Team Leader: Scott Covan
Engineers:
Mondro Barman
Daniel Resendez
Table of Contents
Rigorously Designed Exchangers
Appendix A: Temperature Cascade Calculation Software
The process design team was assigned the tasks of designing the necessary heat transfer equipment and performing heat integration analysis on a base case design for an acrylic acid plant to reduce overall production costs. The plant is to produce 50,000 metric tons per year of 99.9 mole % pure acrylic acid via the catalytic partial oxidation of propylene. The plant will operate for 8,000 hours per year and will have a twenty-year operating life span. After designing the heat transfer equipment, the projected capital costs of the heat transfer equipment were found to be $3,732,000 with utility costs of the base case plant totaling $117,606,000 over the 20 year life span of the plant. A discount rate of 10% and a Chemical Engineering Plant Index of 382 were used. The heat integration analysis showed that only limited heat integration was possible; only two process streams were involved in the heat integrated process. The improved process utility costs for the life span of the plant are $117,341,000, and the capital cost of the heat transfer equipment is $2,648,000. Incorporation of the heat integration plan provided in this report will save $1,349,000 over the life span of the plant
The cost of utilities over twenty years for an acrylic acid plant producing 50,000 metric tons per year is substantial. Good heat exchanger designs and heat integration are thus very important because they can reduce utility costs substantially. The base case plant design included a total of 8 exchangers. Four of these exchangers were pre-heaters or coolers, while two were condensers and two were reboilers. Seven of the eight heat exchangers were selected for heat integration analysis because the exchanger attached to the reactor cooled molten salt and would therefore be difficult to model.
Acrylic acid is produced by the partial oxidation of propylene in a two-step process where propylene is oxidized to acrolein and oxidized again to acrylic acid. This reaction takes place in two independent reactors with different catalysts. Once the reaction occurs the products are quenched with dilute aqueous acrylic acid to prevent further oxidization. The tops are then sent to an absorber, which captures acrylic acid and acetic acid, a by-product. The bottoms of the absorber and the quench are combined, cooled down and sent to a liquid-liquid extractor. In the extractor, the acid fraction is preferentially removed from the aqueous section by the solvent, ethyl acrylate. The organic tops of the extractor are sent to a solvent recovery tower where the ethyl acrylate is removed from the tops and returned to the extractor. The bottoms is a mixture of acrylic acid and acetic acid with very few impurities. This stream is sent to a final purification distillation tower, which produces 95% by mole acetic acid from the tops and 99.9% acrylic acid from the bottoms. The bottoms are further cooled and sent to storage.
Calculations involving the preliminary calculations for heat integration such as the cascade of temperature intervals were performed as described in chapter 7 of the Seider text (Seider 243). A MATLAB program was created to perform these calculations; the code for this program is available in Appendix A. Exchanger design calculations were performed using Aspen Tech’s Aspen Plus 10.1 software.
All of the exchangers in the base case were designed using the shortcut method. This involved assuming a heat transfer coefficient from the literature (Seider 328) based on the state and type of fluids flowing through the exchanger. This U value was then plugged into the governing equation of heat transfer:
Q = U*A*MTDe
where Q is the rate of heat transfer, U is the overall heat transfer coefficient, A is the area through which heat transfer occurs, and MTDe is the effective mean temperature difference between the hot and cold fluid. Use of this equation was further simplified as simulation in HYSYS provided a UA for each exchanger modeled. This required only that this UA value be divided by the assumed U value to obtain the required area of heat transfer. This area value was used when pricing exchangers.
Two of the exchangers, E-302 and E-309, were rigorously modeled in the optimized case. Rigorous design required the specification of exchanger dimensions, number of tubes and tube diameters, pitch, and baffle placement.
The first step in rigorous design was to chose the number of tubes and tube diameter to achieve appropriate pressure drops and Reynold’s numbers on the tube side of the exchanger. Pressure drops were not allowed to exceed 70 kPa, and Reynold’s numbers were maintained at least one order of magnitude above the necessary Reynold’s number for turbulent flow, about 4,000. Choice of tube length was found to have only a small effect on pressure drop through the tube sheet compared to choice of tube diameter. The effect of tube length on Reynold’s number was even smaller.
The next step of the design was to choose shell size and baffle placement in order to meet the same pressure drop and Reynold’s number demands as those for the tube side. Once all of these requirements were met, heat transfer area was considered. Aspen Plus, the simulator with which all rigorous design was carried out, calculated a necessary heat transfer area based on the state of the streams moving through the exchanger. This was compared with the actual heat transfer area in the exchanger as specified considering pressure drops and Reynold’s numbers. If the exchanger was underdesigned, other characteristics of the exchanger were changed in order to meet that specification. Typically, these changes consisted of changing tube and shell length and increasing the number of tubes. This, in turn, changed pressure drops and Reynold’s numbers. An iterative process formed in which the specifications of the exchanger were repetitively changed in order to meet all specifications. In addition, both rigorously designed exchangers were given ten percent more heat exchange area than what was required in order to take account of exchanger fouling and aging.
Equipment was priced using the program CAPCOST. The exchanger costs were based on the heat exchange area, type of heat exchanger and materials of construction. The heat exchange area was calculated by obtaining a UA value from HYSYS and dividing by a heuristic U. The exchanger types used were the types specified by the Turton text (Turton 716-727). The only deviations from these specifications were when the exchanger areas were less then 10 m2. Double pipe heat exchangers were used in these cases because they were more cost effective. Materials of construction specifications were also taken from Turton. Carbon steel tubes are less expensive then stainless steel tubes, so these were used for non-corrosive materials service. Below is a table of exchanger specifications:
Utilities costs were obtained from the Seider text (Seider 376). The twenty year net present value of utility costs was calculated using a ten percent discount rate.
Due to the highly corrosive nature of acrylic and acetic acid, 314 stainless steel was used for all piping, exchanger tubes and other equipment in acrylic acid and acetic acid services. Carbon Steel was used for all non-corrosive services. This combination provided the most cost-effective and safe materials for this process. All low pressure steam exchangers, reboilers and piping were insulated to prevent heat loss.
The base case was derived from the model presented in Turton, et al. 1998. Seven heat exchangers were considered for the base case. These are presented below:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
E-304, E-303 and E-302 are the largest exchangers in the system with areas of 6750, 1150 and 340 m2 respectively. The flows for these exchangers require the most duty, and therefore require more transfer area and greater utilities. Figure 1 is a PFD of the process and Table 1 is a simplified economic analysis.
|
|
||||
|
|
|
|||
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
(BMC = Bare Module Cost) |
||||
|
|
|
|||
The first step in the heat integration procedure was a cascade analysis. A cascade analysis provides two valuable pieces of information, the location of the pinch and the minimum utilities required when using only conventional stream crossing. The pinch indicates the hot and cold temperatures at which heat exchange would violate the minimum approach temperature difference. The minimum exchange temperature was designated as 10 C by the professor. The minimum utilities provide a limit to the amount of improvement possible, and allow the engineer to gauge the effectiveness of any proposed optimizations. The creation of a cascade diagram is described in the Seider text (Seider 247-250).
One of the important challenges the team found in creating the temperature cascade was appropriately handling the various phase changes in the process. Complications arise because the cascade analysis requires that streams exist in only one phase and have a relatively constant heat capacity. The process included 4 multicomponent streams experiencing phase changes. After a great deal of research, the team found that streams with phase changes were best handled by approximating them as multiple pseudo-streams with different heat capacities. Pseudo-streams were chosen such that linear approximations on a heat flow versus temperature plot all had R values of at least .97. This approach is described further in reference (Turton 539). Table 2 presents a comprehensive list of the final stream breakdown with the hot temperatures adjusted by –10 C, the minimum approach temperature difference. The resulting cascade is presented in Figure 2.
The cascade analysis indicates a pinch at 74 C (cold side). This can be seen by the fact that no energy flows into the next temperature interval. The hot and cold utilities can be found at the beginning and the end of the cascade respectively. It is important to note that the total duties of all cooling and heating exchangers in the base case were within one percent of these estimated minimum utilities duties. From this it is safe to assume that little utilities can be saved by conventional stream matching.
After performing the cascade of temperature intervals, hot and cold composite curves were generated and analyzed in order to verify the calculated pinch temperatures. To generate the cold composite curve, all cold streams were assumed to have a heat flow of zero at their lowest temperatures. Then, using the mCp’s calculated during the cascade of temperature intervals analysis, a cumulative heat flow was calculated for each temperature interval through which each stream passed. The total heat flow for each temperature interval was found by summing the cumulative heat flow of each cold stream that passed through that interval. The result was a set of temperatures and heat flow values that could then be plotted against each other. A similar method was used to generate a hot composite stream.
To find the pinch, the cold composite stream must be horizontally shifted in order to meet two criteria. First, the cold stream must lie below the hot stream at every point. Second, the minimum vertical distance between the hot and cold stream cannot be smaller than the minimum temperature approach. The location of that minimum approach on the graph is the pinch. Figure 3 shows the hot curve, cold curve, and shifted cold curve that our analysis generated. As the figure shows, the pinch occurs at a section of the graph that is very nearly vertical. Figure 4 is a detail of Figure 3 demonstrating that, as our cascade of temperature intervals predicted, the pinch temperatures of our process are at about 74 and 84 degrees Celsius.
Stream matching was the first method used to reduce the cost of the heat exchanger network. Figure 5 shows the duties and inlet and outlet temperatures for each exchanger. As the figure demonstrates, the cold streams in the process are generally hotter than the hot streams. This situation makes conventional stream matching impossible. The only opportunity for heat exchange between process streams arose from stream 309. Because the duty in E-302 was so much greater than the duty in E-309, and because the temperatures were sufficiently far apart, heat was exchanged between streams 302 and 309 in the optimized process. This removed the use of steam in E-309, and slightly reduced the use of cooling water in E-302. These savings were accomplished without the need of any additional heat exchange equipment.
Stream Profiles
In addition to the conservative stream matching approach, several innovative designs were considered. Heat pumping, vapor recompression, and reboiler flashing designs were evalutated, but none were found to be cost effective. Each of these methods require compression, and thus they are only cost effective if the savings in utilities are not offset by increased cost in equipment.
Use of steam after it has condensed was considered as a method for maximizing the use of that steam. However, the cost of steam depends on the duty that must be resupplied to the used steam in order to return it to its original state. Use of steam beyond its condensation point would simply increase the amount of heat that must be transferred back to the stream. This would change the cost of that steam. Because of this, this method was not seen as a way of saving money in our optimized design.
Heat pumping consists of a closed loop containing a refrigerant that removes heat from the condenser and delivers it to the reboiler. The heat pump is necessary because the temperature in the condenser (hot stream) is lower than the temperature in the reboiler (cold stream). Heat transfer is thus impossible, because the second law of thermodynamics dictates that heat flows from hotter streams to colder streams. This problem can be resolved with a heat pump because the pressure of the refrigerant can be adjusted so that it is hotter than the condenser feed and colder than the reboiler feed. The refrigerant is initially colder than the condenser feed and isothermally absorbs heat from the condenser feed and vaporizes. The refrigerant is then compressed so that it condenses and converts its latent heat to sensible heat thereby raising the temperature of the refrigerant above that of the reboiler feed. The refrigerant thus transfers heat to the reboiler feed, and is cooled down as much as possible. Additional heat is removed from the condenser in a cooler, and the refrigerant is passed through an expansion valve to decrease the pressure to that of the condenser. The cycle repeats itself continually. Unfortunately, the cost of compression is usually very expensive, and the addition of large compressors decreases the reliability of the process.
A heat pump was implemented for the acrylic acid process. A preliminary analysis was completed using the refrigerant freon 11. Although the team recognized that CFC’s are hazardous to the environment, the team initially wanted to use an efficient refrigerant to see if a heat pump would be effective before beginning to search for an environmentally safe refrigerant. The heat loop included two supplemental exchangers to cool down the refrigerant further and to heat the reboiler feed. The heat pump did not prove to be cost effective. A simplified PFD is presented in Figure 6, and the economics are presented in Table 3. A detailed PFD is included in Appendix B.
Vapor recompression is a method of exchanging heat between the reboiler and the condenser of the same tower. It involves compression of the vapor from the top of the tower, exchange with the fluid to be reboiled at the bottom of the exchanger, and throttling of the overhead flow down to its condensed state.
Two attempts at vapor recompression were made. A special consideration that had to be made was that the reboiler duties of each column were greater than their respective condenser duties. To account for this in the first vapor recompression design, the stream to be reboiled from the distillation tower was split and only one of those streams went through exchange with the recompressed vapor stream. To further make use of the heat in the system, stream 309 exchanged heat with the compressed vapor stream before it was throttled and cooled with cooling water. A simplified PFD of the process is included as Figure 7. The economics of the system are summarized in Table 4. The total cost of the heat exchange equipment and utilities related to the distillation tower and E-309 in the base case was about 3.4 million dollars. The cost of the vapor recompression system was 15.9 million. Clearly, this design did not save any money. The most significant cost of this design was the compressor and compressor drive. Considering the twenty year life span of the plant, a spare compressor was also costed. The total cost of this equipment was 13.4 million, not including the cost of the power to drive the compressor.
Considering the high cost of the compressor in the first vapor recompression system, a second system was designed in which only part of the vapor overhead was compressed. This was done in an effort to reduce the size of the compressor, thereby reducing its cost. The uncompressed stream went through typical cooling water exchange, while the compressed stream exchanged heat with the stream to be reboiled, was throttled down to normal pressure, and cooled off with cooling water. Figure 8 represents a simplified PFD of the process. Table 5 summarizes the economics of the system. The cost in the base case of the related equipment was 3.1 million, while the recompression design cost 15.0 million. Vapor recompression proved itself an inefficient method of heat exchange for our process.
Reboiler flashing is another method of exchanging heat between the reboiler and condenser of the same column. In reboiler flashing, the stream to be reboiled is instead throttled through a valve until its temperature is low enough to condense the overhead vapor. The stream is then compressed back to the correct pressure and recycled back into the bottom of the column as a vapor. This method of exchange was inappropriate for the acrylic acid process, because the towers used operate close to a vacuum, thereby making throttling of the reboiler stream impossible.
None of the alternative methods of heat exchange that were explored for the acrylic acid production process were profitable. There were a number of reasons that these exchange methods were considered unadvisable:
Rigorously Designed Exchangers
Two of the exchangers in the optimized design were rigorously modeled using Aspen Plus. Using the methodology previously described the dimensions outlined in Table 6 were determined to be those maximizing the efficiency of heat transfer. As the table demonstrates, the pressure drops calculated by Aspen Plus were within realistic limits and the Reynold’s values for the two rigorously designed exchangers were high enough to be in turbulent flow. The exchangers were overdesigned to take account for exchanger fouling and aging.
The optimized case is similar to the base case. A PFD of the process is included as Figure 9. After considering other options, the only feasible deviation from the base case was heat integration of E-309 with the quench/absorber product stream. This eliminated the need for low-pressure stream in E-309. These results agree with the results of the pinch analysis. The optimized case saves 1.1 million in capital costs and 300,000 in utility costs over the 20 year NPV. This resulted in an total overall savings of 1.4 million dollars. Table 7 represents a more detailed economic analysis.
|
Optimized Case Economics |
||||
|
|
|
|
||
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
(BMC = Bare Module Cost) |
||||
|
|
|
|||
Plate and frame heat exchangers were examined for condenser and liquid-liquid services. Plate and frame heat exchangers are up to five times more efficient then shell and tube heat exchangers. The continual change in flow direction and the increased turbulence created by the ridged plates results in a larger Reynolds numbers and greater heat transfer efficiency. The compact dimensions of these exchangers take up much less plant real estate allowing for greater flexibility for placement of the exchanger. The flexible construction also allows for easier cleaning and expansion. Because the unit can be taken apart in the field, maintenance costs are reduced and expansion or reduction of size can be performed easily be adding or removing plates.
Unfortunately, we could not integrate these exchangers into our final design. Design details and accurate pricing information could not be obtained. Further research on this type of exchanger would be recommended before the design of the plant is finalized.
After thorough analysis of the base case design, and numerous attempts to produce savings using both conventional and alternative heat exchange systems, it was determined that little could be done to optimize to heat exchange network of the process. By rigorously designing two exchangers and exchanging heat between E-302 and E-309, 1.3 million dollars were saved over the life span of the plant.
Alfa Laval Web Site. Alfa Laval Products. (http://www.alfalaval.com/group/grb1r.htm).
Seider, Warren D., et al. Process Design Principles. New York: John Wiley and Sons, Inc., 1999.
Turton, Richard, et al. Analysis, Synthesis, and Design of Chemical Processes. Upper Saddle River, New
Jersey: Prentice Hall, PTR, 1998.
Temperature Cascade Calculation Software
%Cascade Creation Software
% DataMatrix's rows are of the form [temp in, temp out, mCp] for each stream
% temps is a column vector of all the temperature streams in order from highest to lowest
% Qsteam is the amount of steam added to the process
% Qtotal is the amount of cooling water needed in the process
function [Q_total]=heatcalc(DataMatrix, temps, Qsteam)
num_outloops=length(temps);
num_inloops=length(DataMatrix);
%Initialize the total heat and heat history for this program run
Q_total=Qsteam;
Q_history=0;
%Loop through all of the key temperatures
for k=1:num_outloops
compare=temps(k);
%Initialize the heat for this pass
Q_pass=0;
%Loop through all of the streams (j's) for 1 temperature
for j=1:num_inloops
%initialize the heat value for this particular stream
Q=0;
if DataMatrix(j,1) ~= DataMatrix(j,2);
if (DataMatrix(j,1)-compare) > 0
Q=DataMatrix(j,3)*(DataMatrix(j,1)-compare);
DataMatrix(j,1)=compare;
end
if (DataMatrix(j,2)-compare) > 0
Q=(-1)*DataMatrix(j,3)*(DataMatrix(j,2)-compare);
DataMatrix(j,2)=compare;
end
Q_pass=Q_pass+Q;
Q_total=Q_total+Q;
end
%end the first loop
end
%Process Delta H for this pass
Q_history(k,1)=k;
Q_history(k,2)=temps(k);
Q_history(k,3)=Q_pass;
Q_history(k,4)=Q_total;
disp(temps(k))
DataMatrix
end
%end the second loop
disp ('-----------------------------------')
disp (' temp, H, R')
disp(Q_history)
disp ('-----------------------------------')
disp ('Q_cw for the process:')