International Sugar Journal, 2002, 104, 1247, pp 500-507 Boiling point elevation of technical sugarcane solutions and its use in automatic pan boiling Michael Saska Audubon Sugar Institute Louisiana State University Agricultural Center Baton Rouge, LA 70803 Introduction Despite the advances in recent years in developing new sensors1,2,3,4 for control of sugar crystallization, boiling point elevation (BPE) remains a simple, robust and inexpensive means for monitoring supersaturation over the full range of massecuite purity, from refinery white sugar to low grade massecuites in the raw sugar mills. In the framework of a general program at ASI on optimization and automation of sucrose crystallization, it was deemed of interest to review the commonly used BPE data found in the sugar literature, and, if warranted, perform additional measurements over the pertinent range of purities, concentrations and temperatures. A considerable body of literature has accumulated since the early work of Claassen in the 1900’s, and yet, on closer inspection, original data, particularly for industrial (low purity) sugarcane solutions are surprisingly limited. Of the major sugar technology texts in use today, Honig5 and VanDerPoel6 rely on the sugarbeet-based data of Spengler7, the Cane Sugar Handbook8 presents Holven’s interpretation of Thieme’s data9, and Hugot10 gives the data of Thieme and a nomograph of Othmer and Silvis11 based on Spengler. Thus the historical, and rather sparse data of Thieme from the 1920’s and the fairly recent equilibrium (non-boiling) measurements of Batterham and Norgate12 that appear to have been so far disregarded by the handbooks but are reportedly used in Australia13 are the only data of note derived from impure sugarcane solutions. The scarcity of the original data is contrasted by the profusion in the literature of interpretations, correlations, smoothing and other manipulations of the few limited experimental points11,14-24 Of the two pilot sized vacuum pans at ASI, the larger, all-stainless steel, one (Table I and Figure 1) is fully instrumented and well suited for measuring the relevant parameters of the crystallization process. Forced circulation is provided by two impellers, one located at the top of, the other at the bottom of the central downtake. Calandria is made of two concentric rings, supplied by pressure-regulated steam, and provided with condensate and incondensible gas outlets. A Honeywell UMC 800 controller in conjunction with a PC-based Honeywell PlantScape®, Version 400.0 package, with an in-house developed automatic sugar boiling software provides the parameter control, automation and data collection and display. Table I: Design parameters of the 200 liter ASI pilot vacuum pan. Pan Design Pan I.D. at Calandria, mm Downtake I.D., mm Dished End Volume, L Graining Volume, L Strike Volume, L Total Volume, L 450 216 20 123 200 661 Calandria Details Circulation System Impeller Diameter, mm Pitch, mm Power installed, kW Stirrer Speed, rpm Tip Speed, m/s Heating Elements Large Ring, OD/ID/Height, mm Small Ring, OD/ID/Height, mm Heating Surface, Large Ring/Small Ring, mm2 Heating Surface/Volume, 1/m Signal Absolute pressure Steam pressure Temperature 1 (below calandria) Temperature 2 (above calandria) Level Conductivity RF- capacitance/resistance Microwave density Agitator load Agitator speed Feed flow rate Steam condensate flow rate 178 76 0.37 233 (at 8 Hz) 2.2 (at 8 Hz) 381/330/216 273/216/210 499,000/342,000 4.1 Instrument type Barksdale, 0-15PSIA ABB TEPI11Ex Flow-temp, PT100 , 3-wire, accuracy 0.01% FS Flow-temp, PT100 , 3-wire, accuracy 0.01% FS Foxboro IDP10 differential pressure Dynamic Systems, Model CTX/AC Fletcher Smith, Digital Duotrac pro-M-tec Theisen GmbH, Model uWA-2.5 Load Control, Inc. PH-3A Power Transducer Teco-Westinghouse FM100 Yokogawa AM-105AG magnetic flowmeter Yokogawa AM-110AG magnetic flowmeter Generation of experimental data. The temperatures of the boiling liquors, at various absolute pressures and purities, were measured during slow concentration over 1 to 3 hours from some 65 to about 80 % refractometric dry solids (WDS) with periodic sampling and WDS determination off-line with a standard Abbe refractometer. The absolute pressure, steam pressure, level and agitator speed were automatically maintained at constant levels, while all signals (Table I) - some unused in this work - were recorded at a frequency of some 3 samples per minute on the 3-1/2” disk drive of the UMC 800 controller (Fig. 2 and 3). Care was taken to assure synchronicity of the off-line WDS measurements with the on-line recordings, as both were later imported in Excel for evaluation and graphing (Fig. 5-7). In order to account for and eliminate or minimize the effect of possible minor systematic errors in measuring either the liquor temperatures or absolute pressure (Figure 4), and in order to at least partially compensate for the theoretical effects on the boiling point of the hydrostatic head and any solution superheat, temperature of the boiling water was recorded as with sugar liquors in separate “water tests” (Figure 3) lasting some 1 to 2 hours each, and interspersed between the sugar liquor tests. Again, the data were recorded and averaged, and a correlation (Figure 4) of the averaged data, viz. temperature v. absolute pressure was then used in the BPE evaluations, rather than the literature data. The difficulties with location of the temperature sensors, particularly in large pans, and the effects of superheat at the heating surface are well recognized and documented in the literature. The temperature of the boiling liquid was therefore measured and recorded at two locations, one under the calandria (tb1), and the other just above it (tb2). A small but rather systematic difference (tb1 – tb2) of about 0.2 to 0.8oC was found between the two, and nearly identical, on average, in the water tests (Figure 3) as well as measurements with sugar liquors (Figure 2). As this difference does not appear related to concentration (viscosity) of the boiling liquid or the heat flux (change of steam pressure from 5 to 3 psi in Figure 3), and therefore to any possible liquid superheat at the heating surface, it comes most likely from an offset of one of the two temperature sensors and is not expected therefore to effect BPE evaluation. The average of the two readings was used throughout this work in evaluation and reporting of the data. The average values from the six water tests were correlated (Figure 2) to obtain tbW = 3.3374Pabs + 40.746 (1) where tbW is the water boiling temperature (oC) and Pabs, the absolute pressure (in Hg). From tbW, and the boiling temperature of sugar liquor (tb), the boiling point elevation is calculated as ∆tb= tb- tbW (2) where tbW comes from equation 1 taking into account the actual Pabs for each recorded data point. The experimental correlation is nearly parallel to and offset by some 0.2 to 0.4oC off the points taken from the literature10 (Figure 4). Three levels of (HPLC) purity (100 WS/ WDS) were included in the measurements, viz. a refined sugar liquor (Q=100), a mill (Louisiana) syrup (Q=85) and, for the lowest level (Q=58), a blend of blackstrap molasses and raw sugar. Data Analysis and Presentation Comparison of experimental data with literature Four equations from the literature were chosen for comparison with the experimental data. An extensive, but limited to pure sucrose solutions, set of literature data was fitted by Starzak and Peacock23 with a five-parameter equation t bW + C SP Q 2 2 1 − RB X S 1 + a SP X S + bSP X S t + 273.15 SP bW ∆t b = − 1 (t bW + C SP ) t + C bW SP 1+ Ln(1 − X S ) BSP ( ) (3) with aSP = -1.0038 ; bSP = -0.24653 ; BSP = 3797.06 oC ; CSP = 226.28 oC ; Q = -17638 J/mole and R (8.3143 J/mole K), the universal gas constant. The scatter of the data at the very high solids contents and temperatures (up to 99 % sucrose concentration!) is very large and was expected to affect the applicability of the equation in the narrow range (7< ∆t b <13oC) of ∆t b encountered in sugar crystallization. The sucrose concentration in equation 3 is expressed as a mole fraction, XS = WS / MWS WS / MWS + (100 − WS ) / MWW (4) Batterham and Norgate12 measured BPE of cane solutions at purities 100, 70 and 41 using the equilibrium (non-boiling) technique of Dunning. Theirs appear to be the only data of note for cane solutions besides the limited set of Thieme from the 1920’s. A correlation ∆tb = ABN tbW + BBN + CBN (5) with ABN = 0.3604 - 2.5681 x 10-2WDS + 6.8488 x 10-4WDS2 – 8.0158 x 10-6WDS 3 + 3.5601 x 10-8WDS 4 BBN = 50.84 - 3.516WDS + 9.122 x 10-2WDS 2 – 1.0492 x 10-3WDS 3 + 4.611 x 10-6WDS 4 CBN = -0.272 - 2.27(Q/100) + 2.542(Q/100)2 + 0.05311WDS(1 - Q/100) was developed over the concentration range of WDS 47 to 84% and tbW range 40 to 75 oC. The absence of boiling avoids complications with superheat and hydrostatic head, but a closer inspection of the presented data reveals significant and systematic deviations between the experimental points and the correlation lines, particularly at high concentrations and low purities, but even in pure sucrose liquors at WDS > 70 %, viz. the only regions that matter in sucrose crystallization. This reliance on the more numerous data for dilute solutions and the Duhring rule in fitting of data, and disregard for the experimental points for concentrated liquors introduced some uncertainty as to its applicability. Nevertheless, the comparison offered by Batterham with Thieme’s data is significant and shows much smaller effect on BPE of purity than thought previously. Hoekstra20 fitted the data taken from Honig5, Batterham and Norgate12, Hugot10 and Lyle24 with the integrated form of the Clausius-Clapeyron equation W DS ∆t b = AH 100 − W DS BH 273 + t bW 100 CH Q 100 DH (6) with the coefficients AH = 0.1379, BH = 0.808, CH = 2.327 and DH = -0.42 , Although no comments on the fit with or deviations from the literature were offered, and an apparent omission in the reported equation of a factor of nd rd ( 273 + t bw ) 2 2 2 ∆t b = a B * W DS bB (W DS − 60) + c B 100 in the 2 and 3 terms of the right0.38 (374 .3 − t bw ) hand side of the equation, and some uncertainty about the absolute values of 25 some of the coefficients , it was nevertheless deemed of interest to compare this correlation as it has apparently found some use within the South African sugar industry. [ ] Bubnik etal22 developed a nine-parameter equation (7) aB = a0+a1Q+a2 Q 2 bB = b0+b1Q+b2Q2 cB = c0+c1Q+c2Q2 and a0 = 1.67525*10-4, a1 = -1.85299*10-6, a2 = 7.92284*10-9, b0 = 2.87764*10-6, b1 = -2.77263*10-8, b2 = 1.44306*10-10, c0 = 1.06668*10-3, c1 = 2.10304*10-6 and c2 = 4.28274*10-8, with the mean difference between calculated and experimental data reported to be only 0.15oC. Despite the claim of multiple sources, the underlying data appear limited to the sugar beet set of Spengler. Fitting of the experimental data The NLIN routine (Table II) of SAS26 version 8.2 was used to find new parameters (Ax, Bx, Cx, Dx) of equation 6. The Hoekstra parameters AH, BH, CH and DH were used as the starting points for the iterative fitting. Table II: SAS code for, and output of fitting of the experimental points. SAS code: data one; infile cards; input WDS ∆t b p tbW Q;cards; 70.60 5.14 5.39 58.73 100 72.60 5.92 5.39 58.73 100 74.80 6.44 5.40 58.77 100 .. proc nlin; parameters A = 0.1379 B = 0.808 C = 2.237 D = -0.42; model ∆t b = AX * (WDS / (100 - WDS))**BX * ((273 + tbW) / 100)**CX *(Q/ 100)**DX; run; The SAS output is in the following table: Parameter AX BX CX DX Estimate Approx Std Error 0.1660 1.1394 1.9735 0.1237 W DS ∆t b = AX 100 − W DS BX Approximate 95% Confidence Limits 0.0556 0.0172 0.2750 0.0137 273 + t bW 100 CX 0.0568 1.1056 1.4331 0.0967 Q 100 0.2752 1.1732 2.5138 0.1507 DX AX = 0.1660 BX = 1.1394 CX = 1.9735 DX = 0.1237 (8) Graph of the boiling point elevation v. liquid temperature. QWDS 100(100 − WDS ) Equation. 8 with the new parameters was used over a range of the variables (purity, supersaturation and temperature) and their increments that were chosen such as to be directly comparable with the frequently used older graphs of Holven carried unchanged from the 8th to the 12th edition of the Cane Sugar Handbook8. qS /W = For each set of the variables tb, Q and y, and taking into account the definitions (9) q S / W (100 − Wsat , p ) WS / WW y= = (WS / WW ) sat Wsat , p q sat QWDS 100 = 100 − WDS WDS − q NS / W (10) q sat = WS , sat / WW , sat Wsat , p / WW , sat , p = 1 − 0.088q NS / W (12) (11) the dry solids concentration WDS in solution was calculated as 100 / WDS = 1.088 + q (100 − Wsat , p ) 100 yWsat , p − 0.088Q 100 (13) where the solubility (g/100 g solution) of pure sucrose was taken27 as Wsat,p = 64.407 + 0.0725tb + 0.002057tb2 – 0.00000935tb3 (14) and the solubility coefficient (equation 11), applicable to cane liquors is simplified from Steindl etal28. The boiling point of water tbW is then calculated from the equation t bW W DS = t b − AX 100 − W DS BX 273 + t bW 100 CX Q 100 DX (15) where AX, BX, CX and DX are the new coefficients (equation 8 and Table II). To solve the nonlinear equation, the Maple Version 6 software29 was used to obtain a tbW for every set of Q , tb and y. As the solution was unstable with CX = -1.9735 or -1.97, an approximation CX=-2 had to be introduced, with an negligible effect on the results. From tbW,, BPE is obtained using equation (2). The procedure was then repeated for all combinations of Q, tb and y, and plotted (Figure 8) in the form comparable to the existing literature graphs8. Graphs of BPE v. supersaturation. The data generated in the previous section was replotted (points in Figures 10) at different temperature 55, 60, 65, 70 and 75oC, supersaturation from 1.0 to 1.3, and at five purity levels , 60, 70, 80, 90, and 100. The NLIN procedure of SAS26 was again used to develop a general correlation between BPE, temperature, supersaturation and purity, and, of the several models tested, the fourparameter equation ∆t b /y = aX tbQbX - cXQdX (16) with aX = 0.4702, bX = - 0.2586, cX = 0.0688 and dX = 0.7875 was found to fit the points reasonably well over the full range of purities and temperatures and is represented by lines in the graphs in Figure 10. The threshold or critical supersaturation limit (nucleation limit) was calculated30 as y crit = 1.129 − 0.284(1 − Q / 100) + [2.333 − 0.0709(t b − 60)](1 − Q / 100) 2 (17) Discussion of the results For pure sucrose solutions our data and the correlation, equation 8, labeled “ASI” in Figures 5 - 7 – lie some 0.5 to 1oC above the other lines, with Batterham and Norgate’s and Starzak’s nearly identical, and Bubnik’s some 0.3oC below them . At all purities, Hoekstra’s has a distinctly lower slope than any of the other correlations or the experimental data. At 85 purity (Figure 6), the ASI and Bubnik’s correlations nearly coincide, and Baterham’s lies some 0.5 to 0.7oC below. At the lowest purity (Figure 7), the equation of Bubnik far overpredicts BPE as does, to a smaller degree that of Hoekstra. This, of course, is reflected in the small and opposite effect of purity, (DX > 0 in Table II) on BPE found here in comparison with previous works. The apparently smaller BPE of sugarcane solutions than that of sugarbeet liquors at the same purity was already noted by Thieme. Later, Baterham and Norgate found even smaller effect than Thieme of purity for cane sugar liquors. Hardly any detail of the experimental procedures is given in the historical publications, so the reasons for the deviations are hard to ascertain. Differences in analytical techniques and possibly of non-sugar composition may also be responsible for at least some of the discrepancies. The fact that our data, at the lowest purity lie slightly below, and nearly coincide at the 85 purity with, the non-boiling data of Batterham is a strong indication of absence of any measurable superheat or hydrostatic head effects in our experiments. Although our data and our (ASI) correlation for pure sucrose exceed the other lines by some 0.5 to 1oC, they agree quite well with the curves compiled by Holven14 in his Figure 1. A few points in Figures 5 – 6 were extracted from Holven’s lines and document the agreement. The fit and scatter of the experimental data with and around the ASI equation is reasonable, with the exception of a few tests that may have been affected by poorer than usual control of absolute pressure. In order to make the data useful for control of sugar boiling, the solubility and supersaturation concepts were introduced as described in the preceding sections, and the graphs (Figure 8) constructed as in the often used nomograph in the Cane Sugar Handbook8. A comparison of the two (Figure 9) reveals a relatively good agreement in the middle of the temperature range at high purities, but an increasingly large overprediction of BPE at lower purities by the Cane Sugar Handbook8 data. The straight lines in Figure 10, calculated from the new (ASI) equation 16 fit reasonable well the points in Figure 10, which in turn were calculated from the equation 8, and represent a general correlation in the form that is suited for control of sugar boiling. The equation 16 was incorporated into the UMC 800 controller and the initial tests of automatic boiling at ASI indicate that it is appropriate to accurately indicate the seeding point and control the seed addition. A literature equation30 for the critical supersaturation (dotted lines in Figure 10) is given as a guide for sugar boiling to delineate the “danger” zone for false grain, and its validity is subject of ongoing tests. Summary A series of boiling point elevation measurements was made with sugarcane liquors at three purity levels and three absolute pressures, and an equation W DS ∆t b = AX 100 − W DS BX 273 + t bW 100 CX Q 100 DX (AX = 0.1660 BX = 1.1394 CX = 1.9735 DX = 0.1237) developed that fits well the experimental data. Through an introduction of the solubility and supersaturation concepts a general equation ∆t b /y = aX tbQbX - cXQdX (aX = 0.4702, bX = - 0.2586, cX = 0.0688 and dX = 0.7875) and a series of graphs were produced that are suitable for use in, or direct incorporation into the software of, automatic control of sugar boiling in the sugarcane industry. Acknowledgments A partial funding from the American Sugar Cane League (ASCL) and the Sugar Processing Research Institute, Inc (SPRI) are gratefully acknowledged, as are useful input from prof. P. W. Rein and technical assistance of Lenn Goudeau, Scott Barrow, Joe Bell and Ms. LiFeng. Symbols A, B, C ,D,a,b,c,d MW P Q Q qsat qS/W qNS/W R t ∆t b W X y coefficients in equations 3,5,6,7,8,15,16. molecular weight, g/mole pressure constant (17638 J/mole) purity (%) solubility coefficient sucrose/water mass ratio in the solution impurity/water mass ratio in the solution 8.3143 J/(mole K), the universal gas constant temperature boiling point elevation mass concentration (g/100 g solution) mole fraction supersaturation Index Abs B b absolute Bubnik etal boiling point BN Crit DS H I NS p SP S sat W X Batterham and Norgate critical (refractometric) dry solids Hoekstra impure (WNS > 0) non sucrose pure solution (WNS = 0) Starzak and Peacock sucrose saturated (y = 1) water this work (equations 8,16) References (1) Ehrenberg, J., and Kessler, K. (1997) Application of a high-frequency measurement technique to the determination of dry substance content in solutions and suspensions as well as for the control of evaporating crystallizers. Zuckerindustrie, 122, No. 2, 100-108. (2) Theisen, K.H., Senkowsi, K., and Diringer, T. (2001) Concentration measurement for continuous sugar crystallization processes. 31st General ASSBT Meeting, February 28-March 3, Vancouver, Canada. (3) Saska, M., and Rein, P. W. (2001) Supersaturation and crystal content control in vacuum pans. Proc. 60th Meeting of Sugar Industry Technologists, 251 – 261. (4) Love, D.J., Swan, M.L. and Cox, M.G.S (2001) Combining temperature measurements with the outputs of a radio frequency probe. Proc. ISSCT. (5) Honig P.(1953) Principles of sugar technology, Elsevier Publishing Company. (6) Van der Poel, P.W., Schiwek, H., and Schwartz, T. (1998) Sugar Technology, Verlag D r. A. Bartens, Berlin. (7) Spengler O., Bottger, S. and Werner, E. (1938) Boing point elevation of pure and impure sugar solutions at different pressures. Zeitsch.Wirtsch.Zuckerindustrie, 88,521-608. (Abstract in Intern. Sugar Journal, p. 26, January 1939) (8) Chou C.C. and Chen, J.C.P. (1993) editors, Cane Sugar Handbook, 12th edition, John Wiley and Sons. (9) Thieme J.G. (1927) A discussion of the theory and principles of the art of sugar boiling and their application in sugar factory practice. Facts about Sugar, December 3. (10) Hugot E. (1960) Handbook of cane sugar engineering, Elsevier Publishing Company. (11) Othmer, D.F. and Silvis, S.J. (1948). Correlating boiling point elevations. Sugar, July , 2829. (12) BatterhamR.J. and Norgate, T.E. (1975) Boiling point elevation and superheat impure cane sugar solutions”. Int. Sugar Jnl. 77 , 359—364. (13) Wright P.G. (2001) private communication. (14) Holven A.L.(1942) Supersaturation in sugar boiling operations. Ind. Eng. Chemistry, 34, No. 10, 1234-1240. (15) Holven A.L. (1936) Sucrose solutions. Influence of pressure on boiling point elevation. Ind. Eng. Chemistry, 28, No. 4, 452-455. (16) Macdonald J.C. and Rodgers, T. (1947) Some factors affecting evaporator heat calculations for beet sugar factories. Int. Sugar Jnl.., August, 205-208. (17) Bates F. (1942) Polarimetry, Saccharimetry and the Sugars, Circular of the National Bureau of Standards, C440. (18) Nicol W.M. (1968) Boiling point elevation of pure sucrose solutions. Int. Sugar Jnl, 199-202. (19) Vavrinecz, G. (1973). Siedepunkterhohung und Ubersatigung von Zuckerlosungen. Z. Zuckerindustrie, 23, Nr. 1, 10-17. (20) Hoekstra R.G. (1985) Program for simulating and evaluating a continuous pan. Proceedings of SASTA, 48 – 57 (21) Sheng L.Q. (1990) Calculation of the boiling point elevation of sugar solutions. Int.SugarJnl, Vol. 92, No. 1100, 168 – 169. (22) Bubnik, Z., Kadlec, P., Urban, D. and Bruhns, M.(1995) Sugar Technologists Manual” 8th edition, Verlag Dr. A. Bartens, 203—205. (23) Starzak M. and Peacock, S.D. (1997) Water activity coefficient in aqueous solutions of sucrose – A comprehensive data analysis. Zuckerindustrie, 122, 380 – 387. (24) Lyle, O. (1970) Technology for sugar refinery workers, p. 627, Chapman & Hall, London. (25) Love D.J. (2001) private communication. (26) SAS Institute Inc, Cary, North Carolina, USA (27) Charles D. F.(1960) Int. Sugar J., 62, 126 – 131. (28) Steindl R. J., Broadfoot, R. and Miller, K.F. (2001) Development and application of a model for low-grade cooling crystallizers. Proc. 24th Congress of ISSCT, 116 – 123. (29) Waterloo Maple, Inc., Waterloo, Ontario, Canada (30) Broadfoot R. and Wright, P.G. (1972) Nucleation studies. Proc. 39th Conf. QSSCT, 353 – 362. Legends to the Figures Figure 1: The 200 L ASI vacuum pan and a view of the calandria. Figure 2: Some of the trends recorded during the BPE measurement tests. Purity 85, Pabs = 5.4 in Hg (Test 2-27-02). Only the absolute pressure, and liquor temperatures were used for the BPE calculations. Figure 3: Some of the trends recorded in the course of the “water tests” Note the steam pressure set point change at 10:53 AM. Pabs 5.4 inHg (Test 3-01-02). Figure 4: Correlation of the boiling point of water measurements used in the BPE calculations. Figure 5: Measured BPE points (diamonds), their correlation (ASI) and literature correlations. A few points (round symbols) were read off Holven’s compilation of pure sucrose data (Figure 1 in ref 14). Purity 100. Figure 6: Measured BPE points (diamonds), their correlation (ASI) and three literature correlations. Purity 85. Figure 7: Measured BPE points (diamonds), their correlation (ASI) and three literature correlations. Purity 58. Figure 8: BPE (at various liquor purities) vs. liquor (massecuite) temperature, calculated from equation 8. Figure 9: Difference of BPE (at various liquor purities) calculated from equation 8 and BPE given in ref.8, p.238, respectively, vs. liquor (massecuite) temperature, supersaturation 1.00. At higher supersaturations, the differences become proportionately higher. Figure 10: BPE (at various liquor or massecuite temperatures) vs. supersaturation, calculated from equation 16. Dotted lines: spontaneous nucleation limit calculated from equation 17.