Energy-efficient and cost-effective separation model for solvent recovery from colloidal lignin particles dispersion

Colloidal lignin particles (CLPs) are potentially one of the sustainable alternatives for petroleum-based feedstock. CLPs address the heterogeneity of lignin by enhancing its homogeneous dispersion in aqueous phases. The main production steps are dissolving lignin in tetrahydrofuran, diluting the solution with ethanol, forming CLPs through self-assembly after encountering water, recovering solvents, and finally drying CLPs. In this process, solvent recovery plays an important role in mitigating environmental problems. However, the formation of azeotropes makes the separation process energy-intensive and costly. In this work, two separation methods, evaporation and distillation, are modelled in Aspen Plus ® and compared based on their total annual costs (TACs). Sizing and cost estimations are conducted based on vendor quotations and using design and economic analyzer tools. Results show that distillation reduces costs by up to 37% compared to evaporation. Accord-ingly, as the main separation unit, distillation parameters are optimized based on the minimum TAC. For further assessment of the increase in the rate of costs by reaching nearly pure products, extractive distillation is simulated and examined. Results show that using an entrainer to enhance the tetrahydrofuran concentration from 88 to 99.5 wt.% substantially increases the TAC by over 50%. Finally, based on the results, the desired solvent recovery model is finalized by employing the rate-based approach. Currently, studies with a focus on the techno-economic assessment of pilot-scale separation units are limited, and the developed model offers a good basis for designing optimal solvent recovery units, related to processes where lignin is dissolved, prior to commercialization


| INTRODUCTION
[3] Lignin is rich in different functional groups, which makes it a potential source for producing value-added products.Ultraviolet (UV) protection, moisture resistance, and the capability of forming solid networks are among the interesting properties of this biopolymer.Researchers have examined the application of this natural material in developing eco-friendly products such as adhesives and coatings [4][5][6][7] or applying it as reinforcement in composites. [8]urrently, less than 2% of the available lignin is employed in chemical applications and the development of improved products. [9,10]The heterogeneous nature and poor dispersibility of lignin in different media, including water, have limited its application on commercial scales.Accordingly, in recent years, significant research has been conducted to develop lignin nanoparticles and nanostructures with enhanced properties. [11]Among the different production techniques, extensive studies have been carried out on precipitation methods through pH dropping or solvent shifting.However, they have limited industrial applicability due to their high solvent consumption rate.[14] Lievonen et al. [13] and Figueirêdo et al. [15] prepared stable spherical colloidal lignin nanoparticles (with the size of 200-500 nm) by dissolving softwood kraft lignin in tetrahydrofuran (THF) and precipitating nanolignin by introducing the solution into excess water.The lignin nanoparticles show promising results.Nonetheless, long precipitation times and the highly diluted nanolignin dispersion are not economically desirable on the commercial scale.Leskinen et al. [16] also presented a scalable technology for the production of lignin particles by means of solvent shifting, using a mixture of THF and water as solvents.They used a pilot-scale stirred vessel to test the upscalability of the process.Stirred vessels would commonly offer good mixing conditions depending on the agitator type, rotational speed, fluid viscosity, and so on.However, in industrial processes, larger stirred tanks show high power consumption, inefficient heat and mass transfer, and difficulty in controlling the size range of lignin particles, thereby making scale-up and process implementation problematic. [16,17]Some of the recent colloidal lignin particles (CLPs) production methods also require using hazardous materials, which can threaten the environment. [14,18]ccordingly, for the production of CLPs on the commercial scale, developing a viable and efficient process which excludes hazardous materials and minimizes the use of solvents is desirable.Lintinen et al. [19] presented an industry scalable closed-cycle process to produce water-dispersible CLPs (with the size of 300-400 nm) via a three solvent polarity exchange.Recovering and reusing the solvents in this process reduces the waste stream and the operating costs.Bangalore Ashok et al. [20] carried out the techno-economic assessment of this production method and concluded that the developed CLPs have the potential to compete with conventional petroleum-based polymers, such as those that are applied as adhesives and coatings or utilized in composites.Figure 1 shows the simplified process diagram of the CLPs production technique proposed by Lintinen et al. [17,19] In the beginning, the lignin solution is prepared by dissolving LignoBoost™ softwood kraft lignin in THF (step 1).Reasonable, bio-based solvents, especially THF, are potential alternatives to fossil-derived solvents such as acetone.Afterwards, the lignin solution is diluted by adding ethanol and water to the system (step 2).Ethanol affects the formation of colloidal particles in the next step.The diluted lignin solution encounters water at 20 C using a tubular reactor (TBR) which offers homogeneous mixing, narrower particle size distribution, better mass and heat transfer, and so forth. [17]In this step, water draws out colloidal particles by changing the solubility of lignin and the ratio of the solvents, and consequently, spherical CLPs are obtained (step 3).Next, for solvent recovery, the organic solvents are separated from the CLPs dispersion and after recovering in the second separation step, THF and ethanol are recycled to dissolution and dilution steps, respectively (step 4).In the following, the CLPs dispersion is concentrated in an ultrafiltration unit, which is based on enhanced dynamic crossflow membrane filtration.The installed rotor in the ultrafiltration unit prevents CLPs from passing through membranes or remaining on them.The permeate stream of the ultrafilter, which is rich in water, is recycled back to the CLPs formation step (step 5).Finally, the main product in the form of dry powder is obtained by spray drying the concentrated CLPs dispersion using 180 C hot air (step 6).
For the process technology development, the next stage after laboratory-scale research is the design and construction of the pilot plant.Pilot plants offer the first opportunity for continuous processing and testing of different operating scenarios to reduce the risk and uncertainty before commercialization.Based on the results, optimal operating conditions are established for full-scale operation.According to the production process of CLPs offered by Lintinen et al., [19] the recovery of solvents is the most energy-intensive step due to the formation of three binary azeotropes. [20]The solvent recovery unit also plays a significant role from the environmental aspect.Recovering used solvents and recycling them back to the previous steps offers a closed-loop production process that alleviates waste and sustainability concerns and reduces the consumption of fresh organic solvents.Hence, optimizing the solvent recovery unit is important during the pilot plant's design and operation.This work presents a comprehensive study for modelling and comparing two common separation techniques, evaporation and distillation, on a pilot scale and later optimizing the design parameters of the desired separation method.[23] However, designing and modelling an optimized separation unit on the pilot scale with the ability to preserve CLPs are new aspects that have been considered in this work to investigate the viability of the production technique and the possibility of its application on larger scales.Moreover, limited guidelines are available for preliminary sizing and cost estimation of the optimized pilot unit.Therefore, in this research, equipment sizing and pricing are based on using equipment design and rating tools, economic analyzer software, and vendor data to enhance the accuracy of the proposed model.
For the pilot plant modelling prior to construction, the simulation modelling requires the collection and analysis of the input data collected from the literature review as part of the simulation phase tasks. [24]In this study, the input data include the operating and structural parameters for equipment and choosing suitable thermodynamic and optimization techniques for running the calculations which play an important role in obtaining valid results.By choosing suitable input parameters, more accurate data on the separation efficiency and recovery of the solvents, equipment sizing, utility consumption, and total annual costs (TACs) are collected.Furthermore, by modelling the process units and gathering the required information about separation results and equipment design, the uncertainty level during the pilot plant development decreases and repetitive experimental tests and analysis are minimized.Therefore, choosing appropriate input parameters is an important step during the simulation phase to offer a good basis for process development and optimize the investment costs in the real system.
During the modelling of the first solvent recovery unit, the presence of CLPs requires special considerations to mitigate fouling formation.The feed stream of the second solvent recovery step excludes CLPs, and only contains solvents; therefore, common evaporation and distillation equipment and internals that are often used in process industries are suitable.The mentioned equipment is suitable for processes that do not include solid particles or special chemicals, including corrosive services.
For the evaporation method, the thin film evaporation (TFE) system is a good option in the first separation step since it can process materials that are prone to fouling and are heat sensitive.Short residence time, significant turbulence, and a high surface renewal rate are important factors that support the TFE operation. [25,26]In the case of the distillation technique, using packings and internals that are less susceptible to fouling is important in the first solvent recovery step.
Technological schemes are modelled in Aspen Plus ® V11 to evaluate the recovery of organic solvent outcomes based on the TAC.The TAC includes the annual utility cost and partially the total installed cost of the separation unit.Based on the TAC results, the desired separation method is selected, and related design parameters are optimized considering the minimum energy consumption and defining the product compositions as constraints.
Finally, for further investigation of the increase in the TAC in the case of obtaining almost pure THF, an advanced version of the selected separation technique is modelled.Techniques such as extractive separation can be coupled with the selected method to improve the azeotropic mixture separation.In this work, an entrainer with a high boiling point is introduced into the second solvent recovery unit.Pressure swing distillation is also another method to obtain highly purified THF.However, this technique requires operating at F I G U R E 1 Process flow diagram of the production of dried colloidal lignin particles (CLPs) based on the method proposed by Lintinen et al. [19] THF, tetrahydrofuran high operating pressures, up to 5 bar, which can negatively influence the quality of CLPs. [22]The modified separation unit is optimized based on the minimum TAC, and the results are compared with the conventional version of the selected model.Figure 2 shows the main steps followed in this work for finalizing the optimal solvent recovery unit.

| Thermodynamic method
The solvent recovery modelling is performed in Aspen Plus ® V11.Among different activity coefficient property methods used for modelling highly nonideal systems at low pressure, Universal Quasichemical (UNIQUAC) is reliable to determine the vapour-liquid equilibrium and liquidliquid equilibrium calculations accurately. [27]In addition, according to a study conducted by Wang et al., [22] the estimated properties of water/THF/ethanol azeotropic mixture using the UNIQUAC method are in close agreement with the experimental results.Therefore, this thermodynamic package is employed in this work for process modelling.Table 1 and Table 2 show UNIQUAC model binary interactions and azeotrope data for the water/THF/ethanol system, respectively.The temperature-dependent binary parameters, applied to estimate the activity coefficients, are extracted from the Aspen Physical Property system.

| Evaporator
For separating organic solvents from the CLPs dispersion in the first solvent recovery step, a TFE is employed.To model the evaporator, the fundamental theory, introduced by Billet, [28] is applied.[31] Multistage modelling enables the simulation of the TFE based on the proposed theory in Aspen Plus ® V11. [31,32]o utilize this model, multiple consecutive flash drums are employed, as shown in Figure 3, with a temperature rise of 5 C between each flash drum. [31]The temperature of the first flash drum is the bubble point of the feed stream at 1 bar, 69.7 C .For the second solvent recovery step, a conventional separator is chosen to separate ethanol from THF.

| Distillation column
For modelling the distillation process in the Aspen Plus ® V11 environment, the rigorous RadFrac column is utilized in equilibrium and rate-based modes.The equilibrium mode of the RadFrac column defines the optimized parameters, and the rate-based distillation model applies them to accurately define the compositions and F I G U R E 2 Main steps followed in this work for finalizing the optimal solvent recovery unit.TAC, total annual costs T A B L E 1 UNIQUAC model binary interactions for a water/tetrahydrofuran (THF)/ethanol mixture extracted from the Aspen Physical Property system properties of column products.The rate-based distillation model determines bulk properties and assesses mass and energy fluxes based on the vapour-plug (VPlug) flow model.The bulk properties of the liquid and vapour phases depend on the outlet stream and average conditions at each stage, respectively. [33]For correlation calculations, the HanleyStruc method is employed to estimate the mass transfer and interfacial area.According to Hanley and Chen [34] the results of this correlation are close to experimental data, and it performs better considering other correlations found in the literature for structured metal packings.To calculate the heat transfer coefficient, the Chilton and Colburn technique, which computes heat transfer coefficients from the binary mass transfer coefficients, is applied.
Regarding the global parameters, the flow model condition and transition factor equal 0.5 and 0.2, respectively.Tuning factors, heat transfer value, and flow model condition and transition factors are also assumed to be 1.

| Solvent selection for extractive distillation
As mentioned in Section 1, for the separation of THF from ethanol in the second distillation step, the application of extractive distillation is studied to assess the increase in the related costs by obtaining nearly pure THF.This method is mostly used for azeotropic compositions and close-boiling point mixtures.By adding an entrainer with a high boiling point into the system, the relative volatility of the components in the azeotropic mixture is improved, hence a highly efficient separation is yielded.Organic solvents including dimethyl sulphoxide (DMSO), ethylene glycol (EG), and glycerol are commonly applied as entrainers in extractive distillation.Tang et al. [35] tested the application of different separation agents such as EG and DMSO for enhancing the relative volatility of THF and ethanol.He observed that using EG as an entrainer significantly improves the separation efficiency.Wang et al. [21] and Su et al. [23] also reached $99.9 mol% of product purity by introducing pure or a mixture of EG as the separating agent into the distillation column for THF and ethanol separation.EG with the formula (CH₂OH)₂ is low-toxic, and at normal pressure, the boiling point is 197 C, thereby having a low vapour pressure.From the standpoint of using renewable solvents, bio-based EG, derived from sustainable biomass, is a promising alternative to fossil-based ones.Accordingly, in this work, EG is added as the entrainer into the second solvent recovery unit at a location above the feed stage. [36,37]

| Process optimization
In this study, the separation process is optimized based on minimizing the TAC considering the purity of the products as the main criteria that should be fulfilled.The optimization is conducted by simulating the process with Aspen Plus ® V11 combined with the design of experiments (DOE).Simulating the process coupled with DOE on a pilot scale reduces the number of trials, and the results can prevent revenue loss after scaling up the process. [38,39]In this work, the impact of both structural and operating factors on the response, TAC, is assessed based on the full factorial design to minimize the outcome.At this stage, the optimization tool of Aspen Plus ® V11 increases the calculation speed and offers the opportunity to calculate the responses for all combinations of the factors.

| Economic optimization procedure
The economic analysis and optimization are conducted based on the TAC.TAC includes the annual utility costs F I G U R E 3 Multistage evaporation model applied for modelling thin film evaporation (TFE) in Aspen Plus ® V11 proposed by Ilmanen [31] and a part of the total installed equipment costs.The main utilities that are employed in the solvent recovery process are low-pressure steam, cooling water, and electricity.The utility prices are summarized in Table 3.In this study, the cost of the fresh entrainer, used as the make-up stream, is negligible due to the efficient recovery of EG.The installed equipment expenses include the main equipment costs, including evaporators, columns, heat exchangers, and pumps.The equipment is sized, and the cost estimations are performed using Aspen Process Economic Analyzer™ V11 (APEA), Aspen Exchanger Design and Rating™ V11 (Aspen EDR), and vendor quotations, as shown in Table 3.In this study, piping, instrumentation, and electricity are not considered at this stage, and the installation cost of the equipment excludes foundation, supports, and platforms.In the case of pilot-scale plate heat exchangers (PHEs), APEA™'s capability is limited for calculating the equipment cost since calculations are available for heat exchangers with areas greater than 1 m 2 .Nevertheless, based on the process and physical property of inlet streams, Aspen EDR™ V11 calculates not only the required heat transfer area but also the price of the pilotscale PHEs.For estimating the cost of PHEs, such as reboilers and condensers, versus different areas in varying design scenarios, results of Aspen EDR™ V11 are interpolated using the power regression (Figure 4).The estimated coefficient of determination, R 2 , is 0.97, which shows that regression results are in good agreement with the Aspen EDR™ V11 data.Hence, this regression model can be used in the preliminary process design step.Based on Aspen EDR™ data c 10% Alfa Laval [40] Condenser Aspen EDR™ 780 Â A C 0.16
The optimization approaches for the conventional recovery method and the advanced separation technique are shown in Appendix S1.During the optimization, separation parameters are defined by minimizing the energy consumption and considering the design specifications in Aspen Plus ® V11.For this purpose, the bound optimization by quadratic approximation (BOBYQA) algorithm and the Broyden method are employed to define the optimal feed stage and fulfil design specifications.

| Process description
The flow rate of the primary feed stream, CLPs dispersion, is 50 kg/h, and the liquid phase (composition 75.2 wt.% water, 11.7 wt.% ethanol, and 13.1 wt.% THF) constitutes 97.48 wt.% of the feed stream.During the first solvent recovery step, ethanol and THF are vaporized and separated from the CLPs dispersion.Afterwards, the vapour product is condensed and sent to the second

| Process selection
Figure 5A,B illustrates the flowsheets of conventional evaporation and distillation units including the flow rates and operating conditions which are based on the Aspen Plus ® V11 simulation.The details of the streams for each flowsheet are presented in Appendix S1.According to the results, organic solvents are recovered more efficiently in the distillation unit, and regarding the purity of products, distillation offers better results compared to evaporation.Adding additional separation steps in the evaporation unit can improve the purity of products to the detriment of higher equipment and utility costs.
Table 4 presents the summary of the equipment cost, utility consumption, and TAC results for evaporation and distillation units.To define the payback period, a sensitivity analysis is carried out to observe the change in TAC with the payback period variation.According to Figure 6, changes in TACs of evaporation and distillation units are insignificant at a payback period (t PP ) of greater than 5 years.A long t PP is also riskier from an investment viewpoint.Therefore, t pp is considered to be 5 years in this process.
Figure 7A,B shows the share of the equipment costs in evaporation and distillation units, respectively.As mentioned before, the calculated TAC is a small share of the total capital cost.Based on the selected Lang factor, the total capital cost for the production plant can be 3-6 times as much as the equipment cost. [20,43]However, at the preliminary stage, calculating the accurate Lang factor for the pilot-scale process is difficult due to the limited availability of reliable sources, compared to the industrial-scale process.By comparing the results, the distillation process reduces the TAC by approximately 37%, compared to the evaporation technique.
By comparing the economic results of two methods and the separation efficiency of solvents, ethanol and THF, distillation is selected as the main separation method for further optimization and rigorous modelling.

| Optimization criteria
During the solvent recovery step, the separation of organic solvents, THF and ethanol, from the CLPs dispersion increases the concentration of water in the liquid phase.Following the initial solvent recovery, the mixture of organic solvents enters the second separation step, and the recovered THF and ethanol streams are sent to the dissolving and diluting stages, respectively.Obtaining a concentrated THF solution with low water content is an important criterion during the solvent recovery.Lignin can be rapidly dissolved in a solvent containing a high concentration of THF, minimum $85 wt.%. [19]Moreover, the efficient isolation of THF from the feed stream in the second separation step also leads to obtaining significantly concentrated ethanol, $80 wt.%.
The optimization of distillation units starts with minimizing the TAC of the conventional distillation unit by considering the minimum acceptable purity of recovered solvents as the main constraint.In addition, optimization of the extractive distillation unit based on reaching THF concentration of 99.5 wt.% is also conducted to assess rate of increase in the TAC by enhancing the desired purity of the organic product.Therefore, after the first solvent Effect of payback period (T PP ) on total annual costs (TAC) of evaporation and distillation units recovery step, the entrainer is added to the second distillation unit.Regarding the purity of recovered ethanol, the same criteria, 80 wt.% ethanol, is considered to dilute the solution since it will be contacted with water right after the diluting step.Table 5 shows the purity criteria of products, used as optimization constraints, for conventional and extractive distillation models.For optimizing the system, the considered structural factors are the number of stages (N T ), feed stage (N F ), and entrainer feed stage (N S ).Besides, the reflux ratio (RR), distillate rate (D), and entrainer flow rate are also Ethylene glycol -≥99 examined as operational factors.Table 6 shows the levels and the range of different factors for conventional and extractive distillation.

| Operating pressure
In this process, low-pressure steam (3 bar) and cooling water (20-25 C) are used for heating and cooling.Operating pressures for the first and second columns are considered to be in the ranges of 0.25-2 bar and 0.3-4.5 bar, respectively, to fulfil the temperature approach requirements based on the selected utilities.Figure 8 depicts the effect of pressure on the heat duties of the reboiler (Q R1 ) and condenser (Q C1 ) related to column C 1 .As pressure goes up, Q R1 and Q C1 increase.In addition, CLPs in stream F 1 are sensitive to temperature, and operating at higher pressure and temperature can have negative effects on the quality of CLPs.Therefore, this column is modelled to operate at 0.25 bar.Working under vacuum conditions can not only decrease the energy cost but preserve the quality of CLPs as well.
In the case of separating ethanol and THF using the conventional distillation method, reaching the required concentration at the top and bottom products is possible at 1 bar.Therefore, the conventional column C 2 operates at 1 bar.If higher purity is required for organic solvents, it should operate at a higher pressure which can increase the reboiler duty and the utility consumption.
In the case of extractive distillation, Figure 9 shows the effect of pressure on the condenser and reboiler duties of column C 2 where EG is used as the entrainer.Similar to column C 1 , by increasing the pressure, the reboiler and condenser duties also increase.Therefore, column C 2 is designed to operate at vacuum pressure, 0.3 bar.Regarding column C 3 where ethanol is separated from EG, there is a significant difference between the boiling temperature (T b ) of the entrainer and ethanol (ΔT b ¼ 119 C).Based on the heat transfer requirements such as the effective temperature difference between the process and utility stream, the operating pressure of column C 3 in extractive distillation unit is 0.1 bar.

| Selection of column internals
One of the challenges that has to be addressed during process modelling is mitigating the possibility of the accumulation of CLPs and fouling in the first column.According to Pilling and Holden, [44] when the distillation column diameter and liquid rates are less than 60 cm and 50 m 3 /m 2 h, respectively, packings are the preferred column internals. [33]Therefore, in this work, the properties of different types of packings are examined based on their applications.Fouling resistance, liquid distribution, pressure drop, and residence time are important factors for selecting the desired type of packing.Packings with low residence time and low-pressure drop can mitigate fouling formation. [45]Structured grid and certain types of corrugated structured packings with sufficient open areas can address services tending towards fouling formation.However, specific surface areas of grid packings (40-90 m 2 /m 3 ) are limited, and Column 3 Abbreviation: RR, reflux ratio.
they require a large height equivalent to a theoretical plate (HETP) for efficient separation.Some types of metal perforated plate corrugated packings such as the Mellapak series are resistant to moderate fouling, and they also provide a higher specific area (up to 900 m 2 / m 3 ).Therefore, in this study, Mellapak 750 and Mellapak 250 packings are considered for the rectifying and stripping sections in column C 1 , respectively.[48] For both the conventional and extractive distillation units, Mellapak 750 packing is considered the main internal in columns C 2 and C 3 due to the absence of solid particles.Void fractions of Mellapak 750 and Mellapak 250 packings are 0.963 and 0.987, respectively, and the Stichlmair method is employed to calculate the pressure drop and liquid holdup. [49][52] Therefore, applying a suitable distributor to enhance uniform liquid distribution guarantees the better operation of the packed column. [45,53,54]A forced circulation reboiler in column C 1 also offers high turbulence and high shearing forces, leading to self-cleaning effects and minimizing the fouling formation. [40,55]6 | Optimized distillation units Figure 10A,B shows that, for the conventional distillation unit, the optimal number of stages (N T1 and N T2 ) for columns C 1 and C 2 are 16 and 26, respectively, based on the minimum TAC.By increasing the number of stages, the TAC initially decreases and then rises as expected.In distillation, there is a trade-off between the number of stages and the RR.A decrease in the number of stages initially leads to a lower capital cost; however, it is a transitory impact since the increase of the RR requires a larger column diameter and heat transfer areas for the condenser and reboiler.[56,57] Therefore, by decreasing the number of stages, TAC is initially reduced and later increased.
Regarding extractive distillation, the results of column C 1 are similar to the conventional unit, and the optimal total number of stages is 16.The second unit of this distillation model includes two columns to recover highly purified THF and entrainer, and also ethanol with acceptable concentration.Figure 11A-C follows a similar pattern as the conventional distillation unit, and the required numbers of stages (N T2 and N T3 ) are 25 and 16 for columns C 2 and C 3 , respectively.
For the process modelling, the tolerance and accuracy of the initial and final values of the DOE factors are 0.1 and 10 À6 , and the margin of error in the purity of the product calculations is 0.0005.
By obtaining the optimized operating parameters of the conventional and extractive distillation in the equilibrium mode, the rate-based approach is used to increase the accuracy of the simulation results.Figure 12A,B shows the results of the finalized flowsheets for both distillation models including configuration data, stream flow rate, reboiler and condenser heat duties, and operating pressures and temperatures.More detailed information about the streams is presented in Appendix S1.
Table 7 presents the summary of the equipment cost, utility consumption, and TAC results of the optimized conventional distillation and extractive distillation units, modelled based on the equilibrium mode.According to the results, in the case of recovering almost pure THF, The effect of pressure on the reboiler duty (Q R1 ) and condenser duty (Q C1 ) for column C 1 for both conventional and extractive distillation The effect of pressure on the reboiler duty (Q R2 ) and condenser duty (Q C2 ) of column C 2 related to the extractive distillation unit the TAC will increase by about 56% which is substantial when considering the enhanced purification of 11.5%.Moreover, there is a possibility that if the operation of extractive distillation encounters some problems, a fraction of the entrainer remains in the recovered organic solvents, and after recycling them to the dissolution and dilution steps, the remaining entrainer will exist as an impurity in the CLPs dispersion, which reduces the quality of the final product.In this study, the TAC is calculated based on the major items and equipment, and based on the simplified process, the probable accuracy of TAC is predicted to be ±30% at this stage. [56]The finalized flowsheet of conventional distillation is selected as the primary solvent recovery model.
Temperature and liquid and vapour composition profiles along conventional distillation columns C 1 and C 2 are also shown in Figure 13.According to the separation principles of the distillation columns, the temperature and composition profiles are set based on the difference in relative volatility of feed components.Along the column, the volatility of components is lowered from the top down, since the bottom product is rich in the heavy component, having the highest temperature close to the bubble point of the bottom product.The top section of the column is rich in the light key component, and its temperature is close to the boiling point of the highly volatile component.Ethanol, which is the component with intermediate volatility, follows the same procedure and along the column by travelling from the rectifying to the stripping section; the fraction of THF decreases and the ethanol share, which has the next highest volatility, increases while the vapour pressure is not low enough to have more water.Therefore, somewhere along the column, the fraction of ethanol reaches the highest point, and as it gets closer to the reboiler, the water share increases and the ethanol fraction decreases.In the case of multicomponent distillation, this pattern is used to get Based on the current process, recovery percentages of water and organic solvents in column C 1 are 97.5% and 98%, and in column C 2 , ethanol and THF recovery are 89% and 92%, respectively, which are sufficient for dissolving and diluting steps.In columns C 1 and C 2 , the operating pressures are assumed to be 0.25 and 1 bar, and the estimated pressure drops are 8.2 and 6.7 mbar, respectively.Furthermore, the minimum and maximum operating temperatures, 32.8 and 75 C, are also within the acceptable range to prevent temperature cross with utilities.
Regarding the effect of the feed composition, by increasing the CLPs fraction in the feed stage, at a constant flowrate, the share of solvents will decrease.However, due to the possibility of the accumulation and fouling of CLPs on packings, more open and lowerefficiency column internals are required, which can lead to higher operating and capital costs.According to Lintinen et al., [19] CLPs with a concentration of higher than 2.8 wt.% bring about aggregation and precipitation of the formed product.
Furthermore, by increasing the water fraction in the feed stream, the fraction of the organic solvents will decrease.As a result, at constant RR, N T , and feed flow rate, the consumption of cooling and heating utilities would decrease, thereby lower utility costs.By increasing the water fraction, the bottom product has a higher water purity due to the location of the feed point on the ternary diagram of water/THF/ethanol.On the other hand, by increasing the organic solvent share under the same conditions, heating and cooling utilities increase, and due to the composition of the feed point, the quality of the separation of organic solvents from water is not as high as the previous cases.Therefore, a higher RR or N T are required to enhance the separation.Besides, changing the fraction of the organic solvents in the CLPs dispersion can negatively affect the quality of the final product.Regarding the operating conditions of the distillation unit, by increasing the operating pressure of the columns, the operating temperature will also increase due to the enhancement in the boiling point of solvents, as a result more cooling and heating utilities will be needed, as shown in Section 3.4.
The proposed distillation unit and the procedure followed to simulate and optimize the distillation unit offers beneficial information that is required for setting up the pilot plant.It also provides the opportunity to know about column behaviour and product qualities before the practical tests.However, the multi-objective optimization method used in the current model is time consuming.Therefore, employing stochastic optimization methods such as genetic algorithm (GA) can accelerate the convergence and also enhance the accuracy of results.
Furthermore, in contrast to a large-scale process design, there are limited studies about process modelling, column sizing, and economic studies on the pilot scale.Employing the equations and rules of thumb introduced in different chemical engineering handbooks for industrial equipment sizing and optimization can lead to considerable inaccuracy in the pilot plant design and an unacceptable error margin.Therefore, the model and methods used in this research can also be employed in other pilot-scale models that include distillation unit modelling and economic optimization for solvent recovery or removing impurities from the main product.

| CONCLUSION AND FUTURE WORK
For optimizing the solvent recovery unit of the CLPs production process, two separation methods, evaporation and distillation, are modelled on the pilot scale and compared based on their TACs.Designing and testing the model on the pilot scale reduces the uncertainty level of the operating conditions and evaluates the viability of the process before commercialization.Equipment selection and type of column internals are based on mitigating the accumulation of CLPs and fouling.TAC results show that the distillation process reduces the related costs by over 37% compared to the evaporation method.Therefore, distillation is chosen as the main separation method.For the process optimization, the distillation unit is simulated based on two models, conventional distillation and extractive distillation which is modelled to assess the degree of increase in the TAC by further enhancing the organic solvent purity, especially THF.Both methods are optimized based on minimizing the TAC combined with an iterative optimization algorithm to reduce reboiler and condenser duties.The results show that increasing the recovered THF concentration from 88 to 99.5 wt.% through extractive distillation leads to a significant increase (56%) in the TAC.By comparing the results and obtaining the optimal parameters of the distillation unit in the equilibrium mode, the rigorous distillation model is utilized to improve the accuracy of the results and predict the column behaviour before the experimental tests.Based on the final design, the water concentration in the bottom product is 99.5 wt.%, and the purity of the distillate and bottom products in the second column is 88 wt.% THF and 80 wt.% ethanol, respectively, thereby fulfilling the solvent purity requirements.Based on the results, the recovered solvents can be reused in the CLPs production process and hence minimize the waste produced during the operation.
As the continuation of the model optimization, the developed rigorous distillation model can be improved by combining the simulation software with the multiobjective GA to enhance the accuracy of the model parameters due to the capability of global optimization of the GA, thereby overcoming the related convergence issues.By obtaining the optimal operating and structural parameters of the distillation unit, the pilot scale plant will be at the building phase based on the finalized model.By comparing the actual results with simulation outputs, the model will be improved to get more reliable results for larger-scale simulation studies.

NOMENCLATURE
T A B L E 3 Summary of the employed tools and references for sizing and cost estimation of the main equipment and employed utilities Abbreviations: APEA™, Aspen Process Economic Analyzer™ V11; Aspen EDR™, Aspen Exchanger Design & Rating™; SS, stainless steel; TFE, thin film evaporation.a Aspen Exchanger Design & Rating™.b Stainless steel.c Aspen Process Economic Analyzer™.

F
I G U R E 5 Flowsheets of (A) evaporation unit and (B) distillation unit at 1 bar.CLPs, colloidal lignin particles.Cond, condensed.CWR, cooling water return.CWS, cooling water supply.DIA, diameter.LS, low pressure steam.TFE, thin film evaporation solvent recovery unit.The bottom product of the first separation unit enters the ultrafilter to concentrate CLPs and recycle back recovered water for reuse in the TBR.

F I G U R E 7
Details of the equipment costs share for (A) evaporation unit and (B) distillation unit at 1 bar T A B L E 5 Purity criteria for the conventional and extractive distillation methods

T A B L E 6
Design factors and the related levels

F I G U R E 1 0 2 a
Total annual costs (TAC) based on the number of stages (N T1 and N T2 ) for columns (A) C 1 and (B) C 2 in conventional distillation F I G U R E 1 1 Total annual costs (TAC) based on the number of stages (N T1 , N T2 , and N T3 ) for columns (A) C 1 , (B) C 2 , and (C) C 3 in extractive distillation F I G U R E 1 2 Finalized flowsheets for the recovery of solvents based on the rate-based approach for (A) conventional distillation and (B) extractive distillation.CLPs, colloidal lignin particles.Cond, condensed.CWR, cooling water return.CWS, cooling water supply.DIA, diameter.LS, low pressure steam T A B L E 7 Summary of economic analysis for the optimized conventional and extractive distillation units based on the equilibrium mode side product containing mainly the intermediate components with high purity.Regarding the temperature profile, along the column, the volatility increases at the higher stages, and the THF share increases in the vapour stream as it travels to the top of the column, while water, which is the heaviest component, stays in the liquid F I G U R E 1 3 (A) Temperature profile along columns C1 and C2 and vapour and liquid composition along columns (B) C1 and (C) C2.THF, tetrahydrofuran phase and, due to its low volatility, its fraction increases along from the top of the column to the bottom.The slight temperature decrease which happens close to the feed stage is due to the introduction of a feed stream with a different composition and temperature compared to feed stage liquid and vapour properties.