Millennium Engineering Group
Dave Smith
Carl Williams
Karen Waldner
This project involves the analysis and optimization of a heat exchanger network for the production of acrylic acid via the catalytic partial oxidation of propylene. The goal is to design the best system--in terms of cost, safety, and reliability--to achieve the process stream sensible and latent heat requirements specified in the design by Kangas et al. Note that the scope of the optimization herein precludes the reaction and separation processes of acrylic acid production, and also does not consider the reactor heat exchanger because of modeling difficulties. Initial review of Kangas' design reveals that, given the inlet and outlet temperatures for each exchanger, few opportunities for process stream heat integration exist. Nonetheless, three significant improvements over the suggested design were made.
Applying all three of our safe, reliable engineering optimizations, Millennium's clients will save over $32 million, or 35%, off the total 20-year heat exchange cost of Kangas' suggested design.
Acrylic acid is produced via the catalytic partial oxidation of propylene in a two step process. The propylene is initially oxidized to form acrolein, which is then further oxidized to obtain acrylic acid. The reactor effluent contains a number of impurities due to side reactions and unreacted raw materials, and must then be fed through a number of separation columns. In addition, a heat exchanger network is needed in order to achieve temperature constraints inherent to the process and to provide the condensers and reboilers with heat for the distillation towers. An example of such a process, by which 99.9% pure acrylic acid is produced, is found in Turton. The purpose of this investigation is to analyze in detail the heat exchanger network required for the process and to optimize it. A process design presented by Ryan Kangas et al. will be taken as the base case of the system, and the process stream input and output for each exchanger are based on the stream specifications of this case. By employing the methods of heat integration and pinch analysis, several potential optimized heat exchanger networks will be designed. The economic feasibility of these possibilities will then examined in order to determine which design is both technically and economically superior.
Millennium Engineering was given was the design of Ryan Kangas', et al., which optimized the separations section of the process. The process begins with air, steam and propylene being mixed together and sent to a vertical fluidized bed reactor. The reactor contains heat transfer tubes that are filled with molten salt in order to facilitate proper heat removal. The reactor effluent is then quickly cooled in a quench tower and sent to an absorber fed with deionized water, where all the gases such as carbon dioxide, oxygen, and nitrogen are removed. The remaining compounds, water, acetic acid, and acrylic acid, are then sent to a liquid-liquid extraction tower where ethyl acrylate is used to extract the acids. The tower bottoms is mostly water, with some dissolved solvent, and is sent to a wastewater treatment plant. The materials exiting this tower's overhead (solvent, acetic acid, and acrylic acid) are sent to a distillation tower to recover the ethyl acrylate. The solvent is then recycled back to the extraction tower. The acids are then sent to a second distillation tower for final purification. Eight heat exchangers were used throughout the process in order to implement temperature control measures on the different streams. A process flow diagram of the system can be seen in Figure 1.
The stream flowrates and compositions entering and leaving the
heat exchangers were taken from the suggested design offered by
Kangas, therefore all calculations and the optimized design of the
heat exchanger network were based on these specifications. One of the
heat exchangers, E-301, was not considered in the analysis and
optimization of the suggested design. This exchanger can be found in
the reactor section, and is used to cool the molten salt exiting the
reactor. Due to the difficulty involved with modeling the molten
salt, this exchanger was deemed to be beyond the scope of this
project. The component flowrates that were assumed for the other
seven exchangers are given in Table 1.
Exchanger Water Acetic Acid Acrylic Acid Ethyl Acrylate Total
E-302 1440 25.3 432 0.00 1900
E-303 0.00 70.7 390 129 590
E-304 2.87 0.47 0.00 440 443
E-305 0.00 0.16 13.7 0.00 13.8
E-306 0.00 6.66 3.11 2.79 12.57
E-309 3.58 0.00 0.00 133 136
E-310 0.00 0.00 6.25 0.00 6.26
All of the exchangers in Kangas' design were shell and tube heat
exchangers, with the process stream flowing through the tube side.
In all cases a utility, either cooling water or low-pressure steam,
was used was used as the heat transfer medium. A summary of the
conditions for each exchanger can be found in Tables 2 and 3.
Exchanger Type of Exchanger Temp. In (C) Temp. Out (C) Pressure In (bar) Pressure Out(bar) Flowrate (tonne/hr) Duty (GJ/hr)*
E-302 shell and tube - fixed TS 50 48 3.01 2.94 1900.0 -14.3
E-303 shell and tube - floating head 74 91 0.20 0.19 590.0 348.0
E-304 shell and tube - floating head 46 33 0.13 0.13 443.0 -181.0
E-305 shell and tube - floating head 88 92 0.17 0.17 13.8 9.05
E-306 shell and tube - fixed TS 60 43 0.08 0.06 12.6 -6.26
E-309 shell and tube - floating head 37 40 3.01 2.96 136.0 0.779
E-310 shell and tube - floating head 89 40 2.41 2.37 6.26 -0.685
* duty exchanged "to" process stream
Exchanger Utility* Temp. In(C) Temp. Out(C) Pressure In(bar) Pressure Out(bar) Flowrate(tonne/hr)
E-302 cw 30 40 5.01 2.90 89.4
E-303 lps 160 160 6.01 5.89 162
E-304 cw 30 40 5.01 4.91 1130
E-305 lps 160 160 6.01 5.89 0.0322
E-306 cw 30 40 5.01 4.53 39.0
E-309 lps 160 160 6.01 5.93 0.364
E-310 cw 30 40 5.01 4.97 4.28
* cw = cooling water at 30°C; lps = low-pressure saturated steam at 160°C
Due to the nature of the heat transfer requirements in this process; very little heat integration is possible. That is, there is only a small amount of heat that can be transferred directly between any two process streams without violating the second law of thermodynamics. By completing a temperature interval and cascade diagram analysis, the minimum utilities calculated confirms the lack of streams to match. See Appendix A for temperature interval method analysis and Appendix B for cascade diagram.
The process streams in exchangers E-302 and E-309 are fed into a shell and tube heat exchanger to maximize the small amount of benefits that can be reaped from direct integration. This results in a moderate decrease in the amount of steam and cooling water needed to reach the streams' respective temperature objectives.
The second optimization that Millennium found was to send the acrylic acid product stream to a fin-fan heat exchanger before final cooling in the product cooler E-310. A fin-fan exchanger is a low capital cost and low operating cost solution to cool or heat streams whose temperatures are much higher or much lower than the outside air temperature, respectively. As its name implies, the fin-fan exchanger works by using fans to blow outside air over the process stream tubes which are covered with metal fins to increase the heat transfer area. Cost savings result when using a fin-fan exchanger because instead of requiring costly utilities, it takes advantage of ambient conditions and uses "free" air as its heat transfer medium. Relative to complicated shell and tube exchangers, fin-fan exchangers are more reliable and easier to operate and control. The only other expense besides the cost of the fans and fan motors is a modest amount of electricity to run the motors.
For our process, we have chosen to install five electric motor-driven 5 hp fans which cost a total of $21,000. The acrylic acid product stream will flow over the fans through three tubes, with each tube making three passes over the fans. The exchanger will cool the acrylic acid from 89°C to 50°C, leaving E-310 to provide the final cooling to the desired 40°C. We estimate that the 20-year net savings in capital and operating cost from using the fin-fan exchanger are nearly $15,000. The savings come primarily from using a smaller E-310 exchanger, and from using only 20% of Kangas' suggested cooling water flow in E-310.
Direct transfer of heat between the reboilers and condensers of the process is not possible due to the temperature regime. Each of the condensers lies below the pinch temperature, and each of the reboilers lies above. Thus, we are faced with the difficulty of somehow transmitting heat between the two efficiently. Approximately 99% of the yearly operating cost of the process is due to the steam and cooling water requirements of the reboilers and condensers. There is a large economic motivation for optimization of the distillation column exchanger network.
A compression/expansion refrigeration cycle can be used to "pump" the heat between the reboiler and condenser of a column. Although large compressors are often avoided in sensible design because of their relative difficulty in handling multiple process load disturbances and their high capital cost, the much higher potential savings in operating cost led us to consider the "heat pump." The heat pump loop provides all of the heating and cooling needs for both distillation columns. Using the condensers and reboilers as sinks and sources, a cycle of compression, heat exchange, expansion, heat exchange, compression, heat exchange, etc. is established. The compressor control issue can be dealt with by installing four auxiliary heat exchangers to handle the load disturbances (such as changes in process stream flows and temperatures) and to allow approximately steady-state operation of the heat pump loop.
The heat pump loop uses a stream of mixed pentenes and pentanes as the "refrigerant." The cycle is as follows:
The cycle repeats to provide heat pumping for the reboilers and condensers. The refrigerant is critical, and was initially chosen as R-113 (Freon-113). This chlorofluorocarbon (CFC) has been banned, however, due to its degrading effects on the ozone layer. A common replacement for banned industrial refrigerants are appropriate mixtures of hydrocarbons. Although a larger flowrate of pentenes is necessary due to its lower performance characteristics, it meets the necessary specifications to be our refrigerant. Furthermore, pentene is relatively safe and environmentally friendly.
The following process flow diagram illustrates the flow of the
refrigerant stream through the equipment.
The thermodynamics of the heat pump cycle are detailed in the
following temperature-entropy (T-S) and pressure-enthalpy (P-H)
diagrams. The letter labels on these diagrams correspond to those on
the PFD.
The compression portion of the heat pump cycle involves the addition of 6.14 MW of shaft work. We spoke extensively with a representative from the world's premier industrial compressor manufacturer, Dresser-Rand. A multi-stage reciprocating compressor is implemented, per their recommendation. The large size and power requirements associated with this compressor exceed the limitations of using CAPCOST for costing the equipment. We instead use an estimate provided by Dresser-Rand, which is very conservative. Furthermore, Dresser-Rand can provide the large expansion valve and other equipment needed to implement the heat pump. We include the cost of the valve and associated equipment in the cost of purchasing the compressor. Further study and more detailed specifications would be required to get a truly accurate capital cost. We believe our estimate represents an absolute maximum.
To handle start-up and non-steady state operations of the process, auxiliary reboilers and condensers are included in the design. The heat pump will only be engaged once the process reaches pseudo-steady state after start up. Furthermore, any process condition changes can be handled by using small amounts of steam and cooling water to control column operation until the heat pump system can be adapted. In terms of our economic analysis, the inclusion of auxiliary exchangers will not affect the capital cost. The suggested design capital cost already includes two condensers and two reboilers, and thus we simply carry these costs into our design. The utilities requirements of the auxiliary exchangers will be only transient. We assume, therefore, that the operating costs associated with their inclusion is negligible.
Overall, the innovative addition of a heat pump to the process saves almost all of the cooling water and steam utilities costs. Instead, relatively cheap electricity is purchased to drive the compressor. Our economic analysis will show that, despite increased capital costs, operating costs savings result in a greatly reduced total present value cost when compared to the suggested design.
Millennium performed a detailed analysis of each exchanger in the
network using the HYSYS heat exchanger detailed rating module.
Parameters such as tube length, number of tubes, and shell diameter
were chosen such that the exchanger capital cost was minimized and
outlet temperature and pressure constraints were met. An example of
the results of this rigorous analysis for exchanger E-305, the
product purification reboiler, are shown below in Table 4.
Overall Shell Tube
tube volume 0.09048 m2 shells in series 1 OD 0.02 m
shell volume 0.7069 m2 shells in parallel 1 ID 0.016 m
heat transfer area 28.27 m2 shell diameter 0.6 m tube thickness 0.004 m
tube passes 2 shell fouling 19 C-h-m2/kJ tube length 3 m
orientation horizontal baffle type single number of tubes 150
TEMA type AEL baffle orientation horizontal pitch 0.035 m
baffle cut 20% (% area) tube layout angle triangular
baffle spacing 0.8 m tube fouling 2 C-h-m2/kJ
shell heat transfer coeff. 214.9 kJ/h-m2-C thermal cond. 45 W/m-K
shell pressure drop 0.007791 kPa wall Cp 0.473 kJ/kg-C
wall density 7801 kg/m3
tube heat transfer coeff. 7.21E+04 kJ/h-m2-C
tube pressure drop 7.735 kPa
All capital costs were estimated using CAPCOST. In order to cost
all of the heat exchangers, it was necessary to find the heat
transfer area, which is a characteristic size parameter and can be
used to estimate the cost. In order to find these areas for the
suggested design, the ratio of exchanger duty to heat transfer area
was taken from the Turton case, and a direct correlation was assumed.
Although this kind of assumption is generally not accurate,
experimentation with these parameters showed that this direct ratio
correlation yielded reasonable estimates. The validity of the
assumption was verified by modeling all of the exchangers present in
the suggested design in HYSYS simulation. The utility flowrates (for
cooling water and low pressure steam) were also determined from the
simulation of the suggested case. The price of cooling water was
taken to be $0.05/1000 gal, and the price of low pressure steam was
taken to be $4.00/1000 lb (Seider, et al.). The capital cost and
utilities cost of Kangas' suggested design are shown in Table 5.
Exchanger Capital Cost Utility Flowrate (tonne/hr) Cost($/year)
E-302 $ 149,000 cw 89.4 $ 35,800
E-303 $ 975,000 lps 162. $11,400,000
E-304 $4,000,000 cw 1130. $ 452,000
E-305 $ 41,500 lps 0.0322 $ 2,270
E-306 $ 60,000 cw 39.0 $ 15,600
E-309 $ 16,000 lps 0.364 $ 25,600
E-310 $ 21,200 cw 4.28 $ 1,710
Total Utilities Cost $ 12,000,000
NPV utilities, 20 years, 10% discount rate $103,400,000
TOTAL Capital Cost $ 5,300,000
TOTAL
COST $108,700,000
It is clear from these numbers that the bulk of the cost of the
suggested design comes from the low pressure steam that is fed to the
solvent recovery tower reboiler, E-303. By adding the heat pump
loop, this cost can be avoided. Although the capital cost of the
compressor is extremely high, it, along with the refrigerant and the
expansion valve, are one time costs, and the savings in utilities
more than makes up for the necessary increase in capital investment.
The electricity needed to run the compressor was priced at
$0.04/kWhr. With a necessary power input of 6.14 MW, the electricity
needed to power the compressor system totals $3.2 million. Although
this is no small utilities cost, it is significantly smaller than the
$12 million cost of the suggested design. All capital and operating
costs associated with the optimized design can be found in Table
6.
Equipment* Capital Cost Utility Utility Flowrate(tonne/hr) Utility Cost($/year)
E-302 $ 144,000 cw 84. $ 33,600
E-303a $ 975,000 NONE - $ -
E-303b $ 975,000 lps 162. $ -
E-304a $ 4,000,000 NONE - $ -
E-304b $ 4,000,000 cw 1130. $ -
E-305a $ 41,500 NONE - $ -
E-305b $ 41,500 lps 0.0322 $ -
E-306a $ 60,000 NONE - $ -
E-306b $ 60,000 cw 39.0 $ -
E-309 $ 11,900 NONE - $ -
E-310 $ 10,000 cw 0.856 $ 342
AC-100 $ 201,000 Electricity
Compressor, $38,000,000 4000 kW $ 3,200,000.00
Drive, and Electricity
Valve
Refrigerant $ 180.00
Total Utilities Cost $ 3,234,000
NPV utilities, 20 years, 10% discount rate $28,000,000
TOTAL Capital Cost $48,500,000
TOTAL COST $76,500,000
* Where backup exchangers are used, main
exchangers are designated "a", and auxiliary exchangers are
designated "b"
The effect of the decrease in operating cost is amplified when the
present value of the cost is calculated for a twenty-year plant life.
The comparison between the suggested design costs and those of the
optimized design can be seen in Figure 4.
The large increase in the capital cost is due to the addition of the heat pump loop and the back-up exchangers. Pricing information for the compressor and any necessary auxiliary equipment associated with it was obtained from Dresser-Rand. The total capital cost for the entire compressor system was $38 million. Despite the increased capital cost, over a twenty year plant life, implementation of the design that we have proposed will save an estimated $32 million, a 30% reduction in cost as compared to the suggested design.
