Document LJ9m7Y42ByDv0jQDqKmJBna2b

1700 S. SECOND ST J. R. Savage - JFQ 0. DeGarmo W. J. Wilson - WGK N. F. Mueller H. 0. Hubbard W. R. Richard D, C. Armstrong - WGK J. W. Molloy -JWGK____________ COMPANY CONFIDLN i u'.l. - -------- GENERAL OFFICES ANNISTON D. B. Hosmer W. B. Papageorge H. C. Carder V. R. Haupt (2) R. J. Stratmeyer L. C. Fuhrmeister N. L. Sample J. MacPherson D. Danna (3) G. W. Miller (2) P. G. Benignus J. G. Bryant J. 0. Bright C. Mcllwain R. E. Howard R. P. Steveni R. S. Yates H. 0. flehner IG&&7 PROGRESS REPORT TECHNICAL SERVICES DEPARTMENT ANNISTON, ALABAMA PLANT JOB NO. 370-481 REPORT NO. DATE June 2, 1967 TITLE: OBJECTIVE: QUALITY OF LIQUID AROCLORS (This report) Present details of a systematic study made to identify sources of variability in electrical grade Aroclor. PERSONNEL: W. B. Dunlap, Charles Mcllwain (L. C. Fuhrmeister) - REPORTED BY: W. B. Dunlap SUMMARY: Due to the high quality requirements of export grade electrical Aroclors, a systematic survey was made to isolate existing sources of variability, as planned in Report #4, Job 370-481, Hay 2, 1967. This survey revealed the following facts: , 1. High resistivity (15,000 x 10^ ohm-cm) exists at the blend tank but is de graded by passing through the system to the storage tank. 2. Possible sources of contamination have been found to be atmospheric, exposure to iron through ruptured tank linings and old, threaded galvanized pipe, unacceptable gaskets and packing, and exposure of drum loading to wea ther. 3. Unreliability in reported quality arises in contaminated samples and im precise analysis, with sampling error controlling by a factor of 4:1. Difference between duplicates of single tests on different samples, statistically calculated, could be as high as + 100%. 4. The test method devised by the Anniston PR&D group has better precision than the routine lab test and twice the accuracy (PR&D values average higher by factor of 1.8). RECOMMENDATIONS (including improvements planned and underway by TSD engineers) 1. Replace tanks and lines with aluminum. 2. Close the entire system under a nitrogen pad. QSW 310234 ANNISTON (continued) STLCOPCB4070221 3. Initiate a PR&D study of water washing of crude Aroclor to eliminate air blowing and need for lime in the still. 4. Develop a bomb sampling system. 5. As a referee method upgrade the routine lab test to match the PR&D method. W. B. Dunlap Technical Services Dept. /dp DSW 310235 STLCOPCB4070222 SURVEY OF LOW RESISTIVITY ELECTRICAL GRADE LIQUID AROCLORS I. Objective and Scope CONTENTS II. Procedure III. Observations and Data A. Degradation During Forward Flow 1. From plant survey 2. From previous studies 3. Conclusions on degradation in forward flow B. Sampling - Analytical Variability 1. Relationship of variability to resistivity level 2. Analytical lab test precision 3. Storage tank sampling error 4. Variability of drums and drum sampling 5. PRScD test precision C. Accuracy of Resistivity Data D. Summary of Conclusions on Sampling-Analytical Variability IV. Search for Causes of Quality Degradation A. Factors in Forward Flow of Material 1. Related to upstream components 2. Related.to incidental factors a^ Air-borne contamination b. Operator participation c. Still run sequence d. Corrective measures 3. Equipment, material contamination Page 1 2 3 3 3 4 5 5 5 5 6 7 7 7 8 10 10 10 10 11 3A0^6 os>N STLCOPCB4070223 B. Sampling Factors in Degradation of Reported Quality 1. Sample containers 2. Storage tank sampling 3. Drum samples 4. Tank car samples C. Analytical Factors in Degradation of Reported Quality 1. PRScD method 2. Special lab test 3. Routine control tests V. Conclusions Related to Causes of Low Resistivity VI. Summary of Recommendations to Obtain Future Required High Resistivity Levels A. Related to Plant Equipment Changes B. Related to Materials in Contact with Aroclor C. Related to Analytical and Sampling Capabilities P&Be 13 13 13 14 14 14 14 : 14 14 15 17 qSMM 3^0237 STLCOPCB4070224 SURVEY OF LOW RESISTIVITY ELECTRICAL GRADE LIQUIDAROCLORS ' I. OBJECTIVES>AND SCOPE The high quality requirement of export grade Aroclor (10,000 x 10 Min. resistivity see Appendix L) has made this product extremely sensitive to process contamination. Multiple treatment with earth is frequently required, with some material failing to upgrade to minimum requirements, and consequently being sold as regular. A systematic study has been made to isolate sources of contamination within the process and to spell out the limits of reliability in sampling and electrical testing. This survey has been designed to quantify answers to the following questions: 1. Does material dt the refining step (earth treatment) exceed resistivity specification minimum? 2. If quality is high at the refining step, what are the subsequent sources of contamination which cause minimal quality in final product? 3. What are the analytical and sampling components of the overall observed quality variability? If question No. 1 is answered in the negative, we then have the problem of defining the factors which make it impossible to upgrade Aroclor. Solution of this problem will require systematic laboratory and plant study of chlorine, biphenyl, blowing air, chlorination control, still operation, etc. Defining these variables is out of the scope of the present survey. DSW 310238 STLCOPCB4070225 . , , . ,j ; . . II. PROCEDURE -2 A. Process piping and material flow sheet appear in Appendix A. Over a five-day period a survey was made of resistivity values from blend tank to finished container, analyzing blend tank (after lab filtration), filtrate receiver, storage tank, and finished containers (drums or car). Multiple samples were taken at the storage tank to permit determination of sampling and analytical variability. Selected samples were further checked by the PR&D group in order to compare the two testing situations. During'1 this period supplementary data was recorded to monitor the environ ment, raw materials, and operating controls. Complete details of procedure will be found in Report No. 4, Job 370-481, which set up the study. B. A study was made of previous surveys in this field. A laboratory investigation was made in 1956 and subsequent limited investigations have been carried out by the Aroclor Department supervisors. Lately, J. G. Bryant of Anniston Plant PR&D group made one series of observations. These findings are included in this report. C. A search for causes followed the diagnostic study. These findings are included, along with recommendations for correction and further study. DsW 310239 STLCOPCB4070226 3 III. OBSERVATIONS AND DATA General Note: Weighing the significance of Aroclor electrical data, particularly Resistivity, involves an understanding of certain critical facts related to the test itself: (1) The test procedure is an arbitrary selection of conditions which has been accepted to permit reproducibility, (2) Given this specific test sequence, there are remaining variables in the area of cell cleaning and material handling which profoundly affect test results, and (3) Given this specific test sequence. these outside variables will predominantly cause low values rather than high. This leads to the generality that "the highest value is the most accurate value" in a series of replicate tests on the same material. A. Degradation During Forward Flow 1. The five-day survey is summarized in Appendix B -- "Batch-Sample Summary" and Appendix C -- "Progress of Resistivity Through the System." All resistivities were measured by a special analytical laboratory analyst using a clean test cell for every determination and, therefore, are not necessarily indicative of accuracy or pre cision of routine lab testing. Appendix C shows resistivities of all material flowing to the storage tank and into final containers. Blend tank values are inconsistent, since some are much higher than corresponding filtrated material while others are much lower. Interpretation of this phenomenon is treated later in this report from the viewpoint of one or all of three factors which have strong statistical correlation: operator variability, atmospheric contamination, and run sequence in the still. Each filtrate sample can be paired with the corresponding blend tank, but the corresponding storage tank sample is a composite of all previous material. Hence, the final sample of the storage tank is to be compared with the average of contributing filtrates and blend tanks, as well as the following drums or tank car. These weighted averages are tabulated below: Batches Blend Tank Filtrate Tank Storage Tank #2 #5 Drums Final Lot 535 14,900 4,480 2,900 2,600 AI-123 536-538 20,000 9,810 2,100 1,320 AI-124 539-543 6,100 10,740 7,200 6,900 AI-1006 544-545 3,200 5,260 3,200 3,300 AI-125 546 12,800 5,900 4,000 Discussion : (1) A perfectly consistent drop exists between filt: tank and storage tank. (2) With plant filtration practice unchanged throughout this study, it is difficult to interpret an increase OS\N 310240 STLCOPCB4070227 -4- from blend tank to filtrate, except as a consequence of contamination, either during the lab filtration which was required to remove earth prior to analysis, or by the operator in sampling the tank. Both alter natives are possible and are treated below in detail. With the axiom that errors are on the low side, these low blend tank values would be expected to be at least equal to the corresponding filtrate tanks. (3})The equality of storage tank with shipping containers is surprising in light of the historical evidence of lower drum and higher car values. This probably reflects the extreme care that was used throughout the experiment. _ 2. Previous Studies -- V. R. Haupt, J. W. Molloy, and J. R. Savage con ducted informal experiments during their tenure as Department Super visors. These were primarily designed to define the optimum sojourn time in the blend tank for earth treatment. All studies indicated a rise in resistivity followed by a decline, with a maximum at 1 1/2 21/2 hours earth treatment. In 1956 W. B. Dunlap carried out a laboratory-plant study designed to measure the effect of operating variables. The results were published as "Analytical Laboratory Investigation #3," July 26, 1956 and appeared in Appendix D "1956 Study-Effect of Duration of Earth Treatment and Storage on Plant Batches," and in the following summary. This data confirms and quantifies the observation of the Department Supervisors. Table I Summary of 1956 Investigation (1) Effect of hot aging in original glass sample bottle Average of 10 lots - sample as received after 30 hours @ 90 Q 1790 x 10^ 4370 (2) Value of earth treatment in plant - Resistivity went up with earth treatment to a maximum in 2 hours, then fell off. The decline continued through the filter tank and storage tank until the resistivity in the storage was 7 - 12% of maximum value in the blend tank. . ^3) Effect of variables in laboratory earth treatment - Resistivity increased rapidly with treating time for first hour, then continued at slower rate, but showed no decline as found in plant treatment; 80C was more effective than 150 for treating temperature; higher resistivities were obtained with more earth, but the greatest proportional rise was obtained in the first 0.05%. DSW 310241 STLCOPCB4070228 -5 (4) Decline of resistivity from filtrate tank to storage tank is further demonstrated by an extended study of Aroclor after leaving the filtrate tank. Averages of all Aroclor 1242 production over a 15 day period show: Filtrate tank Storage tank Tank car 5210 x 10^ 1160 2900 3. Conclusions on Degradation in Forward Flow a. Contamination from blend tank to storage tank. b. Progressive contamination jwithin the blend tank which is compensated by earth absorption until earth saturation is reached, after which degradation occurs. c. Stable quality going to drums, improved quality going to tank cars (in the <5,000 resistivity range). B. Sampling - Analytical Variability 1. Appendix E summarizes data comprising the experiment statistically designed to separate sampling, testing, and loading variability. Appendix F summarizes duplicate tests on the same sample by PR&D, using different cells for each element of the pair. 2. Variability related to resistivity level - The PR&D data is extensive enough to get an idea of the correlation between analytical variability and the resistivity level. Statistically the correlation is very significant. Some curvature in the relationship is indicated. More extensive: data might prove that the regression is not linear, but the burden of present evidence, up to 50,000, shows no great deviation from a straight line. Obviously, then, the standard deviation of resistivity test data is not constant, but must be considered relative to the resistivity level. For simplicity, the constant variability parameter has been taken as (T~. Therefore, in the following study of components R of variability, the standard deviation is given in terms oftf'R. 3. Analytical Laboratory test variability - The analytical test precision measured here will inevitably be better than routine testing. In this study one analyst of proven reliability was used; the test cell was cleaned each time before using. Precision figured across all analysts "and procedures would be higher. . Test variability is calculated from 4 pairs of duplicates, expressing the parameter as a function of the resistivity. DSW 310242 STLCOPCB4070229 -6 Table II Averages Difference Lot Sample Betweep Tests d(Test) Lot Sample Average Average d(Test) R 536 1 4400 5200 1800 .34 2 3500 1100 .31 DiffArepce.^between samples 1700 537 1 3900 2600 1200 .46 2 5250 100 .02 Difference between samples 2650 Average d(T) .28 R . Analytical Std. Deviation (t) " (.28)(.886)R TM .25R 4. Storage tank sampling error - The storage tank is sampled from a 1/4" valve and nipple attached to the circulation line. Routine procedure is to rinse a warm sample bottle and cap twice, catch a full bottle, cap tightly, label, and carry to the laboratory. In the laboratory the lip of the bottle is wiped with tissue paper and Aroclor is poured into the test cell. Routine sampling methods were followed in the ' study. Combined sampling and analytical error is calculated from the two "differences between the samples" in Table II, recognizing that each sample average comprises two test values. From this combined error the sampling error is separated by analyses of variance technique. Table III Difference between samples d (sample + test)______ (sample + test) R___________ 1700 .39 2650 .67 Average * .53 \, 0 Std. Dev (sampling + testing) (st) " (.53)(.886) R .47R DSW 310243 STLCOPCB4070230 S 2 - S2 + S? St 8 V N (.47R)2 - S2 + JL25R12 s2 7N * number tests per sample g m storage tank s sampling error 43R 5. Variability of drums and drum sampling > Drum sampling is done with a glass thief, by pressurizing the drum and forcing Aroclor into a sample bottle which has been previously rinsed twice from the storage tank sample port. Drum-to-drum variability cannot be extricated from its combination with drum sampling error in the present study. The combined variability is determined from three ranges of three. - Table IV Lot 123 124 1006 Range of 3 Samples dds 1,600 750 7,600 Average of 3 Samples 2,600 ,1,300 6,700 Average R .61 .58 .88 .69 CO Std. Dev (Drums, drum sampling, testing) (.69)(.591) R .40R S 2 dst S2 + S2 ds t (.40R)2 - S2 + (.25R)2 ds _ _ drum <is sampling error 6. PR&D test precision - Appendix F presents a series of duplicate tests made by the PR&D method. Analytical precision of this method is calculated from the differences and expressed as a function of re sistivity. Ave. d " 5.64 .24 R 24 Std. Dev (PR&D test) = S = (.24)(.886)R = .21R Pfc C. Accuracy of Resistivity Data Precision of test data has been examined above, dealing with inability of resistivity tests to obtain the same values on repetitive tests. Accuracy DSW 310244 STLCOPCB4070231 -8- of the test refers to its ability to determine the true resistivity level by the average of a large number of repetitive tests. (Refer to "General Note" under III above;;) Under the specified conditions of cell charging, temperature and test time, higher test data are considered to be closer to the true resistivity. 1. Table V presents the five test pairs which compare the PR&D test with the special lab procedure employed in the plant study. Table V Batch Lab Test PR&D Test 541 St. Tk. 11,100 10,800 542 St. Tk. 900 9,200 543 St. Tk. 7,200 10,000 544 St. Tk. 1,900 2,900 546 Filt. Tk. 8.300 * 20,700 Average 5,880 10,720 .2 Using the standard deviations for analytical precision previously established for these two test methods, we can easily set up a statistical test to determine the significance of this difference in average. This test takes the form of the following "t" ratio: 1 B o 00 00 t - 10.720 - 5 V\/(. 25R)2 + (.21R)2 5 4.840 \i^[(.25)(5880)12 + ff.21)(10.720)P V 4.0 This ratio is highly significant. We are led to the conclusion, there fore, that the PR&D method gives higher resistivities, in the order of 1.8 magnitude. PR&D test 1.8 X special lab test The superiority of the PR&D test would be greater, of course, over the foutine test used by the lab, in which a cell is not pre-cleaned before each test and several analysts of varying abilities are employed. D. Summary of Conclusions on Sampling-Analytical Variability 1. Specific sources of variability in resistivity test data Analytical (special lab test) Std. Dev; .25R Sampling storage tank Std. Dev. * .43R DSW 310245 STLCOPCB4070232 9 Drum + drum sampling Std. Dev. .32R PR&D analytical .Std. Dev. " .21R 2. Amount of variability in resistivity data is a direct, roughly linear function of the resistivity level itself. 3. The greatest source of error is in storage tank sampling. To statistically bring this error down to magnitude of the test error, it would be necessary to test four samples and average the results. 4. The PR&D test is more accurate than the lab test by a ratio of roughly 2:1. 5. Confidence in various sampling-testing combinations a. One storage tank sample, one lab test Overall std. dev. = .50R Expected range Of 80% of replicate measurements " + .94R i: -Range of the reported value = approximately the test value b. Two storage tank samples, one lab test each Overall std. dev. .35R Expected range of 80% of replicate measurements + .65R : . Range of the reported value = -- 2/3 test value c. One drum sample, one lab test --- ' Overall std. dev. .40R Expected range of 80% replicate measurements m + .58R (This value is less than the corresponding value of (b) due to greater, confidence in a std. dev. figured from 6 pairs as compared to 2 for storage tank.) ; Range of reported value + 1/2 test value d. One drum sample, one PR&D test . Overall std. dev. " .38R Expected range of 80% replicate measurements + .50R c: Range oJ reported value = +1/2 test value DSV\, 310246 STLCOPCB4070233 10 IV. SEARCH FOR CAUSES OF QUALITY DEGRADATION A. Factors in Forward Flow of Material 1. Changes in properties of upstream components - Raw materials, control of chlorinators, personnel on chlorinators and blowing variables are all upstream variables which could be expected to influence liquid Aroclor quality in the blend tank. However, an inspection of flow rates and stream size quickly reveals that batch-to-batch variations from these causes are impossible. Ahead of the still are the continuous blow tank (500 gal., kept 2/3 full), the catch tank (1000 gal., kept 2/3 full), the #4 blow tank (containing 0 - 2400 gal,, pumped out when full). Forward flow from chlorinators is 7.5 gal./min. With the damping effect only very long term effects in the chlorinators will be revealed in the distilled Aroclor. 2. Changes of incidental factors - Appendix G shows graphically the sequence of batches through the system, related to time, and to other variables which are time-oriented. The factors revealed are the only ones of many studied which showed any possible shift during the 7 days covered by the - - survey. / - ' ; . ' ->r; /- - The conspicuous aberration in the experimental data shown in Appendix C is the profound change in relationship of blend tank to filtrate tank samples between batches 535 - 538, 539 - 547, and 546 - 547. As dis cussed above (III A-l) this can be found to coincide approximately with the three factors on the chart. Statistically speaking, these are all strongly correlated with blend tank observations and cannot be separated into specific individual effects. Strong statistical correlation does not prove a cause and effect relationship. However, in considering corrective action these factors certainly should be carefully investi gated. / a. Wind direction. The parathion department releases small amounts 6f mercaptan-type odors and considerable S02J the PNP department releases PNCB vapors and assorted fumes; the chlorine department releases chlorine and hypochlorite fumes. These departments are all west of the Aroclor department and laboratory. During the first period (535 - 539) the wind was steadily toward the west; during the second (540 - 547) it was 180 reversed, blowing these air-borne contaminants toward the Aroclor plant and laboratory. The effect of air-borne contaminants would be pronounced during exposure of samples and sampling equipment. This occurred during " dip-cup sampling of blend tank and filtrate tank. However, the filtrate tank did not show the abrupt drop of the blend tank, and the tanks were sampled identically, using same equipment and method. Air-borne contaminants could also effect the testing, particularly with exposure of blend tank material during the lab filtration re quired to remove earth. Inevitably, considerable air would be pulled through the filter paper. It seems more likely that the harmful effect of outside fumes was suffered in the lab. This is consistent with the experience of two other occasions when very small concentra tions of vapors (paint fumes, scorched phenolic plastic smoke) made DSW 310247 STLCOPCB4070234 11 reliable resistivity testing impossible. This test is apparently very sensitive to extremely small amounts of fouling material. b. Operator participation. Statistically speaking, the participation of operators K-H is strongly involved in the shift of blend tank data. However, K-H also sampled the filtrate tank, without evidence of a shift of any kind. This factor should be discounted. c. Still run sequence. No. 3 still is chain-fed until 6 receivers have been filled, then bottoms are pumped out. This sequence during the experiment is shown on the chart. The initial high period shows perfect correlation with the first sequence (runs 3-6); the low period fits the full middle sequence (runs 1-6); the last high period approximately coincides with the last sequence -(runs 1-3). Only batch 545 is outside the perfect correlation. We cannot ignore the possibility that the middle sequence in the still was heavily fouled (air-borne contaminants, contaminated lime?) and purged only by pumping out, after which a relatively good sequence would start (but still not equal to the high level of the first sequence). d. Corrective measures. (1) Close up the entire system from the stills downstream. Air borne contamination should be excluded. This can be done by sealing the system under N2 Exposure in the blend tank will be prevented by use of a continuous treatment column; process changes have been proposed to eliminate the need for lime, mak ing possible an unbroken N pad in the still. If lime is needed, an airlock for introducing lime can be installed. Closing the system will also prevent the fouling which occurs at present whenever heavy rain blows through the still area over the blend tank. (2) Carefully protect lime and earth. These additives carry with them all fouling material to which they are exposed. At present no protection is provided other than the shipping bags. While drying in the oven, earth is exposed to circulating air from outside. (3) Elimination of Aroclor blowing. All air-borne vapors are included in the air blown through crude Aroclor to remove HC1. If high quality is as sensitive to plant fumes as this survey indicates, serious quality damage may be continually inflicted by present blowing methods. Two proposals have been advanced " to eliminate air blowing. 3. Equipment, materials contamination a. Tank linings. Adequate inspection of linings is not possible without tank entry and cleaning of all interior surfaces. This has not been possible. Prevailing opinion is that the blend tanks have consider able amounts of exposed steel. In the past, exposed tank steel has DS\N 310248 STLCOPCB4070235 12 proved to be heavily oxidized. This oxide is dissolved in the Aroclor, requiring removal by earth. This has been assumed to account for the blend tank degradation observed in past studies. (See above III A-2.) Steel tanks, with fragile zinc-tin linings should be replaced by aluminum or stainless steel. b. Pipe lines. Pipe lines from stills to storage tanks are galvanized iron with threaded joints and a few flanged joints. Much of it is very old. The prevailing opinion is that much of the interior galvanizing has worn away. Threaded joints inevitably expose Aroclor to bare steel. These lines are adequate to produce 500 x 10^ Min. Aroclor but are probably involved in the high level degradation from tank to tank proved by this survey. They should be replaced with aluminum. r c. Pumps and valves. Bronze and stainless steel are specified and are apparently used. Peerless pumps on process tanks and Durco pumps on storage tanks #5. arid #6 have mechanical seals with Teflon inner rings. The Blackmer pumps on the still receivers and the Taber submerged pumps on other storage tanks have packed seals. Valves are Nordstrom plug valves and packed stem gate valves. The plug valves are not lubricated. (Some silicone lubricant has been used very infrequently to open a frozen valve.) Metals in these fixtures are acceptable by present standards. Packing material is not approved. and is discussed below. d. Packings and gaskets. Packing is Raybestos-Manhattan #365; gaskets are Garlock 7021 or its equivalent, SEPCO #200. Appendix, H gives specifications of these materials. The packing is lubricantimpregnated, permitting hydrocarbon contaminant to enter the Aroclor. Hydrocarbons are extremely detrimental to electrical Aroclor quality. Gasket material is a high-sulfur, SBR-bonded material which is not approved for chlorinated aromatics by the manufacturer. J^f present technology is inadequate to specify inert materials here, research should be initiated to provide such information. e. Drumming facilities. The plant study did not show a drop in resistivity in loading. However, significant sources of contamina tion were observed here which could cause serious harm. Appendix I presents photographs of drum loading equipment. Material from the loading line passes through a sac filter, into a funnel resting in the bung hole* Filling continues until gross weight is reached, the loading line is swung over, a drip can suspended under the sac filter, the bung cap screwed in and the drum moved out. During - loading the bung cap rests on a heavily-fouled plywood cover used to protect drum stencils from splashed Aroclor. At this time the drip can rests in a pool of dirty Aroclor on an old drum lid. While drums are being shifted the loading funnel rests on the old drum lid and, in turn, supports the plywood cover. Occasional contamination inevitably enters from contact of drip can with inside of funnel, from contact of bung cap with the fouled plywood cover, and from contact of loading funnel with contaminants on the old drum lid. OS* STLCOPCB4070236 13 During the interim between loadings the funnel and plywood cover rest on the drum lid as shown, in the open, with no cover. It must be considered that HC1 fumes from the blowing tanks is frequently strong in this area; all exposed surfaces are fouled with chlorides. The loading area is open to the west where drums are laid out, stencilled, and the bungs loosened for loading. During rains with even slight wind from the west this area is soaked, with serious risk of admitting water into the drums. Also, the exposed funnel and plywood drum cover are open to rain. The second sheet of Appendix J presents pictures taken by J. G. Bryant during such a rain while loading was being carried out, with rain water drops plainly visible on funnel and plywood.and a pool of water on the old drum lid. - The entire drumming operation needs to be automated with adequate protection. Consideration should be given to a preliminary drum rinse with hot Aroclor, followed by nitrogen pad during drum loading. Anniston PR&D work has indicated a drop in high-level resistivity by merely pouring through the air. Subsurface transfer of Aroclor to the drum would prevent this. Some of these improvement are already being considered by Anniston PR&D. f. Drum materials. J. G. Bryant has done extensive research on drum linings, finding that present double phenolic linings crack in service, exposing Aroclor to steel. This work is reported oh Job 322, Reports #11 arid #7, J.liG. Bryant. It is significant that foreign competitive material with very high resistivity is protected by galvanized drums. B. Sampling Factor in Degradation of Reported Quality 1. Sample containers - Samples are sent to the laboratory in brown glass jugs with polyethylene-lined phenolic caps. Before taking the sample, the jug is rinsed twice with the sample material, cap in place. This study has shown that sampling errors far exceed analytical. A sample container is needed which can be purged with hot Aroclor, remain sealed and padded with N until filled with sample. A sampling bomb of aluminum or stainless, connected to the circulating line with hand disconnects and valves, purged by continuous circulation, then removed and sent to the laboratory would fulfill this need. The laboratory would force out the sample under reduced N pressure, returning the ^-filled bomb to the plant for re-use. 2. -Storage tank sampling - This is the greatest single error in sampling testing procedures. It undoubtedly reflects the exposed conditions of storage tank sampling and the heavily-fouled condition of the sample port valves and all other surfaces in the area. There is no roof to protect the sampling location from rain. Use of the sampling bomb described above would completely eliminate this problem. DSW 310250 STLCOPCB4070237 - 14 - 3. Drum samples - There is no provision at present to obtain a representa tive sample of a drum lot. A sample and reserve sample are generally taken -- each comes from a different single drum. Obviously this fails to provide a valid measure of lot quality. It also fails to monitor drum cleanliness. It exaggerates the sampling error. The sampling bomb described above should be connected through a sample line to what ever drum loading system is eventually installed, using:a-small peri staltic pump and silicone tubing. This would continually sample a small portion from all drums during loading, making the bomb sample a valid lot composite. 4. Tank car samples - The sample system above, consisting of bomb sampler and sampling pump should be installed to obtain a sample during tank car loading. Homogeneity is not a problem, but this system will best protect quality. C. Analytical Factor in Degradation of Reported Quality Test method. Three test methods are pertinent to this study. All use identical test cells, resistance meters and measurement procedures. Essential differences exist in other features treated below. 1. PR&D method - Cells are given routine cleaning, then repeatedly rinsed and tested with high quality Aroclor until a maximum is reached. Verification of cell purity is shown by increase of resistivity when cell is stored in Aroclor overnight. The cell is not cleaned again until low quality material is tested, after which it must be re-cleaned and conditioned to permit high values. This method, its history and supporting data is given by Report #15, Job No. 370-322, J. G. Bryant, 5/3/67. This method has given average values higher than the special lab method by a factor of 1.8 and is the most accurate and precise test available at the Anniston Plant, 2. Special laboratory test - This procedure, used throughout the present study, is identical to the PR&D test except in personnel, environment, and cell preparation. One highly talented analyst performed all tests; all were conducted in the routine laboratory environment; cells were given the routine cleaning before each test, with no conditioning. 3. Routine analytical lab method - This method is identical to the special test above except in personnel and cell preparation. All analysts make the test; the cell is cleaned only when the test falls below 1000 x 10y on regular material (for export grade Aroclor a clean cell is always used) . Accuracy and precision of this method is inevitably at a lower level than the above, thus easily explaining well-known examples of widely divergent checks on the same material. The present study has clearly demonstrated the superiority of the PR&D method. This superiority may be the result of cell cleaning, test environ ment, or personnel. A simple statistically designed study can separate these variables and should be carried out to place plant testing at its maximum practical capability. The design of this test is given in Appendix K. qS\N w0251 STLCOPCB4070238 15 V. SUMMARY CONCLUSIONS RELATED TO CAUSES OF LOW RESISTIVITY A. The answer to the first fundamental question advanced under "I - Objectives and Scope" is affirmative. We do make Aroclor at the refining step which is good enough to meet minimum resistivity specification of 10,000 x 10^ ohm-cm. B. Serious degradation does exist from blend tank to storage tank. This may be as great at 10:1, with greatest drop at the highest values. Possible sources of contamination have been identified as: 1. Atmospheric contamination -7 particularly fumes from other plant departments. ' 2. Broken tank linings which permit contact of Aroclor with iron oxide. 3. Very old galvanized iron pipe lines with threaded joints. Iron is exposed to Aroclor and oxidation in threaded joints. Protective galvanizing on inner surfaces has probably been lost. i 4. Unacceptable packing in pumps and valves. 5. Unacceptable gaskets in equipment and flanged joints. 6. Drum loading facilities seriously exposed to weather, plant fumes, and general contamination. C. Variability (error in reproducibility) was found to be an approximate direct function of the resistivity level and hence must be expressed as a function of R. Checks between duplicate tests will tend to be twice as far apart at 20,000 as compared to 10,000. A test which will be reproduced at + 2,000 at the 10,000 level may show + 4,000 at the 20,000 level. D. Sampling error was found to be the single greatest source of imprecision in reported resistivity data. This numerically amounts to: Storage tank sampling:: std.o.dev. ' .43R Drum and drum sampling : std. dev. m .32R It would be necessary to sample four times, take one test on each sample and average all four to bring the sampling error equal to the error of one analytical test. E. Analytical variability is high. Even under the carefully controlled conditions of the study, the standard deviation of analyses was .25R. This will make duplicate tests on one sample fall within 43.000 to 57.000 at 50.000 level and 4.300 to 5.700 at the 5.000 level. Precision of the routine control test would be expected to fall below that of the controlled study. The above values would probably be ceiling for the routine control test. STLCOPCB4070239 - 16 - F. The PRScD test method gives resistivity values almost twice as high as the special laboratory method used in the study, and hence is much more accurate, especially at the high levels demanded or future production. This advantage would be expected to be higher in comparison with the routine Control procedure. ' DSW 310253 STLCOPCB4070240 17 VI. SUMMARY OF RECOMMENDATIONS TO OBTAIN FUTURE REQUIRED HIGH RESISTIVITY LEVELS A. Related to Plant Equipment Changes (some recommendations have been in progress). 1. Replace present zinc-tin lined tanks with aluminum. 2. Replace existing galvanized threaded pipe with aluminum. 3. Replace present batch earth treatment with the continuous treatment column. 4. Close up the entire system and provide a nitrogen blanket. 5. Initiate a PR&D study to determine feasibility of water washing crude Aroclor to eliminate air blowing and lime in the still. B. Related to Materials in Contact With Aroclor 1. Initiate a research study to select suitable packing and gasket materials. 2, Complete the well-advanced PR&D search for acceptable drum linings. C. Related to Analytical and Sampling Capabilities 1. Design and install a bomb sampler for use in storage tank, drum, and tank car sampling under nitrogen, 2. Execute a carefully designed study to determine what factors in the PR&D test method are responsible for its superiority (cell preparation, personnel, environment). These findings can then be used to upgrade the routine lab test. W. B. Dunlap C>sw 310254 STLCOPCB4070241 ^ssPP?rV***; /f 1 $ ;S V *; .f i*^ '1'..?- - : ' jY ';**' S' - .-,`rs.. " . >i ::fr?s,::^-;:^" v4- :\; /.w.; '-lr. tp*.' *"*r'* '?-- 7<# y'vT" 7 'V V-. '"':V V;.jTT *&P.; ;,; :; y ..v V...; ' ..... ...W I ` 1 '..mi*. - '^''-U*C X |f'' STLCOPCB4070242 18 Appendix B Batch-Sample Data Summary Batch Source Electrical Tests 60 1000 Resist. Cycle Cycle (x 109) P.F. P.F. D.K. Previous Material in Lines or tank Aroclor Batch No. Notes 535 B.T. #3 14,900 .37 F.T. #2 5,600 1.24 S.T. #2 before 4,000 2.40 transfer After transfer #535 4.000 2.35 Before drumming 1,700 2.67 Drums - early drmg. 2,600 2.32 middle drmg. 3,400 1.92 late drmg. 1,800 2.63 .11 4.82 1242 .17 4.84 1242 .24 4.78 1242 .24 4.84 ,25 4.72 .24 4.86 1254 .20 4.82 .24 4.85 534 Treated 2 hrs. 534 534 1021 Lot 123 536 B.T. #3 F.T. #2 S.T. #2 Sple. 1 Test 1 Test 2 Sple..' 2 Test 1 Test 2 23,100 15,900 .34 .88 (4,400 1.59 (6,200 1.36 (2,900 2.55 (4,000 1.08 .11 4.74 1242 .15 4.78 1242 .18 4.83 1242 .17 4.83 .25 4.84 .20 4.80 535 #5 S.T. Treated 2 hrs. 536 537 B.T. #3 F.T. #2 S.T. #2 Sple. 1 Test 1 Test 2 Sple. 2 Test 1 Test 2 12,000 5,600 .34 .99 (3,200 1.89 (2,000 2.27 (5,200 1.54 (5,300 1.45 .10 4.85 1242 .15 4.90 1242 .21 4.83 1242 .22 4.90 .18 4.80 .18 4.77 536 Treated 2 hrs. 536 536 538 B.T. #3 23,400 .35 F.T. #2 9,300 .70 S.T. #2 before drmg . 2,100 Drums - early drmg. 1.700 3.30 middle drmg. 950 4.26 late drmg. 1,300 2.58 .10 4.87 1242 .13 4.82 1248 1242 .30 4.85 1254 .36 4.80 .25 4.82 537 Treated 2 hrs. 59-60 537 18 Lot 124 539 - B.T. #3 F.T. #2 S.T. #2 2,400 4,800 4,100 1.04 1.72 1.20 .16 4.84 1242 .18 4.85 1254 .17 4.90 1242 538 Treated 2 1/4 hrs 231-32 538 540 B.T. #3 F.T. #2 S.T. #2 5,200 19,900 5,200 .96 .57 .87 .15 4.88 1242 .11 4.84 1242 .14 4.82 1242 539 Treated 4 3/4 hrs 539 539 DSW 310256 STLCOPCB4070243 19 Appendix B (continued) Batch Source Previous ______ Electrical Tests__________ Material in Notes 60 1000 Lines or tank Resist. Cycle Cycle D.K. Aroclor Batch (x 109) P.F. P.F. No. 541 B.T. #3 9,100 F.T. #2 7,400 S.T. #2 (11,100 PR&D check (10.800 .32 .77 .70 .36 .10 4.86 1242 .13 4.89 1248 .13 4.81 1242 540 Treated 2 hrs. 61-62 540 542 B.T. #3 4,500 .97 F.T. #2 28,700 .37 S.T. #2 ( 900 3.00 PR&D check ( 9,200 .36 .13 4.82 1242 .11 4.85 1254 .27 4.76 1242 541 Treated 3 hrs. 235- 36 541 543 B.T. #3 8,400 F.T. #2 14,000 S.T. #2 before drag;.(7,200 PR&D check (10,000 .47 .76 ,70 .42 .11 4.89 1242 .13 4.86 1254 .13 4.85 1242 542 Treated 2 1/2 hrs. 237- 38 542 Drums - early drag. 4,900 middle drag. 4,600 late drag. 11,300 1.00 1.20 .63 .15 4.85 1254 .17 4.86 .12 4.82 19 Lot 1006 544 B.T. #3 F.T, #2 S.T. #5 before transfer PR&D check after transfer #544 5,300 4,100 (1,900 .85 .90 1.77 (2,900 1.02 2,900 1.04 .14 4.90 1242 .14 4.82 1242 .19 4.88 1242 .15 4.86 1242 543 Treated 3 1/4 hrs. 543 #5 S.T. S.T. heel #5 S.T. 545 B.T. #3 8,900 .54 F.T. #2 15,800 .47 S.T. #5 3,200 1.41 Tank Car MONX 8606 3,300 1.47 .12 4.87 1242 .11 4.83 1242 .18 4.77 1242 .19 4.82 544 Treated 2 1/2 hrs. 544 544 Lot 125 546 B.T. #3 12,800 F.T. #2 . ( 8,300 PR&D check (20.700 S.T. #5 ,, 4,000 .31 .64 .71 .11 4.77 1242 .12 4.85 1248 .14 4.82 1242 545 Treated 2 hrs. 65 545 547 " B.T. #3 F.T. #2 4,800 .58 3,600 1.27 .12 4.85 1242 .17 4.77 1254 546 241-242 DSW 310257 STLCOPCB4070244 - 20 - Appendix C Progress of Resistivity Through the System (Analytical Lab Data Only - PR&D Not Involved) Batch Blend Filtrate Tank Resist. (Gal.) (x 109) Storage Tank #2 #5 (Gal.) Resist. (Gal.) Resist. Tank Car or Drums Storage tank heel 535 14,900 (2,130) 5,600 (4,900) (7,000) 4,000 4,000)] 1,700}] Lot 123 "(Drums - (2600 (3400 (18Cfo (Ave. 2600 Storage tank heel 536 23,100 (2,040) 15,900 537 12,000 (1,900) 5,600 538 23,400 (2,620) 9,300 ( 450) above (2,480) 1(4,400 6,200 [2,900 4,000 Ave. 4,400 (4,380) [(3,200 1(2,000 [(5,200 (5,300 Ave. * 3,900 (7,010) 2,100 Lot 124 ""(Drums (1700 ( 950 (1300 (Ave. * 1320 Storage tank heel 539 2,400 540 5,200 541 9,100 542 .4,500 543 8,400 (2,050) (2,080) (2,350) (1,920) (2,410) 4,800 19,900 7,400 28j700 14,000 0 (2,050) (4,130) (6,730) (8,380) (10,793) 4,100 5,200 11,100 900 7,200 Lot 1006 ' (Drums (4900 , (4600 [11,100 . _(Ave. = 6900 Storage tank heel 544 ' 5,300 545 8,900 (2,480) (2,170) 4,100 15,800 Storage tank heel 546 12,800 547 4,800 (2,010) (2,170) 8,300 3,600 (5,960) (8,130) (10,300) 1,900 2,900 P.200 13,200 ""(Tank Car -{(3300 (1,780) (3,790) (5,963) t above 4,000 DSW 310258 STLCOPCB4070245 - 21 Appendix C (continued) Batch Blend Filtrate Tank Resist. (Gal.) (x 1(K) Storage Tank _________ #2#5__________ (Gal.) Resist. (Gal.) Resist. Tank Car or Drums Weighted average resistivities going to storage tank compared to final composite resistivities in storage tank and shipment 535 536-538 539-543 544-545 546 14,900 20,000 6,100 3,200 12,800 4,480 9,810 10,740 5,260 5,900 2,900 2,100 7,200 3,200 4,000 2,600 1,320 6,900 3,300 DSW 310259 STLCOPCB4070246 APPENDIX D Effect of Duration of Earth Treatment and Storage on Plant Batches DSW 310260 fiy^/55 STLCOPCB4070247 Appendix E Recapitulation of Data Arranged for Statistical Analyses of Variability Sampling, Testing, Drums .. . .. . LotSampleTestResistivity601000D.K. P.F. 536 1 1 4,400 1.59 .18 4.83 2 6,200 1.36 .17 4.83 2 1 2,900 2.55 .25 4.84 2 4,000 1.08 .20 4.80 537 1 1 3,200 1.89 .21 4.83 2 2,000 2.27 .22 4.90 2 1 5,200 1.54 .18 4.80 2 5,300 1.45 .18 4.77 123 124 1006 Early Middle Late Early Middle Late Early Middle Late - Drums 2,600 3,400 1,800 1,700 950 1,300 4,900 4,600 11,300 2.32 1.92 2.36 3.30 4.26 2.58 1.00 1.20 .63 .24 .20 .24 .30 .36 .25 .15 .17 .12 4.86 4.82 4.85 4.85 4.80 4.82 4.85 4.86 4.82 DSW 310261 STLCOPCB4070248 Appendix F Duplicate Tests by PR&D Method (Resistivity values obtained by multiplying table value by 0.795 x 10^) Test 1 Test 2 2/15/67 #3 22,000 22,000 2/16 2/30 #4 24,000 19,000 #1 16,500 14,500 #2 14,000 11,000 2/22 #1 43,000 37,000 #4 30,000 35,000 2/27 #1 31,000 34,000 n 29,000 32,000 #3 27,000 40,000 #4 25,000 40,000 3/1 #1 22,000 33,000 n 21,000 27,000 #3 24,000 43,000 3/8 #3 25,000 30,000 3/9 #1 60,000 50,000 #2 32,000 27,000 #3 25,000 28,000 3/15 - #1 26,000 21,000 n 22,000 17,000 #3 23,500 15,500 3/17 #1 12,500 12,500 #2 14,500 14,000 #3 12,500 18,000 3/20 #2 14,500 Note: Test 1 and Test 2 made vi:. different cells 14,000 ^0262 STLCOPCB4070249 Da te --> Shift Batch V 535 536 537 538 539 540 541 542 543 544 545 546 547 Appendix G Forward Flow of Material Related to Relevant External Factors 5/4 5/5 5/6 5/7 5/8 5/9 5/10 Wind Directiot Toward Blend Tank Sampler Bo Ri' Run Sequence on #3 Still m N NW W W NW W NW M Bo Ro Ri Bh Bo K Ro Ri H E-} S- SEH NE-} E NE NE E SE SE ES P K P Bo K P Bo Ri H M Ri H M Ri Bo Ri DSW 310263 STLCOPCB4070250 APPENDIX H SPECIFICATIONS ON PACKING MATERIALS R/M SQUARE PLAITED BRAID PACKING R/M No. 365 (Non-metallic) R/M No. 365-R (Rings) R/M No. 366 (Wire-inserted) R/M No. 366-R (Rings) f PACKAGING SPECIFICATIONS Approx, Continuous - Approx, lbs. per Coils lbs. per Feet Size box per box 100 ft. per box '/" Spools 1% -- y,4" Spools -- zv. -- %" Spools -- 4'/j --* y,4" io % 7 5% 200 . %" n% 5 10% 110 * * 7/u" 12% 5 12% 100 ' *\ 'A" 12 /u" 12 5 % 72 * A 20% 59 , %" 12 A 31 - , 40 / %" 12% 3 34 36 %" 14% 3 50 28% 1" 10% - 2 56% 17% . L. DESCRIPTION: Constructed of asbestos yarns, plaited braid, thoroughly impregnated with special, heavy ' bodied lubricant and graphite, resulting in soft, pliable low friction packing. Available as R/M No. 366 with wire insertion. ' .. SERVICE RECOMMENDATIONSi A general purpose pack1 inp for use against steam, water, air, oil and caustics. LdO' O 'l'oc ic~ .O' ; Particularly efficient as centrifugal and rotary pump packing. Use R/M- No. 365-157J-C lubricant for use against gasoline and light oils. STANDARD SIZES: %" to 1%" in increments of 1/16". standard PACKAGES: Coil form, boxed; and 25 lb. reels. %", 3/16", %" available on % lb., 1 lb., 5 lb. spools. 5/16", %" available on 1 lb. and 5 lb. spools. 5/16" and up, coil form, boxed, dSVM 310264 STLCOPCB4070251 - APPENDIX H Continued GARLOCK CCJVi PRESSED ASBESTOS ' f SHEET GASKETING . Garlock Compressed Asbestos Sheet combines white or blue asbestos with a variety of binders . . . bonded under pressure and vulcanized into a homogeneous sheet. All styles are tough and durable with sufficient compressibility to seal properly, yet resist plastic flow under heavy bolt loads. Ideal for high pressure/high temperature applications employing rigid flanges with adequate bolting capacity. For basic information covering four popular styles, see chart below. Contact your Garlock representative on other styles. 7021 MATERIAL TENSILE STRENGTH APPLICATION EQUIPMENT THICKNESSES (inches) SHEET SIZES (inches)NOMINAL WEIGHTS* (pounds per sheet) Compressed Chrysotile asbestos witl^SBR binde^ With grain 6800 psi. Across grain 2500 psi. Average 5000 psi Hot oil service at high temperature--steam or hot gases up to 700*F On flanges of cracking stills and other refinery equipment, pipe lines, compressors, internal combustion engines; other high temperature applications 1 /64.1 '32. 1 /I6, 3/32,1/8, 3/16, 1 /4 40x40 60 X 60 i 60 X 60 j 75x75 | 120x120 | 150x150 6.0 9.4 ! 13.5 j 21.1 j 54.0 j 84.4 STLCOPCB4070252 APPENDIX I Sources of Contamination in Drum Loading Actual loading - drip can on spout, cap on plywood drum cover, funnel in next drum to be filled. Between-loading exposure of equip ment - Funnel resting in pool on old drum lid, plywood drum cover on top. Loading through soc filter in funnel. Cap resting on ply wood drum cover. Drip can on old drum cover. DSW 310266 Actual loading - drip can resting on old drum lid, cap on plywood drum cover. STLCOPCB4070253 APPENDIX J Exposure of Drum Loading to Weather Weather exposure in Drum loading. Showing open end of building toward west. Actual loading of drums after rainstorm - water droplets plainly visible on funnel and plywood drum cover. DSW 310267 STLCOPCB4070254 APPENDIX K Laboratory - PR&D Study Designed to Separate Factors Involved in PR&D Superiority Location -- Personnel * PR&D 'Oo MU a> a Lab PR&D BW DY 11 2 12 3 Lab BW DY 12 3 33 2 Numbers Refer to Samples 1, 2, 3 From Same Drum 12 tests required, using 3 samples from the same drum, four tests per sample. No repetition is necessary. Each factor evaluated between averages of four tests. Possible interaction can be evaluated. Comparisons between conditions will be evaluated against analytical precision information generated by this Study. Note - sample bottles and sampling technique must be very care fully planned to minimize the sample-to-sample differences unavoidably introduced by using 3 Samples. DSW 310268 STLCOPCB4070255 Appendix L Organic Division Quality Goals for Electrical Grade Aroclor Color Water Resistivity Power Factor @ 60 cycles Thermal Chemical Chloride Stability 10 APHA Max. 20 ppm Max. 10,000 x 10^ ohm-cm Min. 0.2% Max. 0.1 ppm Max. OS**'**' STLCOPCB4070256