The focus of this project was to implement an optimized heat exchanger network for the acrylic acid production via the catalytic partial oxidation of propylene. Since heat integration and optimization involved only the heat exchangers, the rest of the process units were omitted from our simulation and analysis. Consequently, the basic specifications for our process where obtained from the literature values provided by Turton et al. The original process flow diagram consisted of ten exchangers totaling $236 million in capital and utility costs over a twenty-year plant life span.

Two process optimizations were identified. The re-vaporizing of the condensed vapor exiting E-303 by E-301 and heating of process stream 7 in exchanger E-309 by process stream 22. The integration of E-303 and E-301generated 83.4 GJ/h accounting for the recovery of 89% of the steam used in E-303 and 56% of the steam used in the whole process. The integration of E-309 and E-302 eliminated 7.9 GJ/h of low-pressure steam and reduced the amount of cooling water needed to reduce the temperature in stream 7 by 7.9 GJ/h.

Design specifications were also considered and altered from the base case design. Basic augmentations allowed for more reasonable condensing and reboiling environments and safer operating conditions. Selected heat exchanger types included kettle reboilers, thermosiphen reboilers, floating head condensers and fixed tube-sheet exchangers. Total design and improvements produced a saving of $25.8 million which accounted for 59% of the base case (not including refrigerated water).

The purpose of this study was to explore the heat exchanging aspects of acrylic acid production via the catalytic partial oxidation of propylene. Acrylic acid is produced through a two-step process involving the oxidization of propylene to acrolein and then further oxidation to acrylic acid. Several side reactions occur, including the production of carbon dioxide, water and acetic acid. Other process set-ups including the use of single or multiple reactors and various separation methods have been used to attain the final objective of 99.9% acrylic acid in the product stream. However, since the foci of this study are the heating and cooling conditions, the case set up by the Turton text on pp. 718-719 is considered to be a plausible process. Consequently, all of the exchanger inputs and outputs are based on this process. Through, the use of heat integration and pinch analysis, possible optimized process flow diagrams were devised and analyzed economically in order to determine their feasibility. Several heating and cooling constraints are important in order to avoid explosive conditions and runaway reactions. Heat exchanger heuristics present further exchanger constraints that are vital to proper operation. Such constraints are considered in the process analysis.

For the initial situation described, ten heat exchangers are used. Three of the exchangers are reboilers used to heat the streams leaving from the bottom of the three distilling towers, while three condensers are used to heat the distillate streams from the top of the columns. The remaining four exchangers are used for the reactor, the acrylic acid product stream, and two other towers. The following table better illustrates the locations and uses of the heat exchangers as dictated in the base case.

Equipment

The base case for the system contains ten separate heat exchangers. Simply put, a heat exchanger is a piece of equipment where heat is transferred from a relatively warm fluid to a relatively cooler fluid. The simplest and most common type of heat exchanger is the fixed tubesheet. Three of the ten exchangers used in the base case of the facility are fixed tubesheet, while the remaining seven are floating head exchangers. Like any other type of shell and tube heat exchanger, both the fixed tubesheet and the floating head exchangers consist of several key parts: tubes, tube sheets, the shell, and baffles. Also present in heat exchangers but of less importance to the optimization at hand are shell-side nozzles, tube-side channels and nozzles, channel covers, and pass dividers.

Because they are the surface for the heat transfer in the system, the tubes are the most important component of a heat exchanger. An important decision to make when designing a heat exchanger is what metal the tubes will be constructed from. Corrosion of the tubes inside an exchanger is a serious problem, as it restricts the flow of fluid and thus the exchange of heat. The heat exchangers specified in the base case are composed of either stainless steel or carbon steel. Stainless steel is considerable less corrosive than carbon steel, withstands high temperature better, and is not much more expensive. Consequently, for the optimized case, stainless steel tubes were used in all of the exchangers. Another important aspect of the tubes is the presence or lack thereof of fins. In order to make the surface larger, fins, essentially jagged edges, are often added; low-fin surfaces can provide up to five times as much surface area for transfer as tubes without fins. By adding fins, one can increase the area of transfer substantially without making the tubes considerably larger in area. More detailed information on the costs and benefits of fins were not available, but are a possibility for further research to determine whether the benefits of the additional surface area are worth the added capital investment.

The tube sheet itself is usually a round plate of metal that is the link between the tubes and the shell. The tubes are held in place inside the exchanger by being inserted into holes in the tube sheet and either welded into place or fit into grooves. Although welding provides a stronger joint, welding has two primary disadvantages: the tubes can not be easily removed and the tubes and the tube sheets must be composed of the same or very similar materials. This can become costly if corrosion occurs in the tubes and the tubes need to be replaced or if an expensive, strong material is needed for the tubes.

The shell is simply the container for the fluid on the outside, or shell-side. Generally, the shell has a circular cross section and is either made from a single, rolled sheet of metal or by cutting an existing pipe (for smaller diameters). Once again, the type of metal used for the shell is very important, although the tube-side fluid is often more corrosive than the shell-side fluid. The base case specifies for a mix of carbon steel and stainless steel, while the optimized case replaces the carbon steel with the more durable stainless steel.

The baffles in the heat exchanger serve two crucial functions: to support the tubes in the proper location during the operation of the exchanger and to "steer" the shell-side fluid along the array of tubes, increasing both velocity and the heat transfer coefficient as the fluid moves. Both the shape and the size of the baffles used can affect the ability of the exchanger to transfer heat. The presence of the baffles limits the volume of shell-side fluid that can be present in the exchanger, decreasing the amount of heat transfer, but a proper spacing of the baffles can allow more fluid to flow more quickly. Once again, the internals of the exchangers used for the base case are unavailable. As another possibility for further research, different arrangements and sizes of baffles can be compared and contrasted.

Types

As its name suggests, the fixed tubesheet heat exchanger is simply an exchanger in which the tube sheet is attached to the tube and the shell with an expansion joint. Although simpler and cheaper in initial investment than other exchangers, fixed tubesheet exchangers are more prone to replacement due to internal problems or corrosion. On the contrary, a floating head exchanger contains a bundle that is removable and replaceable. There are several different types of floating head exchangers available on the market and each have their strengths and deficiencies. For example, the "pull-through bundle" type contains a tube sheet that is small enough that it can be pulled through the shell for cleaning and inspection. However, in order to allow for clearance so that the bundle could be removed, several tubes needed to be eliminated. So, a balance must be found between the need for cleaning and replacement with the added cost of the exchanger and the decreased area accessible for heat exchange.

In order to alleviate the burden of simulating the whole process and then trying to equate all of the process stream information between the three groups, only the exchangers were modeled. Through the information provided by Turton et al. about the inlet and outlet streams of the exchangers, each exchanger was solved independent of the rest of the process equipment. Consequently, analysis and optimization of the exchanger network only involved the heat exchangers.

Validity of the Peng-Robinson EOS

Furthermore, Hyprotech has put forth the largest amount of effort in enhancing this property package for HYSYS. The Peng-Robinson EOS has been improved to yield accurate phase equilibrium for very large temperature and pressure ranges. Such ranges far surpass any of the other EOS data that HYSYS offers. Other property packages lose some of their accuracy due to their additional efforts towards simulating non-ideal scenarios⁶.

Validity of using UNIQUAC

UNIQUAC uses statistical mechanics and the quasi-chemical theory of Guggenheim to represent the liquid structure. Representation of LLE, VLE, and VLLE comparable to the NRTL equation is obtained by UNIQUAC. However, the incorporation of the non-randomness factor is not need with the UNIQUAC activity model.

Since activity models only perform calculations for the liquid phase, an equation of state is required to solve for the vapor phase. Consequently, Uniquac in conjunction with Peng-Robinson were utilized in our simulation work.

Figure 2 depicts the cascade diagram for the process. The cascade diagram shows the net amount of energy at each temperature interval. However, the temperature intervals are not consecutive as the temperature interval diagram shows. Consequently, the various zones represented are those for which actual energy is transferred. It is possible to transfer energy down the temperature gradient. Any excess energy can simply be cascaded down to the next interval. Turton et. al. describes that one will eventually reach a point where the energy transfer is not possible. This point turns out to coincide with the eventual pinch point and is the point where cold and hot utilities have to be employed to achieve energy equilibrium.

Calculated Diagram

A composite temperature-enthalpy diagram was constructed from the calculated values. Such curves should indicate the plausible pinch point. Table 2 and Table 3 indicate the cumulative enthalpy of the hot and cold streams for each temperature interval. Due to the spanning of the temperatures by the process streams, some zones had no energy exchange associated with them. However, for the purposes of correlating temperature intervals, they were included.

Table 2: Cold Stream Composite Energy

Table 2: Hot Stream Composite Energy

From the tables, it is evident certain temperature ranges exhibit no enthalpy changes. This is due to the modeling of the phase changes and the temperature ranges in which they deal with. Indeed when the temperatures are plotted versus the cumulative enthalpies, the curves for the hot and cold streams exhibit very strange behavior. This information is depicted in Figure 3. The curves cross at three different points. However, it is important to note that these crossings are at the three zones Figure 2 depicted as the possible pinch zones.

It is clear that pinch analysis to this point lacked in dealing with the phase changes and the large temperature profiles in the process. However, it has identified large sources of heat and possible streams that could benefit from such sources. Nonetheless, the analysis still has not provided accurate pinch temperatures for the hot and cold streams. Consequently, the use of HYSYS’s pinch analysis was employed.

HYSYS Calculated Diagram

Figure 4 shows the temperature-enthalpy curves for the hot and cold streams. HYSYS obviously dealt with the phase changes differently and in turn produced curves that are more suitable. The exact methodology it employed is not too clear. The two curves employ all of the possible exchangers. If one adds the exchangers one by one and views the changes in the curves between additions, it seems that their depiction of energy at the different intervals are accurate.

Table 4 shows the analysis outputs. The hot and cold pinch temperatures do correlate with the speculated temperatures obtained by the calculated pinch analysis. It is evident that although the calculated pinch analysis identified three possible pinch zones, HYSYS was able to account for the enthalpies accordingly and pick the most plausible pinch zone.

The easiest way to determine the number of exchangers below and above the pinch is to draw boxes representing the energy in the hot and cold streams. If the energy can not be accounted for through the cold-hot integration, then it can be compensated for through the addition of cold and hot utility. The interconnecting arrows indicate the energy transfers between the streams. Consequently, the total amount of heat transfers is indicative of the total amount of exchangers..

It is important to note that the configuration used to determine the number of exchangers is not the only possible one. However, all of the possible configurations do reveal that ten exchangers are needed for our process.

This process presents only limited opportunities for heat integration. One of the fundamental requirements for heat transfer between process streams is that the cold streams (i.e. the streams to be heated) be at a lower temperature than the hot streams (i.e. the streams to be cooled). The table below, which delineates the conditions of the streams requiring heating or cooling, demonstrates that the opposite is typically the case in this process.

To exchange heat, Tin for a hot stream must be greater than Tin for a cold stream. Thus, only two general heat integration possibilities are available. Hot stream 5 may exchange heat with any of the cold streams, while cold stream 22 may exchange heat with any of the hot streams, except stream 15. No other heat integration is possible.

The energy available in stream 5 can be used to heat any of the cold process streams, thereby saving steam usage. If cost is to be minimized, the expense of steam dictates that all the available heat of stream 5 be used. Of the four cold streams, only stream 14 (passing through reboiler E-303) requires enough heat input to sufficiently use all the available energy of stream 5. Thus, it appears that the solution to the problem is to exchange heat from stream 5 to stream 14 in E-301, and then finish heating stream 14 in E-303 with low-pressure steam (LPS). However, this setup could lead to serious control problems. For example, if conditions in the separation tower-producing stream 14 suddenly changed, stream 5, the molten salt stream responsible for cooling the reactor, could suddenly increase or decrease in temperature. Possible consequences of this include a rapid decrease in reactor conversion, or, far worse, a runaway reaction and explosion.

With these control issues in mind, the scheme below was devised as a safer alternative to direct heat exchange between streams 5 and 14.

Figure 5

Here, stream 5, rather than directly heating stream 14, produces steam, which can then be recycled to feed into E-301 and heat stream 14. E-301 produces 2166 kmol/h saturated low-pressure steam, only slightly lower than the 2442 kmol/h required by E-303 (thus, the excess saturated liquid leaving the splitter). The operation of E-301 can be controlled independently of E-303 be changing the split fraction in the splitter, thereby alleviating the control nightmare described above. This heat integration saves 276 kmol/h of steam, and 109,500 kmol/h of cooling water (used to cool stream 5 in the base case).

As expected, the cost savings are tremendous. The chart below demonstrates that $24.3 million is saved by tying the two exchangers together, equivalent to 88% of the base case cost of the two exchangers! Exchanger E-301 essentially produces $23.3 million worth of low-pressure steam over the twenty-year lifespan of the plant, represented by the "negative" operating cost of the exchanger. After integration, E-301 requires more area than in the base case because the temperature difference between stream 5 and saturated steam is less than the temperature difference between stream 5 and cooling water (i.e. the temperature gradient has decreased). Thus, the capital cost of E-301 after integration is larger than its cost in the base case, but this is trivial compared to the utility savings of reduced steam use.

As discussed above, stream 22 may exchange heat with most of the hot streams. Stream 7 was chosen for two reasons. First, stream 7 is physically located very near to stream 22, reducing the need for piping all across the plant. Stream 22 is the solvent feed to the liquid-liquid extractor, while stream 7 is the high-volume stream leaving the quench tower, part of which will be diverted to the same extractor. Second, problems of control are not essential for these two streams. The extractor does not need to operate at exactly 40 degrees, so the exact heating of stream 22 is not an issue. In addition, stream 7 is so large that the cooling resulting from heat exchange with stream 22 is not too significant; a great deal of cooling water is still necessary to cool stream 7 all the way to 40^oC. As shown in the PFD below, stream 22 is only capable of cooling stream 7 to 48.9^oC.

Although stream 22 has only a limited effect on stream 7, the cost savings are significant, as shown in the table bellow

Table 6: Integration Savings

The total savings are $2.3 million, equivalent to 63% of the base case cost of the two exchangers. As was the case with E-301, the bulk of the savings come from a reduction in steam needs. Instead of heating stream 22 with low-pressure steam, stream 7 can do the job, eliminating $2.2 million in steam costs. The remainder of the savings are the result of the decreased cooling water needs for stream 7.

Typical overall heat-transfer coefficients were obtained for fluids resembling the physical properties and interactions of our fluids. Though rigorous calculations deriving these numbers were not performed, the selected values proved to provide reasonable areas using the HYSYS calculated UA values. Furthermore, the areas agreed well to those provided by Turton et al.

The correlation accounted for fluids flowing through the shell-side and tube-side. Approximate U-values were also provided for systems experiencing phase changes, which were implemented in the fluids flowing through our reboilers and condensers.

Turbulent flow is vital to obtain good heat exchange between to fluids; consequently, the Reynolds numbers for all of our flowing fluids were calculated. The following equation was used to calculate the dimensionless values.

Number of Tubes

An average tube length of 16 feet was chosen in accordance with the heuristics provided by Turton et al. A tube diameter of 0.75-inch OD pipe was chosen since it proved successful in providing adequate Reynolds numbers for our flows. The average diameter of the tubes were estimated by taking the mean of the inner and outer diameters of the pipe. The total areas of the tubes were then determined by dividing the total exchange area by the area of each tube.

Baffle Spacing

Minimum baffle spacing was suggested by Perry’s handbook to be about one-fifth of the shell diameter and no less than two in. In order to provide the adequate support for the tubes within the exchanger, the maximum unsupported tube span in inches should equal no more than 74*D^0.75 (where D is the outer diameter of the tube). Since our material of construction consist of sturdy stainless steal and our flows were sufficiently turbulent, most of our exchangers were incorporated with the maximum baffle spacing. Furthermore, all of our baffles were placed vertically in order to prevent entrapment and entrainment.

Materials of Construction

The process consisted of various condensing and reboiling stages; consequently, in order to aid this phenomena the process stream was fed through the shell-side of each exchanger. Since our components were considerably corrosive, the decision was made to convert our shell-sides to stainless-steal. Although the cost for each exchanger increases due to the modification, it change was deemed necessary in order to prevent safety complications and inhance phase transitions in or process streams.

A summary of our specifications of all of our exchangers is provided in Appendix I.

The shell-side pressure drop was calculated for exchanger E-310. According to Kern². , the change in pressure is given by the following equation:

The change in pressure for this exchanger was found to be 2.7x10-² psf.

The tube side pressure drop for E-310 was determined using the standard Fanning friction factor method as described in Perry’s. Given a Reynolds number of 4.2 *10⁵ and a relative roughness of .0026, the friction factor was found to be .0063. A head loss of 0.14m²/s² was calculated using the following equation:

H_L = 0.5 * f * L/D * V²

Multiplication by the liquid density of 1004 kg/m³ yielded a tube side pressure drop of 138 Pa, equal to 0.02 psi.

Capital Costs

The size parameter needed to cost a heater is the heat transfer (m²). The constants are listed in the book and there are no deviations from the standardized equations.

Utility Costs

The utility costs were calculated using the utility information and formulas given in the text on page 87 (Turton). The basic formula used to calculate the yearly cost for each unit was

Yearly cost = Duty * Cost of utility used * time * Stream Factor (SF)

Where duty is heat in the case of heat exchangers and electric power for pumps and SF is the percentage of the year that the plant is open. The duty values are given in table 3.4 (Turton) and are in the units $/GJ or $/kWh. SF is taken to be 0.95 as suggested by the book.

Heater Utility Cost

To calculate the utility cost for a heat exchanger the duty in GJ/hr is substituted into the yearly utility cost equation. The two types of heating/cooling medians used in our project are low-pressure steam, cooling water and refrigerated water costing $3.17/GJ, $0.16/GJ and $20/GJ respectively.

Incremental Economic Analysis

Once the yearly utility cost per unit is calculated, this value must be adjusted to correspond to the total operating cost for the assumed life of the plant, which we set to be 10 years. We accomplished this by doing an incremental economic analysis. The theory behind an incremental economic analysis is that the cost of utility each year remains constant although inflation continues to rise. Therefore, one is paying the utility companies the exact same amount of money each year; however, this money is worth more due to inflation, so the utilities cost less. Using a 7% discount rate, we solved the following equation given in the book to calculate our discount factor.

In conclusion, our total costs including design modifications and our integrated process may have increased our capital costs, but over a Twenty year life span, the savings are considerable and well over $25 million as depicted in Chart 3

Turton, Richard; et al. Analysis, Synthesis, and Design of Chemical Processes. Prentice Hall: Upper Saddle River, New Jersey. 1998.

GPSA Engineering Data Book. Gas Processors Suppliers Association, 1987.

Perry, Robert H. and Green, Don W. Perry's Chemical Engineer's Handbook. McGraw- Hill. New York. 1997.

Kern, D. Process Heat Transfer. New York: McGraw, Hill, 1950.