Dec.. 1921 THE JOURNAL OF INDUSTRIAL AND ENGINEEBINQ CHEMISTRY 1149

The Determination of Moisture in Insulating Oils'a By C. J. Rodman

RESEARCH DEPARTMENT. wESTINGHOUS& ELECTRIC AND MANUPACTURINQ CO., EAST PITTSBURQH, P A .

Many methods of determining water in oils have appeared in the literature, but no reliable simple method has come to the author's notice. By examination of the literature in detail, a classification of methods has been made. This classification includes five qualitative, a dozen approximate, and few quantitative methods. In dealing with moisture in insulating oils, a method cannot be called quantitative unless an accuracy of closer than 0.002 per cent is obtained. The approximate methods include those which determine moisture content to an accuracy of 0.01 to 0.05 per cent. The qualitative tests are those which have been recommended here a i d there for the use of the electrical engineer as a rapid means of finding the presence of water. No insulating oil should be allowed to take up enough moisture to show positive tests by qualitative methods, because the dielectric strength of an oil becomes less considerably before the moisture can be thus detected.

QWALITATIVS TESTS I-Addition of anhydrous CuSO4. &Hot nail "spdts," indicating Hz0.L' 3-Addition of certain water-soluble, but oil-insoluhlc dyes.' 4-Centrifugation.4 5-Electrical precipitation:b

l-Loss of weight by heating. For nonvolatile oils and greases.6.1* !&-Diluting with solvent and settling by gravity.

oils, but a diluent is to be avoided. &Use of color comparator tube.6 4-Tleating with normal acids.6 +Treating with CaCn.

Turus blue with HzO.,%*,*

APPRVXIMATE ANALYSIS

Applicable to thin Centrifugation may be used to hasten.'

This is convenient, hut with petroleums ac- curate to 3 per cent of the Hz0 value only, as CLHZ is absorbed by them. The 3 pcr cent accuracy holds only above 0.01 per Sent content.'

One cc. Hz represents 0,0016 g. HzO in the sample used. This is accurate and convenient. Complicated apparatus necessary to insure thorough contact of sodium with oil and to eliminate quantitatively all of the hydrogen.6

Water-saturated xylene, a mixture of xylene and toluene, or benzene and toluene are used. Accurate to ap- proximately 0.033 g. of Hz0 per 100 cc. of distillate.'

8-Directly distilling off the water. This is accurate to within 0.03 per cent, and does not apply to light oi1s.a Inert gases used to dry oils.2

9-Water in oils may be tested by conductivity providing no other de- composition products are present. Only approximate results obtainable.0

1O-Oil mixture with organic solvent to produce cloud. A more general case than No. 7.

11-The Tyndall effect is used to determine finely divided water in SUS- pension.

6-2'1eating with sodium.

7-Distilling with a immiscible liquid.

QUANTITATIVE METHODS I-By special application of No. 3 (qualitative tests) less than 0.003

per cent Hz0 can be detected. Necessitates color scale, clear oils-standard- ization for quantitative work.

2-Vacuum treatment while shaking sample, heating, and collecting Hz0 in PZOS tube. Method quantitative if oil is free from nitrogenous vola- tile compounds, as is usually the case. By special method we are able to determine water to f 0.001 per cent.

3-Same treatment as No. 2, using freezing mixture to collect HpO and volatile components, afterwards estimating water by No. 4 method. Good results obtainable, but complicated apparatus must be employed.

4-Vacuum treatment of thin film of oil running over heated surface. Vapor eliminated and measured by oil-water vapor pressure difference.

5-Mixing oil with dry ether (solvent), treating with CaCz, drawing off CzHz by vacuum, precipitating out in alcoholic AgNO8 solution and noting change in resistance with precipitation of silver acetylide.8 This gives time-rate curve and is very desirable where both hygroscopic and constitu- tion moisture ere involved. It is accurate to less than 0.01 per cent and is chie5y applicable to certain insulating materials and oils.

Inasmuch as the dielectric strength of an oil is greatly influenced by the moisture present,lo it is highly desirable to '

1 Preqented before the Section of Petroleum Chemistry a t the 61st Meeting of the American Chemical Society, Rochester, N. Y., April 26 t o 29, 1921

2 Published as Scientific Paper 92 of the Westinghouse Electric and Manufacturing Company.

* Numbers refer to Bibliography.

determine the moisture content accurately. The different forms in which the moisture is present within the oil may vary greatly. This is dependent upon the composition of the oil and its impurities. Water may be in finely divided suspension, condensed nuclearly upon dust particles, ab- sorbed by fibers or soaps, in colloidal form or in true solution. The true dielectric strength of pure oil is most strongly in- fluenced by water in solution, but the usual conductivity takes place by the alignment of particles within the oil.

Very pure mineral oil of the paraffin series does not dis- solve more than 3 X per cent H20.11 The unusually high percentage of moisture sometimes found in oils is at- tracted by impurities which are soluble in oil. Such im- purities (unsaturated compounds, fatty acids, etc.) possess a greater coefficient of solubility for water than do the pure paraffins .

A Sample bulb C : Tnermometer E = E lec t r i ca l w ndinq F -Freez inq mixture G = Manomete r

FIG. 1

Of the numerous possible methods of determination of water in transformer and similar oils only two or three may be depended upon for the accuracy desired. A modification of Method 2 (Quantitative Methods) has given excellent results.

PRINCIPLE OF METHOD Water is fractionated from the oil a t reduced pressure and

elevated temperature, during rapid agitation of the sample. Under these conditions the partial pressures of oil and water are far apart and fractionation is rapid. The water is col- lected with some oil distillate in a trap cooled by liquid air. Redistillation of the water and collection in a P20h weighing bottle take place upon the removal of liquid air. The small amount of oil vapor distilling over passes on through the phosphorus pentoxide. The method depends for its success upon the production of a great number of thin films within the oil sample. A low freezing mixture may be substituted fac: the liquid air.

APPARATUS AND PROCEDURE Fig. 1 shows diagrammatically the arrangement of the

apparatus. The weighing bulb* is of special design con- sisting of two glass parts, an inner and outer chamber. The inner chamber (2 cm. in diameter X 10 em. long) is slightly constricted a t the lower end and has a stopcock handle a t the upper end, This fits concentrically within the outer chamber (2.5 cm. in diameter x 10 cm. long), and is made airtight by the ground joint near the top. An inlet allows gas to enter the outer chamber first. The gas passes down-

* Modification of bulb used by Dr. H. C. P. Weber, Westinghouse Electric and Manufacturing Company.

1150 THE JOURNAL OF INDUSTRIAL

ward through the outer chamber, entering the inner from the bottom. Pure P205 packed between two plugs of glass wool within the central chamber is used as the moisture absorbent. The gas passes up through the PZOS and out through a small tube directly opposite the inlet. By turning the inner member through 90" the bulb is sealed to any gas pressure difference. Such a bulb holds a vacuum many weeks.

The sample bulb, which fits within the heater shaker, is of about 50-cc. capacity. A tube, closed a t the bottom and just large enough to accommodate a small thermometer (range 0" to 150" C.), passes down into the bulb. The bulb can be kept a t any desired temperature by the electric heater surrounding it. The motor speed can be controlled so that the shaker runs from 50 to 250 per minute. The liquid air trap consists of a closely bent U-tube of 1-cm. tubing, sur- rounded by a Dewar flask containing liquid air.

PRocEnvRE-Before making a run the apparatus is thor- oughly cleaned and dried. A few cc. of oil to be tested are used to rinse the sample bulb before pipetting in the 25 to 40 cc. of the oil. The bulb is connected to the liquid air trap .by means of a heavy, pure rubber, pressure tube. The Pz06 tube is fastened in place also by means of a rubber connection. Rubber cement prevents gas leakage a t the joints. The vacuum is now turned on and the pressure very quickly reduced to a millimeter. The manometer reading is taken and the P206 bulb closed. Closing a screw pinchcock tightly a t either connection to the PZOE bulb suffices to hold the vacuum in the system when taking the P20a bulb out for weighing. The Pz05 bulb is replaced and screw cocks opened. While shaking and heating the bulb to 140' C . the liquid air trap is covered with COz snow and acetone or liquid air contained in a silvered Dewar flask. The water vapor is quickly and completely eliminated from the oil and froeen in the trap, which also condenses any light oil distillate. Upon removing the freezing mixture the frozen water is rapidly vaporized and caught quantita- tively within the P& bulb. By closing the Pz06 bulb a t the same pressure as in the blank weighing one can readily obtain the eliminated water by difference. Phosphorus pentoxide does not absorb any oil vapors at the pressures used.

The more rapid method does not account entirely for the moisture in the air within any part of the apparatus. As shown in Table I, it gives comparatively good results, varying *0.002 from the mean. A single determination is not to be recommended where the best accuracy is desired.

Two methods of procedure may be followed.

TABLE I

Per cent Per cent Difference Water Added Water Found

0.0042 0.0047 +O ,0005 0.0194 0.0216 +o. 0022 0.0110 0.0092 -0.0018 O.OOS8 0.0078 -0.001 0.0074 0.0086 +o .0012

d dl + dz, etc.' - 0.0012

In this case the vacuum is cracked without shaking $he sample bottle, thus taking out the air above the oil sample when the weighing tube or freezing mixture is not yet in place. A pinchcock is immediately cloeed. The freezing mixture is then raised around the trap and the weighing thbe put in place, and the determination is made as described above.

The second method of procedure takes into account the amount of moisture in the air. Two blank runs are first made to determine the humidity. Several runs made for humidity moisture check the theoretical values. With the addition of this correction to the first procedure, the moisture in hydrocarbon oils may be determined accurately.

AND ENGINEERING CHEMISTRY Vol. 13, No. 12

NOTES ON THE METHOD In determining the moisture content to ~ 0 . 0 0 1 care must

be exercised a t every point in the determination. The oil sample must be representative of that which is to be examined. Any apparatus used in connection with the oil should be carefully cleaned and dried and then rinsed with some of tJhe oil sample. Humidity changes influence the water content appreciably if the sample is open to the air. Care should be exercised in keeping the breath away from the sample, as surface condensation of water vapor may grent'ly influence the result.* The time of a run should not exceed 20 min. This gives ample time to insure complete elimination of the moisture.

Factors which give high results, and hence are additive, may be briefly summarized as follows:

1-Surface condensation of water from humid air. 2-Contamination of any of the apparatus by water or dust. 3-Volatile nitrogenous and basic substances within the oil. 4-External leaks of any kind. 5-Increase of pressure during run so that excess gas is trapped in PzO,

6-Adsorption of some oil vapor in weighing bulb. bulb.

Subtractive errors are as follows: 1-Slimination of water not complete as a result. of high viscosity of oil

2-Bath temperature too low. 3-Shaking not vigorous enough to expose all t.he oil to thin film condition. 4-Decrease in pressure over the initid pressure within the bulb. 5-Time of Bhaking oil too brief.

sample.

Weighing errors may be positive or negative. Using the representative transformer oil little difficulty

is experienced in the determination of water due to its vis- cosity, impurities present, or to the volatility of the sample. Leaks are readily taken care of by using rubber cement between the glass-rubber connections. Other possible leaks are prevented by .proper construction of apparatus. Freshly sublimed P205 free from organic matter will not ab- sorb and hold any oil under low vacuum conditions. Should. any oil distillate come over in appreciable quantities i t is readily eliminated by surrounding the Pa05 tube by a temper- ature bath of 100" to 150' C.I2 Though the possible errors seem numerous, it is only necessary to recognize them to eliminate them by following out carefully the procedure recommended, and little difficulty will be experienced in ob- taining an accurate result.

BIBLIOGRAPHY 1--Elect. Club J., U. S . A., May 1904. 2--E!ect., 66 (1911), 490. 3-A. H. Gill, "Oil Analysis,'' 6th Ed., 1913, p. 22. P-Chem.-Ztg. , 33 (l909), 1259. 5-Bureaii of Mines, Technical Paper 26 (1912).

6-Orig. Com. 8th Intern. Congr. A g p l i e d Chem., 10, 17. 7-J. Ind. Ens. Chem., 10 (191S), 357; Direccion General de Mines,

S-U. S. Dept. Agriculture, Bureau of Chemistry, Circular 97; J. SOC.

9-J. SOC. Leather Chem., 3, 101-4; 14, 654-74; 3, 2fJ6-8; J. I n d . Eng.

See also patent litera- ture.

Buenos Aires, Bol. SB (1915).

Chem., 29 (1910). 197; Chem.-Ztg., 6 (1892j, 29.

Chem., 12 (19LO), 486-98 1 0 - ' E l ~ ~ t . , 66, 490; 11-See also Elect., 82, 103, 67, 81s; Z. QnoYg. Chem., 81, 24-39;

12-Z. anovg. Chcm., 81, 24-30. 13-J. SOC. Chem. I n d . , 39 (1920), 305.

Elect. J.: 16, 74-6.

2. Elektvochcm., 17, 346.

*The rubber connections used in the vacuum line do not give up any appreciable amount of moisture after having been oncc subjected t o low vacuum, if care is taken subsequently to keep out humid atmosphere.

A committee of eleven has been appointed to draft tentative plans for an organization of oil men, to comprise refiners, jobbers, producers, supply and tank men, which it is expected will em- brace the entire oil industry of the southwest, probably with headquarters in KanSas City. Part of the plan includes the formation of an arbitration board to settle differences between conflicting interests, and an oil exposition to be held yearly.