Electrical Stability of Oil-Impregnated Paper - Relation to the


Electrical Stability of Oil-Impregnated Paper - Relation to the...

0 downloads 48 Views 625KB Size

Electrical Stability of J

Oil-Impregnated Paper Relation to the Properties of the Oil C. E. TRAUTMAN AND W. N. ARNQUIST Gulf Research & Development Company, Pittsburgh, Penna.

R

EFINED petroleum oils

tion is evaluated by following For oils derived by various refining treathave been widely used as changes in power factor occurments from one Texas Coastal oil, the ring a t high electrical stress impregnants for paper to power factor stability and high voltage life produce composite electrical inand elevated t e m p e r a t u r e s . of the oil-impregnated paper were found sulation in industrial applicaThese tests are performed in to be related to specific dispersion and other tions such a s h i g h - v o l t a g e "stamp capacitors" and furnish c a p a c i t o r s and underground better methods of evaluating properties reflecting the composition of the the oils in insulation than do cables. For these applications oil. the insulation must have a tests on the oil in the absence Some results on other oils suggest that of paper and metals. The work s u f f i c i e n t l y h i g h dielectric these correlations are not general. In this strength to prevent electrical described below is an extension of that described by Clark, failure and also a relatively connection it is assumed that beneficial or low power factor to limit the with particular emphasis on the detrimental substances are present in the generation of heat. Both the chemical nature of oils, the oils in amounts too small to be detected by high dielectric strength effect of antioxidants on power the tests reflecting the composition of the and the low power factor must factor stability, and the effect oil. be maintained during the life of oil composition on resistance to breakdown under highof the insulation. The exact The addition of a small amount of an mechanism of electrical failure voltage stress. Results are antioxidant to the refined oils was found to in this type of insulation is not described to show the effect of improve markedly the power factor stability completely understood. It is refining the oils on the electriof the impregnated insulation. k n o w n , h o w e v e r , t h a t oxical stability of the oil-impregdation of the oil impregnant nated paper in stamp capacic o n t r i b u t e s t o t h e Dower tors. Investigation was limited to oils with viscosities in the neighborho:d of 100 Saybolt factor increase of the insulation during use. Universal seconds a t 100" F. The selection of suitable impregnating oils or liquids for such high-voltage insulation applications involves the consideration of physical and chemical properties of the impregElectrical Tests nant. These properties may be divided into two broad classifications : POWER FACTOR STABILITY. Essentially the same procedure described by Clark (3) was used to evaluate the power factor stability of the oil-impregnated paper. Stamp capacitors were 1. Physical properties, such as viscosity, coefficient of exmade with eight layers of 0.5-mil unbleached linen capacitor pansion, and pour point, which are of importance in paper separating aluminum foil electrodes and held together befabricating and handling the insulation and also in the tween glass plates with phosphor-bronze clips. The electrode mechanical and structural properties of the insulation. area was 24 sq. cm., giving a capacitance, after impregnation, 2. Chemical nature of the impregnant which influences: of about 750 micromicrofarads. Groups of four capacitors, in a. Its resistance to breakdown under high voltage stress. 600-ml. beakers, were dried for 48 hours at 110" C. in a vacuum b. Its oxidation resistance and the stability of its power oven, where the pressure was maintained at less than 500 microns factor. of mercury. The oils to be studied were then de assed by d r o p ping slowly into their respective beakers, with t f e low pressure still maintained, until the capacitors were completely impregThe composition and properties of mineral oils as affecting nated and the beakers filled. The oven and contents were then dielectric stability have been found difficult to evaluate. allowed t o cool to room temperature under vacuum, and the beakers were finally removed for tests on the capacitors. Many attempts have been made to develop tests which will The power factor stability of the impregnated insulation was predict the stability in advance of using the materials in servdetermined by measuring the power factor of these capacitors as a ice. Such tests include accelerated aging on samples of the function of the time they were exposed to a stress of 800 volts per impregnated insulation and tests for evaluation of the rate of mil, 60 cycles, at 75' C. The power factor measurements were made at 30" C., at about 7 volts per mil, 60 cycles, with a conjudeterioration of the impregnant under specified conditions. gate Schering bridge. Since the measurements were made at low Clark (3) in a recent review of problems concerning oilvoltages, the observed power factor changes are believed to indiimpregnated high-voItage insuIation, places special emphasis cate predominately the chemical changes which took place in the on the chemical nature of the oils. He describes a test in insulation as a result of the aging treatment. The stability tests were terminated \Then sufficiently definite trends in the which the chemical stability of the oil-impregnated insula1535

INDUSTRIAL AXD ENGINEERING CHEMISTRY

1536

VOL. 32, NO. 11

power factor changes had been established. For the more stable oils the tests were conducted for about 50 days. HIGHVOLTAGELIFE. The "life" or time to failure under a steady 60-cycle voltage of the impregnated insulation was determined in a manner similar to that followed by Bousman (@. The stamp capacitors, prepared as described above, were connected in parallel to a 5.8-kv. transformer, 60 cycles, at 75' C., with a 10,000-ohm resistance in series with each capacitor to prevent overvoltage transients from affecting the specimens. Considerable spread in the life values for the capacitors in a beaker was often found. In some cases one or two values were very low as compared to other tests on the same oil, and they were discarded. The life values reported are averages of the individual values obtained from the capacitors in one or more beakers. GAS E V O L ~ T I OThe N . evolution of gas from the oils under ionic bombardment at reduced pressures was studied. The testing apparatus and procedure were similar to that described by Berberich ( 1 ) . Thirty-five milliliters of the degassed oil sample was bombarded in a glass concentric tube ozonizer with a 60-cycle, 13,000-volt potential applied directly to the cell. The evolved gases were collected in a reservoir of about 1100 cc. total volume, and the amount of gas was determined by pressure measurements with a McLeod gage. The volume of gas after 2 hours of discharge, corrected to normal pressure and temperature, was taken as a measure of the gas evolution.

Chemical and Physical Tests

0.8

0.6

0.4

l

a

I 1

e

o

la

1

OIL 3A 0

u-L

UNSATURATION. The total unsaturation was determined by the usual sulfuric acid reaction following a method used by Fenske (4). Kattwinkel's variation ( 7 ) of the sulfuric acid method was found to give about the same values for this unsaturation in many cases. In addition, both Kattwinkel's test and the one recommended by Fisher and Eisner ( 5 ) indicated a negligible olefinic content of the oils. Thus, the total unsaturation values were considered to reflect the aromatic content of the oils. WATERMAN ASALYSIS. The method of Vlugter, Waterman, and van Weston (9) was used for determining the apparent composition of the oil samples in terms of the aromatic, naphthenic, and paraffinic (free and side-chain) content from the following data: aniline point, index of refraction for the D line, density, and molecular weight. Aniline points were measured by the Waterman method; densities were determined a t 20" C. by the pycnometer method; and refractive index measurements, also at 20" C., were made on an Abbe refractometer. The molecular weights were estimated from viscosity measurements according to the empirical relation of Keith and Roess (8). The viscosity values were obtained with Ostwald pipet viscometers and ranged from 100 to 108 Saybolt Universal seconds a t 100' F. and around 38 at 210" F. SPECIFICDISPERSION. It has been known that aromatic hydrocarbons have a considerably greater dispersion than most other types of hydrocarbons. Von Fuchs and Anderson (6) showed that dispersion measurements serve as an excellent means for reflecting comparative aromatic content in oil mixtures. In the present work, dispersion measurements were made on an Abbe-type refractometer at 20' c.

Refining Treatments The various oil samples tested were prepared by refining treatments with sulfuric acid, with aluminum chloride, and in one instance with furfural solvent extraction. The acid-refined samples were prepared by thoroughly agitating the oil with the percentage by weight of concentrated sulfuric acid indicated under refining treatment in Table I. After agitation for 20 minutes a t 80" F. (26.7' C.), the mixtures were allowed to settle for at least 2 hours. Most of the clear oil could be decanted and separated from the acid sludge that had settled to the bottom of the container. The oils were then neutralized by treatment with about 10 per cent by weight of contact clay for 10 to 20 minutes at 275' F. (135' C.). After settling a short time, the clay was removed by filtration. In some cases the samples required two clay treatments for complete neutralization. All of. the treated samples had neutralization values of 0.02 mg. of potassium hydroxide per gram of oil or less. The aluminum-chloride-treated samples were prepared in a manner similar to the acid-treated samples. The temperature of treatment was 175" F. (79.4' C.) and the time of contact 2 hours. The oils were clay-neutralized by the same procedure used for the acid-treated oils. The furfural-refined sample, 21, was prepared in a commercial extraction tower. The solvent-oil ratio was about 2.5 to 1,

R;OVEMBER, 1940

INDUSTRIAL AND ENGINEERING CHEMISTRY

and the average temperature of extraction was about 225' F. (107.2' C.), In order to improve the color of the oil, it was treated with a small percentage of acid before neutralization with clay.

Results The results of the chemical and physical tests on the oils are given in Table I. I n the designation of the oils, the numbers refer to the distillate and the letters distinguish the various oils refined from the distillate. Subscripts are used to indicate successive samples of the same oil; the comparative properties of these samples illustrate the reproducibility of the refining treatments.

1537

refining. The results show t h a t increased acid treatments are of decreasing effectiveness in improving the power factor stability, although the aromatic content is progressively decreased as shown in the table. I n Figure IC the d a t a for oils 2A1 and 2F1 illustrate the effects of comparable amounts of acid- and aluminum-chloriderefining on distillate 2. These results show further that small changes in aromatic content may be associated with large changes in power factor stability. Figure I d shows the capacitor power factor stability of an impregnant derived by intensive solvent extraction followed by a light acid treatment. The intensity of treatment is

T ~ B I .I. E PROPERTIES OF OILS Oil Designation

Refining T r e a t m e n t

1 1A

Unt 6% Unt

2Ai 2Aa 2B 2C 2D 2E 2Fi

6'7

2F2

5%

2

Density, d x

Index of Refraction In%o)

0.9140 0.9074 0.9080 0.9062

1.5057 1.5020 1.5005

63

0 9010 0.9011 0,8985 0.8977 0.8994 0.8992 0.9046

1.4990 1.4966 1.4963 1.4944 1.4934 1.4962 1.4958 1.4982

6%

0,9036

2Gs

6%

0,9039

2H

63% AlCls Solvent (furfural) follow ed

2G1 2G2

21

3A

E$11% 16% 21% 5%

0.

0.

by small % HzS1 0 4 5% AlCls (Penna. dist.)

0.9055

1.4992

Sp. Optical Dispersion X 10' (20" C.)

126 123 120 120 120 115 114 111 110

Aniline Point, n C . 69.3

72.3 71 . O

72.5 73.5 76.0 76.0 77.0 78.2 76.0

Compn. by Waterman Analysis % 70 naph70 aromatics thenes paraffins

16.0 14.4 13.6 12.3 12.0 10.5 10.3 9.5

31.8 31.9 34.9 36.2 36.8 36.3 36.9

10.9 10.0

35.1 36.0

37.7

% Total Unsatn.

C c . Gas Evolved

(N. P. T.) per 2 Hr.

... ...

52.2 53.7 51.5 51.5 51.2 53.2 52.8 52.8 52.6 54.0 54.0 50.5

37 30 31 27 26

19 21 27

6.0

50.4

22

5.5

50.4

23

6.0

5.8 5.3

22 23

22

8.7

38.7

11s

77.0 74.0

11.4

1 ,4979

117

74.0

11.6

1 ,4980

118

74.0

11.6

38.1 38.0 38.0

0.8974

1,4938

112

77.0

9.6

37.3

53.1

23

6.0

0.8816

1.4829

98

89.0

0.7

41.4

57.9

4

8.7

0.8527

1.4730

108

103.0

1.5

16.4

82.1

2

8.6

...

116

The various oil properties listed are directly related to the actual percentage of a particular refining agent used. For example, the density, index of refraction. and specific dispersion decrease regularly with the amount of refining. While these relations may be of some interest, the discussion has been confined to the correlation of the results of the electrical tests with those properties of the oils which reflect their composition. The relation of the electrical properties of the oils and the actual refining treatments will be evident without further mention from the correlations to be shown and the oil properties listed in the table. Some typical results of the power factor stability tests are given in Figure 1. Each point represents the average of the power factor values for four capacitors in one beaker. Individual values were found to vary by a small percentage from the average except for the large power factor values where the variation was somewhat more. The effect of the addition of a small percentage of an antioxidant compound to some of the oils was studied. The physical properties of these oils were not modified noticeably b y the addition of this compound, Figure la shows the power factor stability of three sets of capacitors. The first set Tvas impregnated with the untreated Texas Coastal distillate 1, and the second set with oil 1A derived from the distillate by moderate acid refining. The lowest curve shows the stabilizing effect of the antioxidant when added to oil 1A. Table I shows that while the acid treatment improved the power factor stability to a great extent, the apparent composition of the oil as indicated by the aromatic content was only slightly changed. The effect of more intensive acid treatments is illustrated in Figure Ib. Here oils 2Az, 2B, and 2D mere derived from oil 2 by progressively increasing the proportion of acid used in

22

...

...

indicated by the fact that the aromatic content was reduced from about 14 to less than 1 per cent. The insulation is relatively unstable which, according to Clark (S), is characteristic of highly treated Texas oils. However, the stability of this highly refined oil was markedly improved by the addition of the antioxidant. Not all oils with low aromatic content had poor power factor stability characteristics in these capacitor tests: Figure l e shows the results obtained with oil 3A, a moderately treated paraffin oil with an aromatic content of about 2 per cent. I n subsequent tests the addition of the antioxidant to this oil was found to improve further the stability of the impregnated insulation. All of the tests with the antioxidant strongly suggest that the deterioration measured by the stamp capacitor test is largely an oxidation phenomenon. I n order to compare the instability of the oils, as determined by the stamp capacitor power factor tests, the average rate of change of power factor, S , was calculated in each case. This was done by neglecting the initial part of the instability curve and by using the relatively straight portion of the curve as a measure of this rate. There was little question in assigning approximate magnitudes to these slopes except for the less stable oils. I n these instances it was found that any reasonable values given to the slopes would not materially affect the correlations to be discussed below. The rate of power factor change appears to be a more consistent measure of instability than a comparison of the actual power factor values after a n arbitrary aging time. Explanations for this may be advanced in connection with such factors as differences in initial power factor, possible induction periods in the oxidation reaction, and readjustments between paper and oil which are associated with the mechanical forces

1538

IIVDUSTRIAL AND ENGINEERING CHEMISTRY

VOL. 32, NO. 11

FIGURE 2. RATEOF CHAXGE OF POWER FACTOR OF OIL-IMPREGNATED PAPERUNDER TESTCONDITIONS vs. TOTAL VNSATURATION, AROMATICCONTENT, AND SPECIFIC OPTICALDISPERSION OF THE IMPREGNANT

caused by the electrical stress. I n connection with the latter, it may be noted t h a t the average mechanical stress on the impregnated paper was calculated to be about 2 pounds per square inch (0.14 kg. per sq. cm.). One of the effects of this stress was to cause the capacitance values of the test specimens t o increase abruptly by 1 or 2 per cent during the first few days of the test. Thereafter these values increased slowly a t rates of the order of a few hundredths per cent per day with generally higher rates for greater values of S. These latter rates of increase in capacitance suggest increases in the effective dielectric constant of the insulation. Three correlations of the instability factor, S, with the initial properties of the oils are shown in Figure 2. The first of these shows the relation between S and the total unsaturation. This is a similar correlation to the one found by Clark (S), indicating t h a t there is a n optimum refining treatment for greatest power factor stability. A somewhat better correlation of S with aromatic content was found. However, the best correlation was with specific dispersion, as illustrated in the lower curve of Figure 2, where the results with the antioxidant compound are also included for comparison. The dispersion measurements are direct and comparatively simple t o make, and for this reason they are relatively the most reliable from the experimental standpoint. Thus the departure of the points from the specific dispersion curve in Figure 2 indicates the relative reliability of the value of the power factor instability factor, S , while the increased scattering of the points from the other curves reflects the difficulties encountered in the unsaturation and aromatic determinations. I n Figure 2 the results for oil 1A were plotted as triangles, and the dashed lines were approximately established by the results for oil 1, with an S value of about 90 X which was far off scale (Figure la). The results for oil 1A were checked, and it is certain t h a t they do not correlate with the results for the No. 2 oils. Unfortunately time did not permit the study of the effect of increasing refining treatments on oil 1 in order to establish better the dashed lines in Figure 2. The lack of correlation of the results for the N o . 1 and 2 oils in Figure 2 emphasizes that instability factor S is not determined uniquely b y any of the three measures of aromaticity. Apparently the relative presence of active substances not indicated directly by these measures of the composition is very important in determining the instability as measured by the stamp capacitor tests. It was pointed out above that the addition of an antioxidant can improve the stability markedly without changing noticeably the initial properties of these oils (Figure 1, a and d ) . Thus the measures of oil composition used in this work cannot be relied upon to indicate the relative presence of such beneficial substances. It is also probable that these measures are not quantitative evaluations of all detrimental substances which may be present. The lack of correlation between the results for XO.1

and 2 oils may be explained by assuming that there were relatively more detrimental substances or fewer beneficial substances, or both, in No. 2 oils as compared with No. 1 oils. The suggestion that both beneficial and detrimental substances exist in mineral oils has often been made. Clark (3) explains his results on Texas oils, which are similar to the Ushaped curves of Figure 2, b y assuming t h a t progressive refining treatments remove first the substances t h a t contribute preferentially to instability, and then those t h a t largely promote stability.

Life Tests The results of the high-voltage life tests have shown generally t h a t the highly treated oils fail first. Bousman ( 2 ) pointed out a correlation between specific dispersion and the high voltage life of paper impregnated with oils of 100 seconds Saybolt Universal viscosity. A similar relation is illustrated in Figure 3 where the log of the life of some of the oils is plotted against the specific dispersion. A small change in the specific dispersion of the oil has a large influence on the life of the impregnated insulation. T o the extent that this

IMPREGNANTS 0 -0DERlVED

'*I95

FROM OIL 2

105 115 125 SPEC1FIC DISPERSION

FIGURE 3. LOG^^ OF THE LIFEAT 1450 VOLTS PER MIL, 75" C., OF OIGIMPREGNATED PAPERus. THE SPECIFIC OPTICALDISPERSION OF THE IMPREGNANT

test measures the suitability of an oil for high-voltage insulation, the results indicate that a minimum refining treatment is desirable. The life values associated Rith distillate 1 and oil 1A do not agree with the results for the S o . 2 oils; this indicates that the straight line drawn in Figure 3 is not a general correlation for all oils. A similar conclusion was reached above with regard to the S values associated with these oils.

NOVEMBER, 1940

INDUSTRIAL AND ENGINEERIKG CHEMISTRY

The results of the gas evolution test were somewhat disappointing in that they did not give much information about the oils. Since considerable emphasis (1) on the relation between gaseous ionization in impregnated insulation and gas evolution has appeared in recent years, it was hoped t h a t a close correlation could be found between the results of the gas evolution test and the stability tests in the present work. However, the tabulated results show t h a t for these oils, representing a wide range of life values and power factor stability, the gas evolution values varied b u t little; only slightly higher values appeared for the more refined oils. The measures of aromaticity afford a much more sensitive means for predicting the life and power factor stability of the insulation than the gas evolution test.

Acknowledgment The authors are indebted to Paul D. Foote, Gulf Research & Development Company, for permission to publish this paper, and to H. A. Ambrose and M. Muskat of this labo-

1539

ratory for helpful criticisms in the course of the work. Acknowledgment is also made of the help of several members of the Testing Division in obtaining test data on numerous oil samples.

Literature Cited (1) Berberich, L. J., IND. ENG.CHEM.,30, 280 (1938).

(2) Bousman, H. W., Conf. on Electrical Insulation, Natl. Research Council, New York, 1937. (3) Clark, F. M., IND. ENG.CHEM.,31, 327 (1939). (4) Fenske, M . R., unpublished test developed a t Penna. State College. ( 5 ) Fisher, C. H., and Eisner, A.. U. S.Bur. Mines, R e p t . Investigations 3356 (1937). (6) Fuchs, G. H. von, and Anderson, A . P., ISD.EXG.CHEM.,29, 319 (1937). (7) Kattwinkel, R., Brennstof-Chem., 8,353 (1927). (8) Keith, J. R., and Roess, L. C., IND. ENG.CHEM.,29, 460 (1937). (9) Vlugter, J . C., Waterman, H. I., and Weston, H. A. van, J . Inst. Petroleum Tech., 21, 661 (1935). PRESENTED before t h e Conference on Electrical Insulation of the National Research Council, Cambridge, hlass.

SPIRIT VARNISHES Incorporation of Ethylcellulose R. C. ERNST, J. B. TEPE, AND I. W. HUTCHISON, JR.' University of Louisville, Louisville, Ky.

Spirit varnishes in which ethylcellulose was substituted for from 5 to 25 per cent of the resin content were tested to determine the effect on hardness, moisture resistance, and resistance to cold check and abrasion. Results of experimental work indicated that in all properties tested, spirit varnishes can be materially improved by the substitution of ethylcellulose for small percentages of the resin content.

P I R I T solutions of solubleresins have for some time found extensive application, fulfilling the demand for an inexpensive, quick-drying, protective coating. One of the most universal uses of spirit varnishes has been in label and carton finishes. A more recent outlet has been created by the widespread usage of traffic paints for marking highways and streets. During the present European conflict, black-outs in many cities have resulted in marked increases in traffic accidents. The number of casualties has, however, been mat,erially reduced through extensive marking of street intersections, traffic lanes, and crosswalks. Pigmented spirit varnishes have been used for this work, and the resulting increase in demand for raw materials is being met only with difficulty.

S

1 I. W. Hutchison, J r . , is the D o w Chemioal Company Fellow a t the University of Kentucky.

Resins employed in spirit varnish formulation include not only natural resins such as Manila, pontianak, shellac, and dammar, but also some synthetics. However, the less expensive grades of natural resins are most widely used. The dry resin film necessarily lacks many desirable properties, and this investigation was undertaken to determine if the substitution of ethylcellulose for small percentages of the resin content would materially improve the moisture resistance, hardness, and resistance to cold check and abrasion of the film. Results of experimental work reported in this paper are on varnishes made up with pontianak and Manila resins. The resin cut was 3.0 pounds per gallon of solvent. I n each series of varnishes, standard ethoxy ethylcellulose (47.5-49.0 per cent ethoxy) of 20-centipoise viscosity was substituted for from 0 t o 25 per cent of the resin content in increment8 of 5 per cent by weight. The solvent consisted of 75 per cent alcohol and 25 per cent toluene by volume. Resins employed were pontianak chips and DBB Manila chips. Varnish films were tested as described below, and results of tests are given in Table I.

Hardness The most satisfactory hardness test employed mas a determination of scratch hardness, using a series of Venus drawing pencils. Films of equal thickness were cast on polished plateglass panels. The softest pencil which was capable of scratching the varnish film with ordinary writing pressure was selected. Hardness of pencil lead increases with the H number. In both varnish series, the substitution of ethylcellulose resulted in an increase in scratch hardness which was proportional to the amount of ethylcellulose incorporated.