# Optimisation of an Agitated Thin Film Evaporator for Concentrating Orange Juice Using Aspen Plus.

код для вставкиСкачатьDev. Chem. Eng, Mineral Process., 11(3/4),pp. 309-322, 2003. Optimisation of an Agitated Thin Film Evaporator for Concentrating Orange Juice Using Aspenplus N. Chawankul, S. Chuaprasert', P.L. Douglas* and W. Luewisutthichat' Department of Chemical Engineering, University of Waterloo. Waterloo, Ontario N2L 3GI CANADA 'Department of Chemical Engineering, King Mongkut's University of Technologv, Bangkk, I0140 THAILAND This paper presents an application for the optimisation of an existing agitated thin film evaporator (A TFE) for concentrating orange juice using Aspenplus? A rigorous heat exchanger mode (Heaa) and a rigorous two-phase flash model (Flash2) were used to simulate the dominant effects of the ATFE. The thermo-physical properties of orange juice, not available in Aspen Plus, were determined experimentally and correlated as functions of temperature and solids content by Boonsriudomsuk [I]. Effective heat transfer coeflcients were calculated fLom measured temperatures andflow rates. Experiments were performed on a laboratory-scale and pilot plant system and compared with the simulation results. The Aspen Plus simulation model using experimentally determined heat transfer coeficients and thermo-physical properties of orange juice compared well with the experimental data fiom the ATFE. When the mass and energy balance data were reconciled the errors between both experimental and simulation resuIts were significantly decreased. The optimisation results indicated that by operating at the optimum operating conditions the operating costs could be reduced by about 10%. This translates into savings of more than $10,00O/year in the case of the laboratory-scale evaporator and $33,00O/year in the case of the pilot plant. r f a commercial ATFE process was optimised then the potential savings could approximate to $330,00O/year. Clearly, process optimisation is a valuable tool in the design and operation of these processes. "Authorfor correspondence (pdouglas@cape.uwaterloo.ca). 309 N.Chawankul, S. Chuaprasert, P.L. Douglas and W. Luewisutthichat Introduction Simulation and optimisation are used as tools to analyse plant design and operating conditions with a goal to improving the plant operation. Simulation packages such as AspenPlusW have built-in process models and optimisation routines, thus offering a convenient and time-saving means of examining an entire chemical process [2]. Many process optimisation applications can be found in the chemical process industries, however, there are relatively few applications in the food industry. Evaporators are widely used in the food industry as concentrators and separators. Agitated thin film evaporators (ATFE) having short residence times and relatively high heat transfer coefficients are best used for concentrating foods that are heat sensitive and cannot tolerate high temperatures for more than a few seconds. Chuaprasert et al. [3] used AspenPlus to simulate and perform data reconciliation of experimental measurements for a laboratory-scale ATFE used to concentrate sugar syrup. Chawankul et al. [4] extended this model to the case of tangerine orange juice. This paper presents a practical application for optimisation of the operation of a tangerine orange juice concentrator using an ATFE. Experiments were performed on two different ATFE systems, one was a laboratory-scale system used by Chuaprasert et al. [3] and the other was of pilot plant-scale capable of handling larger flow rates. The process was simulated using the AspenPlus simulation model developed by Chuaprasert et al. [3]. The model consisted of a rigorous heat exchanger model (Heatx) followed by a rigorous two-phase flash (Flash2). The heat exchanger model was used to simulate the evaporator; it required information on the configuration (cocurrent or countercurrent), the heat transfer area (A), and the overall heat transfer coefficient (U). The output stream from the heat exchanger (a two-phase stream) consisting of concentrated orange juice and water vapour, was fed to the two-phase flash unit operating at the same pressure. The flash unit then separated the water vapour from the concentrated orange juice liquid. The built-in SQP (Sequential Quadratic Programming) optimisation routine was used for both the data reconciliation and the economic optimisation problem. For the economic optimisation problem, the feed flow rate, steam flow rate and evaporator vacuum pressure were simultaneously adjusted to minimise either the energy consumption or the energy cost, while maintaining product purity and production rate constraints. Agitated Thin Film Evaporator (ATFE) A schematic diagram of the ATFE system in the Department of Chemical Engineering at King Mongkut's University of Technology Thonburi (KMUTT) is shown in Figure 1. Both the laboratory-scale system and the pilot plant system have the same configuration, the only difference being that the pilot plant system can handle larger flow rates. The specificationsof the evaporator unit are the same for the laboratory and pilot scale units as follows: 9 material stainless steel (3.4 mm thick) 9 height 4.5 m diameter 0.108 m 9 heat transfer area 0.251 mz 9 agitator blades 4 Luwa fixed clearance, (smooth or meshed) > 310 Optimisation of an Evaporatorfor Concentrating Orange Juice Using Aspenplus Figure I . Schematic diagram of agitated thinfirm evaporator system. e l STEAM PUMP FEE0 1 A 1 fl VAPOUR FLASH b PRODUCT Figure 2. Aspenplus simulationflowsheet. 311 N. Chawankul, S. Chuaprasert,P.L. Douglas and W.Luewisutthichat Dilute orange juice feed solution was pumped from the feed tank; the flow rate was controlled by adjusting valve V1. The solution flowed through a flow meter and entered the ATFE and was distributed over the inner circumference of the heated ATFE cylinder. The rotating agitator blades created a thin liquid film, which flowed down the heat transfer walls under the force of gravity. The orange juice liquid was concentrated through evaporation by the steam in the outer shell of the evaporator. The concentrated liquid then flowed into tank RI and left the system via valve V5. The water vapour formed during the evaporation process left the top of the ATFE unit and was condensed in the condenser, and flowed into tank R2 and left the system via valve V7. The ATFE systems were operated under vacuum to reduce the boiling point of the orange juice. A vacuum pump was used to create the vacuum, valve V6 was used to control the vacuum pressure in the system. AspenPlus Simulation of the ATFE Process An AspenPlus model of ATFE process was developed [3, 41 using the AspenPlus rigorous heat exchanger model, Heatx, followed by a rigorous two-phase flash model, Flash2, as shown in Figure 2. The feed to the Heatx block represents the dilute orange juice-water feed to the process. Steam enters the Heatx and leaves as condensqte. The product stream from the Heatx block is a two-phase mixture of water vapour and concentrated orange juice-water liquid; this stream does not exist in the real process. The FIash2 model is used to separate the concentrated orange juice from the water vapour. The products from the Flash2 represent the two products from the real process. The pressure in the Flash2 unit was assumed to be the same as in the Heatx unit and adiabatic operation was assumed. Adiabatic operation is a reasonable assumption since the heat exchanger is insulated. The H e m model requires the heat transfer coefficient and heat transfer area as inputs. The area was measured directly and found to be 0.25 1 mz. The heat transfer coefficient used in the Heatx model was calculated directly from process measurements using Equation (l), where T, was used as the reference temperature. The effective heat transfer coefficient, calculated from Equation (l), combines all heat transfer factors including heat loss from the process and fouling factors. Therefore, good agreement would be expected between the simulation results using the effective heat transfer coefficient and the experiments. Alternatively, the U-value could be calculated from correlations such as those developed by Sae Tae [5] and Chawankul et al. [4]. However, because this work focuses on the optimisation of existing evaporators, presumably measurements of temperature and flow rates would be available enabling the calculation of the effective heat transfer coefficient (U) using Equation (1). 312 Optimisation of an Evaporator for Concentrating Orange Juice Using AspenPIus Optimisation Results and Discussion Once data reconciliation has been completed we are in a position to use the reconciled data as the base case and attempt to improve the economic performance of the evaporator. The energy consumption and energy cost were minimised using the builtin SQP optimiser in AspenPlus. Both the energy consumption and energy cost were optimised to determine if the location of the optimum changes when the objective function was modified. A Pump was added to the AspenPlus evaporation process flow sheet to account for the electricity used by the vacuum pump. The modified flow sheet is shown in Figure 2. The dilute orange juice is fed to the Pump at atmospheric pressure and leaves the Pump under a vacuum. The Pump module calculated the electricity used to produce the vacuum. Minimising Energy Consumption Usually it would be necessary to maximise profit or minimise operating costs, however, the development of these objective functions can be complex requiring information that is not readily available. Therefore, it is quite common to minimise other functions such as the energy consumption, with the assumption that the minimum energy consumption corresponds to the same operating conditions as a minimum cost or maximum profit. It is important to note that, although this may seem logical, this is not always the case. To minimise the energy consumption used in the ATFE process the objective function was defined as the sum of the energy associated with the steam and the electricity used in the vacuum pump; the objective function was written as Equation (2): Energy = (Steamflow rate) (Steam enthalpy) + Electricity used in vacuum pump (2) The optimum was limited by the feasible search space defined by the equality constraints (i.e. the model equations) and the inequality constraints on the product flow rate and the product concentration. The optimisation problem can then be written as: Choose: feed flow rate of dilute orange juice steam flow rate pressure inside the evaporator To minimise: Equation (2) h(x) = 0 Such that: g(x) 2 0 where h(x) = 0 represents the process model equations, and g(x) 2 0 represents the production and purity constraints. The product concentration must be 2 28 % (mass basis), and the product flow rate must be 2 5 k g h , in the case of the laboratory scale and 2 15 kg/hr in the case of the pilot plant. 313 N. Chawankul, S. Chuaprasert,P.L. Douglas and W. Luewisutthichat The objective function, Equation (2), was incorporated into the Aspen Plus optimiser. The optimiser adjusted the decision variables simultaneously, to minimise the objective function. Aspenplus requires that each decision variable have an upper and lower limit to restrict the search space for each variable. In this work, the laboratory scale and pilot plant systems were given different ranges due to their different sizes as follows: Laboratory-scale: feed flow rate was varied between 15-35 k g h steam flow rate was varied between 10-30 kghr pressure in the evaporator was varied between 0.1-0.9 bar Pilot plant: feed flow rate was varied between 55-90 k g h steam flow rate was varied between 40-70 kg/hr pressure in the evaporator was varied between 0.1-0.9 bar The location of the optimum was calculated using the built-in SQP optimisation routine. Laboratory-Scale Evaporator The optimisation results, shown in Table 1, indicate that the minimum energy used in the process is obtained when the feed flow rate is 17.484 kghr, the steam flow rate is 13.973 kg/hr and the pressure in the evaporator is 0.56 bar. The resulting product flow rate is 4.979 kg/hr, slightly violating the product flow rate constraint, (25 kglhr), only by the tolerance which is allowed in the constraint definition. The orange juice product was 28.1% which is slightly above the constraint of 28%. Therefore, the optimum operating point lies on the production constraint (5 kg/hr) and on the concentration constraint (28%). This would be expected when trying to minimise energy consumption. The approach would be to try and reduce both the amount and purity of the product, while observing the production and purity constraints. The optimiser determined the optimum point by keeping the feed flow rate low, (an increased feed flow rate requires an increased steam flow rate). The optimiser lowered the vacuum pressure to reduce the steam requirement. The electricity required by the Pump is low when it is compared with the steam energy (only 0.001% of total energy), so the optimiser prefers to create a low pressure in the evaporator to reduce the steam requirement. From Equation (2), when the steam flow rate was increased, the energy fiom steam was also increased. The optimiser tried to search for the optimum steam flow rate that matched the feed flow rate and steam flow rate. In this case the steam flow rate at the optimum is 13.97 kgihr. Pilot Plant Evaporator In the pilot plant evaporator, the feed flow rate and steam flow rate are higher than in the laboratory scale. Therefore, the ranges of decision variables such as the feed flow rate and steam flow rate were extended as discussed above. The optimisation results are presented in Table 2. The optimum feed flow rate is 60.272 kg/hr, the evaporator 314 Optimisation of an Evaporatorfor Concentrating Orange Juice Using Aspenplus pressure is 0.2 bar and steam flow rate is 47.623 kghr. The resulting product flow rate is equal to 16.935 kg/hr and the product concentration is 0.285, which satisfies both constraints. Because the initial feed flow rate requires a high steam flow rate, the optimiser again tries to lower the feed flow rate. The energy source is mainly the steam supplied to the evaporator whereas the electricity used in the vacuum pump is very low (only 0.0008%of the total energy requirement). Feed Product Vapour Steam Temperature ("C) 30 84.8 84.8 120.2 Pressure (bar) 1.01 0.56 0.56 2 17.484 4.979 12.505 13.973 0.08 0.281 Mass Flow (kghr) Orange juice concentration (Brix) Energy consumption (kW) 10.67 Feed Product Vapour Steam Temperature ("C) 30 60 60 127 Pressure (bar) 1.01 0.2 0.2 2.46 60.272 16.935 43.336 47.623 0.08 0.285 Mass Flow (kg/hr) Orange juice concentration (Brix) Energy consumption (kW) 36.38 Minimising Energy Cost In the previous section the minimisation of energy consumption was discussed. In this section the energy cost will be minimised. The unit cost of steam and electricity were estimated as follows: - Cost of steam = $0.0173 k g - Cost of electricity = $0.0615 / kW-hr 315 N. Chawankul, S. Chuaprasert, P.L. Douglas and W. Luewisutthichat The objective function was modified by multiplying both energy terms in Equation (2) with their respective costs. The modified objective fimction is shown by Equation (3): Energy cost = (Steamflow rate x Enthalpy of steam x 0.0173) + (Electricity used in vacuum pump x 0.0615) (3) The new optirnisation problem becomes: Choose: To minimise: Such that: feed flow rate of dilute orange juice steam flow rate pressure inside the evaporator Equation (3) h(x) = 0 g(x) 2 0 The constraints remain unchanged. The results of the optimisation are shown in Tables 3 and 4. From Table 3, the location of the laboratory scale optimum is almost the same as in the energy minimisation case. This is because the steam cost is dominant and the cost of electricity used in the Pump is minimal. In the pilot plant ATFE, the location of the optimum is also nearly the same as was found fiom the energy minimisation study and is shown in Table 4. The results of this study are not surprising since the cost and amount of steam needed dwarf the electricity term in the objective function, Equation (3). Although, in this case the location of the optimum did not change when the objective function was modified, this phenomenon is by no means common. In many applications a change in the parameters in the objective function (e.g. a change in the cost of utilities) does change the location of the optimum. This in turn requires either that the parameters need to be known more accurately, or that the optimisation be re-run whenever the parameter changes. This then leads to the notion of Real Time Optimisation or RTO. Table 3. Typical cost minimisation results: lab scale ATFE. Feed --m-Product Vapour Steam 82.1 82.1 120.2 Temperature (“C) 30 Pressure (bar) 1.01 0.5 1 0.5 1 2 Mass Flow (k@) 17.491 4.96 12.531 13.876 0.282 - Orange juice concentration (Brix) I 316 I I Optimisation of an Evaporator for Concentrating Orange Juice Using Aspenplus Overall Cost Savings The objective of the optimisation was to determine the operating conditions that result in the minimum operating costs. The operating cost of the process depends mainly on the cost of steam and the cost of the feed; the cost of electricity is minimal. The feed cost was estimated to be $0.64/kg of feed and the steam cost was assumed to be %O.O173/kgof steam. The total saving was determined over a period of 8000 stream hours per year by: - Savings = [{(Feedflow rate at nominal conditions FeedJlow rate at optimum conditions)x 0.64 + {(Steamflow rate at nominal conditions - Steamflow rate at optimum conditions)x 0.0173}] x 8000 (4) Tables 5 and 6 show the total savings at the optimum operating condition on the laboratory scale and pilot plant systems. It can be seen that the savings on the feed cost dominates the total savings since the feed costs much more per unit than the steam. The optimisation can save as much as $10,7lS/year on the Iaboratory scale evaporator and $32,864/year on the pilot plant system. The economic incentives for optimisation on a commercial evaporator would increase due to the large flow rates involved. Commercial evaporators operate at feed flow rates of about 10 times that used in the pilot plant; this could then translate into a potential total savings of about $330,00O/year. Savings of this magnitude would certainly justify an optimisation study of the process. Effect of Starting Point on Final Solution It is important to ensure that the optima, found by the optimiser, is the global optimal and not a local optima. The starting point for the optimisation, i.e. the initial guess, was varied and each solution was determined. The results of these experiments are reported in Tables 7 and 8. 31 7 N.Chawankul, S. Chuaprasert, P.L.Douglas and W. Luewisutthichat Table 5. Total cost savings afier optimisation: lab scale AFTE. Operating Conditions Details Feed flow rate, ( k g h ) Steam flow rate, (kghr) iEnergy cost, ($ Iyr.) Optimum Nominal I I 19.45 I 17.48 18.38 13.97 2,544.62 1,934.3 1 I Energy cost savings, (S /yr.) 610.31 Table 6. Total cost savings afier optimisation: pilot plant ATFE. Operating Conditions Details Nominal Optimum Feed flow rate ( k g h ) 63.15 56.89 Steam flow rate (kghr) 5 1.62 46.17 Energy cost ($ /yr.) 7,147.38 6,392.77 I Energy cost savings, ($ /yr.) 754.62 Feed cost savings, ($ /yr.) 32,109 318 Table 7. Efsect of the starting point on the final solution: laboratory scale ATFE. 3 9% 9 P Table 8. Eflect of the starting point on thefinal solution:pilot plant A TFE. 9 Optimisation of an Evaporatorfor Concentrating Orange Juice Using AspenPIus Table 7 shows the results for the laboratory-scale ATFE unit. The starting points (Ff, F, and P) were adjusted and the optimal values (Ff*, F,* and P* and OF) were calculated. Table 8 shows the results for the pilot plant ATFE unit; while the results for the pilot plant are also good, there is a slightly larger variation in the optimum values. This is due to the fact that the equipment is larger and the flow is larger. The vacuum pressure (P) appears to have a relatively insignificant effect on the objective function because its large variation does not seem to affect the objective function very much. The difference in results for each of the various runs is due to the effect of the tolerance in the optimisation routine; there is a tolerance set on the objective function value and the decision variables. That is, the optimiser will stop when the difference between the current value of either the objective function or the decision variables, and the previous iteration, is less than the tolerance. As a result of these tests we are confident that the SQP optimiser was able to find the true or global optimum for this problem. Conclusions The steady-state AspenPlus simulation model combined with a built-in optimisation routine was used to simulate and reconcile experimental data gathered from experimental evaporator systems. When the process was operated at the optimum condition, the operating cost was reduced by as much as 10%. Savings of more than $10,000 per year in the case of the laboratory scale evaporator and $33,000 per year in the case of the pilot plant could be realised. Estimated saving of $330,000 per year could be realised for commercial operations. The Aspen Plus evaporation model represents a good model for orange juice evaporators, and it can be applied to other evaporation processes if the physical properties of the feed are available. Acknowledgements The authors are very appreciative of research finding fiom The National Science and Technology Development Agency of Thailand (NSTDA), which enabled this research to be undertaken. Nomenclature A C, Ff F, OF P Tf Overall heat transfer area (m2) Specific heat capacity (kJfl<g"C) Flow rate of feed stream (kg/hr) Flow rate of steam (kg/hr) Objective function Pressure in evaporator(bar) Feed temperature ("C) Temperature of orange juice product ("C) Overall heat transfer coefficient U (kw/m2 "C) Vapour flow rate (kgkr) V ATh Log mean temperature difference ("C) Latent heat of evaporation li (kJW Tp 32 1 N. Chawankul, S. Chuaprasert, P.L. Douglas and W.Luewisutthichat References 1. 2. 3. 4. Boonsriudomsuk, S. 1999. Thermophysical properties of orange juice. BEng Thesis, Department of Chemical Engineering, King Mongkut's University of Technology Thonburi, Bangkok, Thailand, (in Thai). Aspen Technology. 1993. AspenPlusM Users Manual. AspenTech Ltd., Cambridge, Mass.,USA. Chuaprasert, S.,Douglas, P.L., and Nguyen, M. 1999. Data reconciliation of an agitated thin film evaporator using AspenPlus. J. Food Eng.. 39,261-267. Chawankul, N., Chuaprasert, S., Douglas, P.L., and Luewisuttichat, W. 2001. Simulation of an agitated thin film evaporator for concentrating orange juice using AspenPlus. J. Food Eng., 47(4), 247-253. 5. 6. 7. Sae Tae, A. 1999. Heat transfer coefficients in an agitated thin film evaporator for concentrating sugar syrup. Masten Thesis, Department of Chemical Engineering, King Mongkut's University of Technology Thonburi, Bangkok, Thailand, (in Thai). Tjoa, I.B., and Biegler, L.T. 1991. Simultaneous strategies for data reconciliation and gross error detection of non-linear systems. Comput. Chem. Eng., 15(10), 679-690. Piccolo, M., and Douglas, P.L. 1996. Data reconciliation using AspenPlus. Dev. Chem. Eng. Mineral Process., 4(3/4), 157-182. Received: 13 December 2000; Accepted afrer revision: 6 August 2002. 322

1/--страниц