Chemical Stack Losses from Soda Recovery Furnaces HOWARD S. GARDNER‘ AND ROY P. WHITNEY School of Chemical Engineering Practice, Massachusetts Institute of Technology, Cambridge, Ililass.
C
HEMICAL stack losses from black liquor recovery furnaces constitute a serious economic loss in producing pulp by either the soda or sulfate process. Each of these processes uses a cooking liquor in which the chief constituents are sodium compounds, and the soda recovery cycle in the pulp mill is operated to recover these compounds from the spent cooking liquor and convert them into materials required for re-use in the digesters. The soda recovery cycle begins by removal of the spent cooking liquor from the digested pulp as a “black liquor,” usually by countercurrent mashing in order to maintain as high a solid content as possible. The black liquor, containing organic matter dissolved from the wood in digestion, is then concentrated by multiple-effect evaporation to approximately 50 per cent solids. This strong black liquor is fed to a recovery furnace where the residual water is evaporated, the organic material is burned, and the sodium compounds are converted principally to sodium carbonate. In addition, in the sulfate process, make-up sodium sulfate is reduced to sodium sulfide. Depending upon the type of equipment used, the recovery furnace discharges either a black ash or a smelt, which is leached or dissolved. The resulting “green liquor” is causticified with lime to convert most of the sodium carbonate to sodium hydroxide for re-use in the cooking process. Profitable operation of an alkaline pulping process requires that the soda recovery be maintained a t a high figure, usually 90 per cent or better, when based on the soda charged to the digesters. Certain losses are inevitable, but careful attention to operations can do much to keep them a t a minimum. One of the major sources of loss in the recovery cycle is that of chemicals in stack dust from the recovery furnace, and this loss probably accounts for one third to one half of the total disappearance of soda from the cycle. The presence of dust in the flue gas can be attributed to:
The black liquor in the soda mill was burned in a battery of four conventional rotary burners (Figure l),each 15 feet long and 9 feet in external diameter, rotating a t 1.2revolutions per minute. In addition to burning the black liquor, each rotary burner was fired with flue gases from a small auxiliary coalburning furnace, located a t the discharge end of the rotary.
In producing pulp by the alkaline processes, chemical stack losses from the recovery furnaces constitute one of the major economic losses. This paper reports the results of stack loss measurements from a battery of four rotary furnaces burning black liquor from soda pulp manufacture and from a modern spray-type stationary furnace burning sulfate black liquor. Sampling traverses were made in each flue, and the samples were withdrawn a t duct velocity a t each point. Dust was effectively absorbed in water by a series of bottles, in which the gas stream was broken up by being passed through finely woven cloth bags. Experiments indicated that this method of absorption was more satisfactory than the use of Alundum crucibles or towers packed with glass wool. Losses from the rotary furnaces were mainly sodium carbonate, amounting to 35 pounds per ton of pulp production and representing 3.3 per cent of the chemical input to the furnaces. Losses from the spray furnace were mainly sodium sulfate, 98 pounds per ton of pulp or 5.9 per cent of the chemical input. Dust loadings w-ere 0.44 grain per cubic foot of Bue gas a t the rotary furnaces, and 1.1 grains per cubic foot a t the spray furnace. The value of the dust lost from each of the installations amounted to approximately $30 per day.
1. Entrainment of dust particles in the gas stream from the solid mass within the furnace. 2. The presence in the gas stream of minute droplets of black liquor which are dried and burned to ash particles so small that they remain suspended. 3. Volatilization of sodium compounds (1, 2 ) at the high temperature of the furnace.
Flue gas from each rotary passed through a fire-tube waste heat boiler and was then discharged into a common round flue leading t o an induced draft fan. Black liquor in the sulfate mill was burned by spraying into a modern Babcock and Wilcox-Tomlinson furnace (6). Flue gases leaving the boiler section of the furnace (Figure 2) passed vertically downward through an air heater and then
Because of the difficulties involved, stack losses are not normally measured in mill operation, This paper presents the results of two tests to determine stack losses from rotary furnaces in a soda pulp mill and from a stationary spray furnace in a sulfate mill. 1
Present address. University of Rochester, Rochester, N. Y.
181
182
INDUSTRIAL AND ENGINEERING CHEMISTRY
through a scmbbing chamber where they came in contact with & coarse spray of dilute black liquor, After the scrubher, an induced draft fan discharged the flue gas through a rectangular flue to the stack.
Test Procedure Measurement of stack losses requires a determination of the total vdume of flue gas and of the concentration Of the stack dust suspended in the gas. To determine the dust concentraSAMPLES TAKEN
VOL. 31, NO. 2
volume of sample withdrawn was calculated from flowmeter readings and corresponding temperatures and Pressures. Total sample volumes for the 6-hour runs varied from 25.6 to 26.7 cubic feet (corrected to 32" F., 1 atmosphere) at the soda pulp rotaries and from 30.6 to 35.5 cubic feet a t the sulfate pulp spray furnace. The dust was removed from the gas stream by a series of three absorption bottles, each a one-quart, widemouthed glass jar, about one third full of water. As indicated in Figure 3, the glass
~~~~a,ex,,"n$~$.?$~$'$ th:JZ:$'' g,t"dhe"~ ~ ; l ~ & " cotton cloth (No. 60 Cambric, Londsdale), tied over the end of each tube. At the end of each run, the sampling tube was washed carefully with water and the washings were added to the first bottle.
The amount and composition of the materials absorbed from the gas stream were determined by chemical analysis. Solid material present in the solutions was very slight and was removed by filtration. In cases where solutions from the first bottles of the absorption train were sufficiently colored t o interfere with end points, the solutions were decolorized by boiling with animal charcoal. It was also found that objectionable color could be eliminated by evaporating to dryness and igniting. For each 6-hour run, only the principal constituent of the dust was determined. I n the dust from the sulfate spray furnace, total sulfate was determined gravimetrically as barium sulfate. Composite samples representing the 24-hour test period were analyzed for the lesser constituents. Chloride was determined by titrating neutral solutions with silver nitrate, using potassium chromate as an indicator. Since solutions were acid to methyl orange, there was no carbonate present. Sodium bisulfate was calculated as equivalent to the acidity determined by titration with sodium hydroxide to a methyl orange end point. The absence of sulfite was indicated by the qualitative iodo-starch test. In the soda mill dust, aarbonate was determined by titration with hydrochloric acid to both the phenolphthalein and methyl orange end points. Composite samples were analyzed for sulfate and chloride by the methods outlined above. '
FIGURE1. ROTARYFURNACE INSTALLATION
tion, a sample of gas and suspended dust must be withdrawn from the stream, the dust separated and analyzed, and the gas volume measured. The problem of sampling a gas stream carrying a suspended solid requires that the sample be withdrawn by a tube facing against the direction of flow, a t a velocity equal t o that at which the stream is flowing. If the velocity of sampling is less than stack velocity, the inertia of the dust particles should result in too high a dust concentration. Conversely, a high rate of sampling should result in a dust concentration which is too low. Another factor to be considered is the possibility of nonuniform distribution of the solid particles over the entire cross section of the duct at the sampling point. Each of the two tests lasted 24 hours and consisted of four consecutive 6-hour runs. Samples were withdrawn from the main flues continuously at stack velocity by the apparatus illustrated in Figure 3. A Pitot tube and sampling tube were fastened together, with the tips 0.5 inch apart. For each 6-hour run the tubes made two complete traverses of the duct in order to reduce the possibility of error due to nonuniform distribution of flow and dust particles over the sampling cross section. At the soda rotary furnaces, measurement and sam ling stations were at ten points representing equal areas on bot! horizontal and vertical diameters of the circular flue, 66 inches i. d. The tube assembly was held at each station for 10 minutes on each traverse. The 33 x 30-inch rectangular flue from the sulfate spray furnace was divided into twenty-five equal areas, and the sampling assembly was held at the center of each area for 8-minute periods. The velocity of flue gas at the sampling point was measured by a Pitot tube, using an inclined manometer. The rate of sampling was indicated by a capillary flowmeter, connected t o a source of suction, as shown in Figure 3. Previous to the test, the flowmeter was calibrated by the volumetric dis lacement of air. By converting the calibration to assumed condttions for the gas during the test, a graphical relation was developed between Pitot tube reading and the corresponding flowmeter differential which would give a sampling velocity equal to the duct velocity. During the test the Pitot-tube manometer was read at intervals of 2 to 4 minutes, depending on the variation in flow, and the rate of sampling was regulated according t o this relation. The total
1 1
DISSOLVING TANK
I
FIGURE 2.
SPRAY FURN.4CE INST.4LLATION
Orsat analyses of the stack gas for carbon dioxide, carbon monoxide, and oxygen, and measurements of the temperature and pressure of the flue gas were made to provide data necessary for calculation of the losses.
~ ~
FEBRUARY, 1939
INDUSTRIAL AND ENGINEERING CHEMISTRY
183
Figures for plant production and for chemical input (sodium compounds) to the furnaces were obtained from the plant operating records.
sure drop through the cloth filter bags makes them better suited for test work. The glass wool towers are less effective in absorbing the dust, but it is possible that this might be remedied by the use of towers of larger diameter. This submerged bag method of dust removal would probaResults bly be unsatisfactory with insoluble dusts, such as fly ash or The average results of the stack loss measurements are mineral dusts, owing to plugging of the cloth. However, for given in Table I. Each of these values represents the 24-hour measurement of stack losses in the wood pulp industry, where period covered by the test, and is the average of four consecuthe principal constituents of the dust are water soluble, the tive 6-hour runs. method seems well adapted. The presence of chloride in the dust from the soda mill (Table I) may be explained by the fact that soda losses at the TABLEI. CHEMICAL STACKLOSSESFROM SODARECOVERY mill are made up by the addition of electrolytic caustic soda, which contains a small amount of salt. Consequently, FURNACES sodium chloride tends to build up in concentration in the Sulfate Soda spray cycle until the losses are equal to the additions of salt to the Rotaries Furnace system. Since the ratio of sodium chloride to other sodium 4 1 Number of furnaces 115 Max. capaoity tons pulp/24 hr. 75 compounds in the black liquor or in the black ash is much less 100 Production &e, tons pulp/24 hr. 44 than it is in the stack dust, it seems probable that sodium Dust loss: 35 Lb./ton pulp 98 chloride is more volatile than sodium carbonate under the 41 19 Lb. of equivalent NalO/ton pulp 3.3 5.9 % of chemical input (NarO) to furnace operating conditions. The most probable source of the 0.44 1.1 Dust loading, grains/cu. ft. (flue oond/tioni) sodium sulfate in the soda mill dust lies in the formation of Volume of flue gas, CU. ft. (flue oonditions)/ton 640,000 560,000 sulfate from sulfur originally present in the coal used to fie None 71 the auxiliary furnaces of the rotaries. It seems probable that 21 87 sulfur dioxide from the burning of the coal reacts with the ..... 11 8 2 basic sodium compounds suspended in the furnace gases, None ..... forming sodium sulfite which is oxidized to sodium sulfate at the high temperature in the presence of excess air. The TABLE11. AVERAGEOPERATINGDATADURING TESTON RECOVERY FURNACES
Gravity of black liquorb Bb. Temp. of blaok liquor, F. Flue gas temp. at sampling point, Orsat analysis of flue gas, %
co2 02 co
Nz Excess air, %
O
F.
Sulfate Spray Furnace 32.6 219 158
Sods, Rotaries 38.4 140 480 Test Station 4.7 16.2 0 79.1 360
Furnace Outlet 15.8 4.2 0 80.0 25
Y
Test Station 8.8 11.1
FLUE
0 80.1 109
The average values of significant operating data obtained during the runs are presented in Table XI. It is believed that the test method gives reasonably complete removal of stack dust from the gas sample. For the four 6-hour runs on the soda rotary furnaces, the third bottles in the absorption train were found to contain 1.6, 0.5, 0.5, and 0.5 per cent of the total absorbed for each run. The same method of dust absorption was used in the test on the sulfate spray furnace, and a similar value for the amount absorbed in one run in the third bottle was 0.6 per cent. Subsequent to the tests, the effectiveness of various methods of absorbing the dust was further investigated. This work was carried out on the flue from the soda rotaries and included the use of cloth filter bags submerged in water, 35-cc. Alundum thimbles submerged in water and used in the same manner as the cloth bags, and towers 1.25 inches in diameter and 15 inches tall, packed with glass wool. Absorption trains composed of various combinations of these units were made up and operated, with the dust absorbed in each unit being analyzed for sodium carbonate only. The results in Table 111 are typical. The cloth filter bags and the Alundum thimbles are equally effective means of dust removal from the gas stream. Also apparent is the decreased effectiveness of absorption as the rate of sampling is increased, which may be due t o entrainment of water from one bottle to the next. The lower pres-
WATER ASPIRATOR TER
FIGURE 3. DIAGRAM OF TESTAPPARATUS
OF DUSTABSORPTION TRAINS TABLE111. COMPARISON
Absorption Train 4 cloth filter bags
2 Alundum thimbles and 1 cloth bag 4 glass wool towers Q
% of Total Na!COa Pressure Drop Rate,of through Absorbed in Unit No: Sampling4 Train 1 2 3 4 CU.ft./hr. In.Ho 7.2 3.4 98.1 1.9 0 .. 95.4 4.6 0 0 11.0 5.3 7.2 0 0 6.4 92.8 14.7 7.6 5.2
14.0 1.2
Corrected to 32' F. and 1 atmosphere.
95.6 88.0
4.4 8.5
0 3.5
.. 0
184
INDUSTRIAL AND ENGINEERING CHEMISTRY
amount of sulfur present in the stack dust is approximately equal to that of the coal fired in the auxiliary furnaces. In measuring stack losses, determination of the rate of flow of flue gas is of equal importance to the problems of dust sampling and analysis. In the duct from the rotary burners, the Pitot tube location was quite satisfactory, and little difficulty was experienced. However, the only available test location at the spray furnace was directly behind a set of three horizontal dampers mounted across the rectangular flue. Because of the difficulty of obtaining accurate measurements with the Pitot tube at this location, the total flow of flue gas was estimated by an additional method. Plant figures for the rate of feed and carbon content of the black liquor were used, together with Orsat analyses of the flue gas at the test station. Assuming a material balance between the carbon in the black liquor and that in the flue gas, the volume of flue gas was calculated as 650,000 cubic feet per ton of pulp. The close agreement of this value with 640,000 cubic feet per ton of pulp as determined by the Pitot tube and reported in Table I must be regarded as fortuitous, for the data and assumptions made in calculating the material balance put the entire calculation on an approximate basis. On the other hand, the general agreement of the two independent methods of flow measurement does indicate that the value obtained is probably reasonably accurate. A further indication of the reliability of the results can be given by the variation in the individual values for the separate 6-hour runs. During the 24-hour period the operation of the furnaces appeared to be normal and fairly steady, and variations in the results calculated from the individual runs cannot be significantly correlated with changes in operation. The four 6-hour values for the dust loss showed average deviations from the mean result of 7 and 9 per cent for the soda rotaries and sulfate spray furnace, respectively. From these facts and from a general consideration of the test work, it is believed that the results are accurate to within 10 or 15 per cent of the true dust loss. It should be pointed out that the stack losses from the spray furnace at the sulfate mill are probably higher than they should normally be. A comparison of the Orsat analyses of the flue gas in Table I1indicates that there was a large leakage of air into the flue gas between the furnace outlet and the dustsampling point near the entrance to the stack. From the decrease in carbon dioxide content and the increase in oxygen content, it is calculated that approximately 40 per cent of the gas a t the sampling point was air leakage into the flue gas. At the time of the test it was not possible to determine the source of this air leakage, but subsequent investigation by the plant operating staff indicated large leaks in the air heater. The effect of this leakage, in addition to reducing the thermal efficiency, was t o increase the flue gas volume and velocity by approximately 70 per cent. Without leakage, the provisions for dust recovery by removal of suspended solids in the air heater and by the black liquor spray in the scrubber would probably have been more effective and would have resulted in somewhat lower dust losses at this rate of production. However, since the Orsat analysis at the furnace outlet indicates that furnace operation was normal, chemical carry-over to this point may be regarded as normal. The loss from the furnace would probably increase because of higher gas velocities if the furnace capacity, then running at 60 per cent of maximum, were increased. Stack losses measured in this test compare well with other published values. Sultzer and Beaver (4) report results from stack loss measurements in sulfate and soda pulp mills. Losses from three sulfate pulp spray furnaces varied from 40 to 85 pounds of sodium oxide per ton of pulp, averaging 65 pounds per ton. Losses from three soda pulp rotary kiln furnaces varied from 30 to 67 pounds of sodium oxide per ton
VOL. 31, NO. 2
of pulp, averaging 43 pounds per ton. These figures may be compared with the losses given in Table I, of 41 and 19 pounds per ton, respectively. Millidge, Taylor, and Heimrod (3) reported data from the operation on soda black liquor of three spray-type modified Wagner furnaces, indicating losses of approximately 58 pounds of sodium oxide per ton of pulp before recovery of 90 per cent of the stack dust by Cottrell precipitation. This represents a furnace loss of 7 per cent of the chemical input to the furnace, at a dust loading of 1.4 grains of sodium carbonate per cubic foot. Wilcoxson (6) reported test data from the first Babcock and Wilcox-Tomlinson furnace, similar in type to the spray furnace which is the subject of this paper. Operating on sulfate black liquor, at a production of 75 tons of pulp per day, the furnace losses were 46 pounds sodium oxide per ton of pulp, or 5.6 per cent of the chemical input. A recent source (6) states that losses reported from rotary furnaces vary from 0.2 to 3.0 per cent of the chemical recovery, or from 2 to 30 pounds of sodium oxide per ton of pulp. The same source mentions a chemical loss of 5 per cent for the Babcock and Wilcox-Tomlinson furnace. Among the factors which have contributed to the higher stack losses from the spray furnace are higher operating temperatures, introduction of the liquor as a spray, and higher gas veldcities throughout the furnace installation. For comparison, the average velocity of the flue gas a t the dust sampling station was 28 feet per second at the rotary furnaces, and 48 feet per second at the spray furnace. When considering the over-all operation of spray and rotary burners, it must be remembered that stack losses form only one basis of comparison. Among other points for comparison are initial cost of equipment and depreciation rate, operating costs, maintenance costs, space requirements, ease of control, degree of chemical conversion effected, and thermal efficiency of the furnace in utiliiing the heat of combustible materials in the black liquor. In a number of these points the modern spray furnaces have a definite superiority. The value of the materials lost in stack dust is great enough to justify considerations of dust recovery. Based on soda ash a t $18 per ton, the dust from the rotary furnaces represents a loss of $0.30 per ton of pulp, or $30 per day. With 96 per cent salt cake a t $14 per ton, the dust from the spray furnace represents a loss of $0.70 per ton of pulp, or $30 per day. At the present time dust recovery by Cottrell precipitation (4) is the method generally used by the industry when losses are high enough to justify the cost of installation.
Aclrnowledgment The authors wish to acknowledge the contributions of students in the School of Chemical Engineering Practice, especially A. W. Barry, J. R. Fitz-Hugh, W. Squires, Jr., and B. H. Wilcoxon, who took part in planning the tests and carried out the experimental work. Thanks are also due the managements of the Advance Bag and Paper Company and the Penobscot Chemical Fibre Company for their cooperation during the conduct of the tests in their respective plants and for permission to publish these results.
Literature Cited (1) Day, G. A., Paper Trade J., 105, 16 (1937). (2) Liander and Olsson, Iua., 1937, 145. (3) Millidge, Taylor, and Heimrod, Paper Trade J., 98, 321 (1934). (4) Sultzer and Beaver, Ibid., 102, 45 (1936). (5) Tech. Assoc. of Pulp and Paper Industry, “Manufacture of Pulp and Paper,” 3rd ed., Vol. 111, Sect. 5, pp. 145, 173, New York, McGraw-Hill Book Co., 1937. (6) Wilooxson, L. S., Paper Trade J., 100, 298 (1935). RBC~IV October ~ D 29, 1938. Presented before the meeting of the Amerioan Inetitute of Chemical Engineers, Philadelphia, Pa.,November 9 t o 11, 1938.