will be over-stressed even at the normal line voltage of the insulator.
In order to make a comparison of the dielectric fields of various insulators, their field forms were determined as in the investigation of the theoretical designs. It is believed that the following field forms and illustrations, Figs. 8 to 20 inclusive, sufficiently indicate that many present types have not been designed with a full appreciation of the advantages of shapes that conform to the electrostatic flow lines in obtaining the most efficient distribution of the stresses in the dielectric field.
Fig. 8 (insulator C) gives the dielectric field of a unit of the type used in the early developments of high voltage transmission. The air between sheds just below the cement section is highly stressed. Because of the height of the pin in proportion to other dimensions of the unit the stress toward the base of the pin and the supporting cross arm is very low. Moreover, the third shell of the insulator is spaced so close to the insulator pin that it does not take its proportion of voltage stress when either dry or wet flashover occurs.
Fig. 9 (insulator D) shows the dielectric field of a three piece insulator of a somewhat more recent design. The center shed is better spaced than in insulator C. However, the air just below the cement sections is highly stressed and the short rain shed of the second shell gives an unequal voltage distribution at flashover, dry or wet.
Fig. to (insulator E) shows the dielectric field of a four-piece unit of comparatively recent design. The sheds of this design are more uniformly spaced, but the air between sheds just below the cement sections is highly stressed. The stress throughout the dielectric field of this unit is an improvement over the types C and D. However, the short second shed and protected fourth shed give unequal voltage distribution at flashover dry or wet.
Fig. II (insulator F) shows the dielectric field of a unit of the proposed design. The shortest air path between shells is sufficient so that the air is not overstressed at the working voltage of the insulator or until flashover occurs. Moreover, the rain sheds are so spaced that each section of the unit takes its share of the stress at flashover, dry or wet.
Fig. 12 shows the dielectric field of insulator F having the upper surfaces of the rain sheds covered with a conducting paint. This field form, which approximates the rain conditions, indicates that the stress per shell on the unit during rain would be approximately equal. Moreover it indicates that the stress in the dielectric field is more uniform during rain.
Fig. 13 shows the dielectric field of insulator F when equipped with Nicholson arcing rings, and indicates that the most highly stressed portion of the field about the insulator is not changed. However, the most highly stressed portion of the field between the line wire and the cross arm is now between the arcing rings and flashovers would, therefore, occur between the rings.
Fig. 14 shows the dielectric field of insulator F when static shields are placed at the top and base of the insulator. This combination would give a very fine distribution of stresses in the dielectric but would be rather expensive commercially.
60-CYCLE FLASHOVER TESTS
Flashover on most of the older insulator types is caused by the corona formation at the line and tie wires and the edges of the cement joints between shells. As the voltage applied to the insulator is increased, the area of the corona formation increases and static streamers gradually spread over the surface of the insulator sheds. The static streamers increase in length until the air insulation between them finally fails and flashover follows. Obviously, the path of the flashover will start along the path of these streamers. Trouble may thus be caused by the intense heat of the power arc and the rain sheds may be stripped from the insulator.
In the proposed type of design there are no static streamers from the edges of the cement section between shells up to flashover voltage. The corona formation at the tie and line wires therefore, builds up until flashover occurs by breaking down an air path between the line and pin or cross arm. The proof of these statements may be seen in the following illustrations. The axes of the two units in each of Figs. 18, 19 and 20, giving comparative flashovers, were at the same distance from the camera lens and hence the dimensions are directly comparable.
The difference in the stress in the air around the insulators just below flashover voltage dry was very marked. Insulators of the proposed type, F, G and I in Figs. 18, 19 and 20, showed no appreciable corona except at the line and tie wires until flashover occurred. Static streamers began to spread out over the surfaces between sheds of insulators H and .1 at 80 percent flashover voltage and the arc therefore formed over the insulator surface. Of course the old type design in Fig. 20 has been entirely superseded, but it clearly indicated the entire neglect of a consideration of the dielectric field.
The difference in distribution of stress before wet flashover is even more noticeable. In insulators H and J the unequal spacing of rain sheds, combined with a highly stressed air between the sheds below the cement sections, produces preliminary discharges (marked p) between the sheds. These preliminary discharges throw electrical impacts onto parts of the insulator and short circuit portions of the porcelain between the line and pin. Consequently, when a line surge occurs during a rain storm or when the unit is wet and dirty, the factor of safety of these insulators in resisting puncture or flashover is actually no more, and sometimes is less, than it would be minus one of the shells.
Insulators F, G and I of the proposed design show no preliminary discharges except static from the tie or line wires to the pin or cross arm. Static discharges (marked s) are shown in Fig. 19. All of the leakage surface and the thickness of porcelain between the line and the pin, are, therefore, effective up to failure by flashover.
2—Surface Leakage—Table II gives the resistance per shell of various insulators tested during this investigation. The values were obtained by an integration of the surface, i.e., surface resistance equals where ds is an element of surface and y the radius of that element from the axis of the insulator.
It is obvious from Table II that certain of the older designs especially those having a short second shell, long inner shells, etc., have a very unequal surface resistance per shell. If the insulator surface becomes dirty and wet, so as to pass a leakage current of even a thousandth of an ampere, the voltage distribution would depend upon this current and the capacity current could be neglected. The voltage gradient over the insulator surface thus often becomes sufficient to cause discharge between the sheds and pin or cross arm or over the short sheds. An electrical impact is thereby applied to parts of the insulator and portions of the porcelain body between line and pin are short circuited. It is believed that the continued overstressing of parts has been the cause of many insulator failures in the past.
The surface resistance of the proposed designs as typified by insulators A and B in Table II is gradually increased from the top to the center shed, the increase being considered as an advantage, since the center sheds will usually become dirtiest.
A novel feature of the proposed design is illustrated in Fig. 21 showing insulators D, E, F and H. These units were set on a cross-arm, line and tie wire attached as in service, voltage applied and plaster of paris dust blown around them. The surfaces along the lines a of the proposed design (Fig. 6) are practically free of dust.
The reason for this is quite apparent. All the force acting in the dielectric field along this surface a is tangential and would tend to force the particles to the sheds above or below. The same action was noted when the units were subjected to atomized salt water. This feature would doubtless have some value in dust laden sections since the dust would tend to settle mostly on the lower shed and rain and wind would clean this to some extent.
It is necessary to clean the insulators in long portions of line in certain sections of country as the coast districts of California. It is very apparent that the proposed type of design may be cleaned much more readily and thoroughly than any of the older types.
3—Porosity—The deterioration of porcelain insulators in service was given little consideration during the early days of transmission engineering. The majority of transmission engineers preferred an insulator having a porcelain body which offered a high resistance to mechanical breakage. As a consequence, the porosity of the material, which varies inversely with the mechanical strength as regards resistance to mechanical impact, was considered of secondary importance. The results that this condition have caused in service have been clearly presented.(4) Since porosity is a function of the body compositions, manufacturing processes and burning, more scientific and painstaking factory control must minimize its effect in the future.
4—Mechanical Breakage—(a) From Handling: The increase of thickness of the rain sheds and addition of a drip edge will materially decrease the percentage loss from this cause.
(b) Mischievous Stone Throwing and Rifle Shooting:—Figure 18 shows the dry flashover on units G and H, and Table III gives the flashover voltages of broken units in percent of the unit when unbroken. Figure 19 shows the wet flashover on units J and I and Table IV gives comparative flashovers of broken units, as in Table III. Figs. 22, 23 and 24 show the flashovers on units having broken sheds.
As would be expected from a study of the dielectric field diagrams, the breaking of the second shed of the proposed type of design has practically no effect upon the flashover values of the insulator. In fact, as shown in the illustrations, the paths of the dry and wet flashovers did not follow over the broken shed. When sheds are broken, the corona formation and static streamers build out over the surface of the older type of design at a lower voltage than when the units are intact. The paths of flashover over these older types, therefore, follow the surface of the insulator. In the proposed type of design the absence of streamers from the porcelain surface causes the arc to keep clear of the insulator. A power arc will, therefore, be less liable to cause complete failure of a broken unit of the proposed design.
One of the most important features of the proposed design is that when the units are hit by stones, etc., the rain sheds will not crack or break beyond line a Fig. 6, due to the shape of the individual parts. The rain sheds of the older types of designs when hit are very likely to crack or break up into the cemented section. The first voltage surge or even normal line voltage will, therefore, often puncture the remaining shells. In fact, in the two series of tests photographed, both the older type of units punctured during the dry arcover after the sheds were broken.
One each of units H, I avid K and two of J were subjected to rifle shots. Twenty-two caliber long bullets were shot at the insulators from about 3o yards distance and in a line at 45 degrees to their axes. Figs. 25, 26 and 27 show the comparative breakage and the ability of the broken units to thereafter withstand electrical test. The shooting was done by men disinterested in the design of the insulators who were requested to do as much damage as possible.
Fig. 25 shows insulators I and .1 after 15 shots were fired at each. The top, second and third shells of I were broken, the second shell being cracked into the cemented section. The second shell of J was chipped in two places, the rest of the insulator being intact.
Fig. 26 shows insulators, H, K and J after 14 shots were fired at H, 12 at K and 28 at J. The second and center shells of H were cracked and the center of K. The sheds of .1 were chipped off in a few places but the shells were not cracked. These five units were then set with their axes at right angles to the line of fire. Not more than 5 or to shots were necessary to strip the main part of the remaining sheds from insulators H, K and I while one unit of type I still retained a considerable portion of its sheds after approximately 100 shots had been fired at it. The sheds remaining on the two units J were then knocked off by a hammer, to illustrate to those present that the surface of the insulator that follows flow line a would not be cracked thereby.
Fig. 27 shows the first dry flashover test made on these units after the shooting. Units H, K and I punctured at 33, 43 and 56 kilovolts, respectively. Unit J of the proposed type of design flashed over at a 105 kilovolts, the remaining porcelain body bounded by line a still being intact.
(c) Insufficient Strength as a Support: Two samples as per Fig. 28, were tested to determine the resistance to side pull. In each case load was applied at the wire groove which was one foot from the base of the pin. The parts from which insulator L was formed were obtained by trimming off the rain sheds of individual shells of unit J before burning and the mechanical test should, therefore, be about the same as that of unit J. The pin of unit L was cemented directly into the insulator. A separable pressed steel thimble was cemented into insulator F. The one-inch bolt of the pin cemented into insulator L failed at 4400 ft-lb., and 3100 ft-lb. bent the pin of insulator F as shown, the position of the insulator being such that additional load could not be applied. Both units were electrically intact after these tests.
(d) Brittle Material: All units used in these comparative tests were made of the same porcelain body and hence the question of brittleness, which is a ceramic problem, does not enter.
5—Lightning—It is generally conceded that a direct stroke of lightning will destroy any insulator that comes within its wake. However, some of the older designs, especially those having deep inner shells and heads of large diameter, were very vulnerable to any sudden impact voltage. In the first place, the impulse ratio (flashover voltage at high frequency divided by flashover voltage at normal frequency) of such insulators is rather high, and in the second place the ratio between flashover voltage in air and puncture voltage under oil is comparatively low. Furthermore, the body of the porcelain bounded by the flow lines a should have an impulse ratio close to one. A very high impulse voltage might, therefore, puncture through the rain sheds of the insulator, leaving this body of the unit intact. The thicker section of porcelain between line and pin will also materially increase the factor of safety of the unit.
6—Unequal Expansion of Metal, Cement and Porcelain—The introduction of a resilient material be-tween the tops of shells should eliminate the tendency of certain older designs to split off. Greater radii of curvature at the tops of the insulator shells and a cement section sloped from the axis should tend to eliminate the trouble from any difference of coefficient of expansion of the porcelain and cement.
7—Internal Stresses in the Material—Internal stresses set up in the insulator parts during manufacture should be very much decreased by the elimination of small radii in corners and sudden changes of cross section of the material.
CONCLUSIONS
Briefly stated, it is believed that the advantages of the proposed type, Fig. 6, over the older commercial types in resisting failure in service would be as follows:
1—When the insulator is dry, the corona and static formations are practically limited to the tie wire and line wire, up to flashover voltage.
2—When the insulator is wet, no corona or static formation occurs up to the flashover voltage. The flashover voltages for given overall dimensions are thereby increased.
3—The leakage resistance per shell is increased gradually from the head to the center shell. This takes into account the probability of the lower sheds becoming dirtier than the tops. The voltage distribution per shell is, therefore, equal when the insulator becomes dirty and wet and a heavy leakage current passes over the insulator.
4—Since the capacity per shell is about equal, the voltage distribution per shell will be equal when the insulator is clean and in dry air.
5—Since the distribution of voltage per shell depends upon the capacity current and leakage current, the distribution of voltage per shell in these designs should be approximately equal under all operating conditions.
6—The resistance of the insulator to side pull for a given weight and given electrical strength is relatively high. This is due to the feature of the design whereby the flow line a of the electrostatic field and the mechanical stress lines coincide.
7—The design of the individual shells is such that when they are tested before assembly the surface conforms to the electrostatic flow lines a. This allows testing of the individual parts to a higher percentage of service voltage than was possible with the individual shells of older designs.
8—Due to the shape of individual parts and of the assembled unit, the insulator sheds when hit by stones, rifle, balls, etc., do not break beyond the surface a. The unit, therefore, offers a considerable percentage of its original resistance to flashover after the sheds are broken. The same feature tends-to protect the insulator from complete failure during flashover in service.
9—Each characteristic of the insulator which would vitally affect durability in service has been treated uniformly throughout the line.
(1) In the JOURNAL for February, p. 36; March, p. 77; and November, p. 443, 1918.
(2) Revised by the authors from a paper before the American Institute of Electrical Engineers, June 1918.
(3) "Distribution of Potential about High Voltage Line Insulators," by C. T. Allcutt and W. K. Skolfield, in the Journal of Electricity, Power and Gas, June 17, 1916. "Electrostatic Problems," by C. W. Rice, in the Trans. A. I. E. E., Vol. XXXVI, 1917.
(4) "Ceramics in Relation to the Durability of Porcelain Suspension Insulators," Trans. A.I.E.E., Vol. XXXV, 1916.
