Air Separation Units (ASUs) Simulation Using Aspen Hysys® at Oxinor I of Air Liquid Chile S.A Plant

Abstract The method used to extract copper from its ores depends on the nature of the ore. The main process currently to separate copper from sulphide ores is the smelting process. The concentrated ore is heated strongly with silicon dioxide (silica), calcium carbonate and oxygen enriched air in a furnace or series of furnaces which is carried out using the injection of the air for oxidation the Fe and Si present in the raw material. Oxygen can be produced using several different methods. One of these methods is Air separation process, which separates atmospheric air into its primary components, typically nitrogen and oxygen, and sometimes also argon and other rare inert gases by cryogenic distillation. In this paper, simulation of air separation units (ASUs) was studied using Aspen Hysys®. The obtained simulation and model was validated with the operational data from the Oxinor I of Air Liquide S.A Plant. The ASU was divided into subsystems to perform the simulations. Each subsystem was validated separately and later on integrated into a single simulation. An absolute error of 1% and 1.5% was achieved between the simulated and observed the process variables(s). This indicated that Aspen Hysys® has the thermodynamic packages and required tools to perform simulations in cryogenic processes at industrial scale.


INTRODUCTION
The main process to separate copper sulphides from ferrous metals is the copper smelting process, which is carried out by injecting pure oxygen into the minerals in order to oxidize the Fe and Si present in the raw material. Only in Chile between 2010 and 2016, 14531.8 kMT of ore were processed at the smelter plants 1 . Taking into account that each ton of material requires 42 m 3 of Oxygen, it can be seen that the need for this supply is of high value for foundry companies.
Due to the importance of the oxygen supply in copper sulphide treatment plants, an emphasis is needed on the production of this raw material. Obtaining O 2 is performed in several ways, some of them are absorption of gases using zeolite or other synthetics materials ,chemicals reactions and the cryogenic process, the main one being a unit called ASU (Air Separation Unit) because of its capability to produce large volumes of high purity products, it uses cryogenic processes to obtain pure oxygen in liquid or in its gaseous form 2 There has been investigation to optimize the processes of an ASU with the purpose of determining its use in power generation systems 3 . The complexity of the cryogenic processes, due to its low temperatures, makes the simulation process diffi cult, which is why simulation software is used to help with this task. The software that has been used and tested by different researchers is Aspen Hysys ®4 . Other software used in this area is gProm 5 .
Different approaches have been taken to simulation of oxygen plants and their energy effi ciency. Studies have been carried out emphasizing the separation process by modifying variables or stages of the process 6 . Others have performed simulations in complete plants, optimizing electricity consumption and Oxygen production 7 .
Many researchers have used the Aspen Hysys tool to simulate power generation plants through integrated gas cycles, within which ASU systems are an essential part of the process 8 . Up to now the published work focused on qualitative changes in the gas processes for energy production 9 . Other investigations have been carried out simulations in specifi c equipment such as cryogenic distillation columns or pumps of subcooled fl uids 10 some researchers have used Aspen Hysys to simulate cryogenic CO 2 removal plants 11 .
In this paper, simulation of air separation units (ASUs) was studied using Aspen Hysys ® . The obtained simulation and model was validated with the operational data from the Oxinor I of Air Liquide S.A Plant. The aim of this work is to provide Airliquide Chile S.A. a validated tool to simulate the operation variables for the purpose of process optimization.

Air separation process
The Cryogenic Air separation process involves several main and a few secondary stages. First, the air must be fi ltered so that impurities such as dust or other contaminants are eliminated. Once the air is fi ltered the process begins with the gas compression. For this, the fl uid must pass through a three-stage turbocharger. Tube & Shell heat exchangers are then used to make the compression adiabatic.
After compression the washing stage begins, which is to clean the air from contaminants that may exist, such as carbon dioxide and water vapor this is done by cooling the compress air with cool water in a washing tower then the humid air pass through a fi xed bed column that adsorb the CO 2 and water vapour. After cleaning the gases, the stream is separated into two streams, one of them, which we call MP (medium pressure), goes directly to the next stage called cold box, the rest of the fl uid goes through a second compression, which increases the gas pressure using a two-stage turbocharger and a tube Polish Journal of Chemical Technology, 22, 1, 10-17, 10.2478/pjct-2020-0003 & shell heat exchanger. The resulting stream, called HP (High Pressure), enters the cold box stage.
In the cold box stage we have the heart of all ASU processes which consist of several items of equipment such as heat exchangers, vaporizers, pumps and the distillation columns, both HP and MP. Figure 1 shows a block diagram exemplifying the process of an ASU. There are two distillation columns one working at high pressure(HP) and one working at medium pressure(MP), in the HP column the air is separated into a stream of pure nitrogen at the top and at the bottom an impure oxygen stream, the MP column has the purpose of purify the O 2 stream.

Software
As software, the Aspen Hysys ® Software acquired by the Universidad Católica del Norte, Antofagasta Chile, as shown in the Figure 2 , was used to carry out the simu-lations. Aspen Hysys ® developed by Aspen Technology has a library of more than ten thousand components added to a library of thermodynamic equations that includes more than ten equations of state each with their corresponding models and parameters. Specializes in steady-state analysis. The validations were performed using the data obtained from the Oxinor I operation from Air Liquide S.A. As a thermodynamic package, the Equilibrium model from Peng-Robinson modifi ed by Stryjek and Vera 12 was used.

Methodology
As a methodology, it was decided to address the problem of the simulation of an ASU plant in the form of its internal operations by dividing the plant into the following subsystems. This decision was made because the processes and operations that involve an ASU are so complex that to try a unique simulation where it involves the whole process would be chaotic and could not be appreciated in detail the errors and deviations in

Water cooling plant
As mentioned above, the tube and shell heat exchangers use water to cool the various gases, this is why the Oxinor I plant has two water circulation systems called Closed and Open Circuit. Both circuits are connected by plate heat exchangers.
The open circuit consists of a cooling tower, a pump and the plate exchanger. The cooling tower is a device that uses atmospheric air to cool the water by direct contact and uses a fan system to circulate the air through the water, which is cooled by saturation of the air. The purpose of this system is to cool the water that can be used in the wash tower.
The closed circuit is composed of a pump, a plate heat exchanger and an accumulator. This circuit is intended to provide cooling water to the heat exchangers during the compression stages. Each circuit has two A and B connections, which are two different heat exchangers.

Validation methodology
For the verifi cation of the compression stages, it was decided to adjust the adiabatic effi ciencies of each step and the heat transfer coeffi cients of the heat exchangers. As input variables, the value of temperature, pressure and fl ow sensors from the plant were taken into the input of each compression and cooling step. The temperatures and pressures were compared to the output of each equipment.
For the washing step, it was decided that the best way to adjust the simulation to the plant data would be to calculate the pressure losses in both the cooling column and the adsorber, in addition the pump design data was used to obtain the characteristics curves of the pumps. The pressures obtained with the plant data were compared in the cooling columns, adsorber and centrifugal pumps.
For the validation of the turbine compressor system, the adiabatic effi ciencies of both the compressor and the turbine and the heat transfer coeffi cient of the heat exchanger were used, the output values were used to compare the simulation results with the plant data.
Based on the information that was obtained from the cold box, it was decided to maintain certain temperatures, that are constant in the process, and to modify the infl ow and production fl ows of the distillation columns in order to be able to compare the quality of gaseous oxygen upon exit.
For the validation of the water cooling plant it was decided to determine the outlet temperatures of the plate exchangers depending on the input temperatures of both closed and open circuits. For the cooling tower of air, the temperature at exit was found to be a function of the one at entrance.

Compression results
From the results of the compression the temperatures obtained by the simulator and the temperatures of the plant sensors were analyzed (T1: input to second compression stage and T2: outlet of compression stage). The fl owsheet used to simulate this stage is shown on the variables of process. If they are simulated separately you can ensure that each stage works on its own and then simulate the entire process in one step: -Compression; -Washed; -Cold box; and -Water cooling plant. These divisions were validated independently with operational data. Once the parts were simulated and validated we proceeded to join the stages to form the process. The plant was validated with the same data.

Compression
The compression consisted of two separate phases, a primary compression performed by a three-stage turbo compressor and a secondary compression performed by a two-stage turbo compressor. The fi rst compression occurs at the introduction of the atmospheric air and the second compression after adsorption stage.

Primary compression
This fi rst compression consists of a three-stage turbo compressor already mentioned, after each stage there are tube and shell heat exchangers that cool the air coming out of the compression sections in counter fl ow with water. The purpose of this stage is to raise the atmospheric air pressure to the working pressures of the cryogenic distillation zones. There is a fi lter at the opening of the compressor to prevent the ingress of particulate material.

Secondary compression
The second compression uses two compression stages and a shell and tube heat exchanger to compress and cool a portion of the air from the washing step in order to achieve the pressure required to operate the distillation columns.

Washing step
This stage consists of two main parts, which are gas cooling and an adsorption of impurities. The removal of water and CO 2 present as impurity is carried out in adsorption columns. The gases are cooled by backwashing in direct contact with cold water.

Cold Box
The process of the cold box has the purpose of producing products desired by the needs of the market. This stage consists of plate heat exchangers, HP and LP distillation towers, auxiliary vaporizers, and a booster / turbine system.

Booster /turbine system
The compressor/turbine is equipment that provides the cooling required for the separation of air to occur. The compressor compresses the air and later in the turbine, the air is expanded causing it to cool and thus resulting in the normal operating conditions necessary to compensate for the cold losses of the system and to provide the cooling power for the generation of liquid products. The equipment has a tube and shell heat exchanger.  Figure 4. A regression index of 0.9979 was obtained. As it can be seen in the Figure  4 the data simulated a high concordance with the data from the plant sensors.

Washing stage results
From the washing step, the outlet pressures of the absorption and wash towers were taken and compared with the data from the plant sensors. The simulation fl owsheet used to run the programs is show on Figure 5. The results are shown in Figure 6. A regression index of 0.9648 was obtained (P1). If we look close in the Figure  6 it can be seen that the data obtain from the simulator and the plant data has a high similarity however it can still be improve changing some parameters. As Aspen

Cold box
In the cold box stage, shown on Figure 7, the results of the O 2 fl ow (F1) were obtained at the exit of this stage and de temperature inlet to the plate exchange. The plant data were used to validate the obtained data. The results were plotted Figure 8 (F1 and T3). Regression indices 0.9475 and 0.8782 respectively were obtained. Figure 8 shows a lower regression index, but taking into account the complexity and quantity of equipment and fl ows within the system it can be said that the index is suffi ciently valid to consider this stage validated with the simulation data in contrast to the values Plant

Water Cooling plant
For the water plant, we took the data from two circuits plant A and B. The output temperatures of closed circuits A, closed B, open A and open B were simulated (T4, T5, T6 and T7). The simulation fl ow sheet is shown on Figure 9. The results are plotted in Figure 10. It can be observed that circuit A has a higher regression index than circuit B. This can be due to many reasons, such as a discrepancy of data in the sensors of both circuits, the difference in the maintenance of both circuits, and the use that is given to each system.

Error analysis
In order to establish the errors in the different validations, a variance analysis was performed on the absolute errors of the studied and validated variables, fi nding with a certainty of 95% that the error is between 1% and 1.5% comparing the value delivered by the simulator and the value of the sensor. The error distribution chart is shown on the Figure 11.

DISCUSSIONS
Observing the results of the simulated stages of the process, we can see that the regression coeffi cients are mostly between 0.9 and 0.95, which allows us to concisely state that the values used in the simulation are in line with the reality of the plant allowing achievement of outputs near the process values.
Certain stages such as the cold box, specifi cally the fl ow of O 2 must be looked at more extensively since its regression index is less than 0.9. This may be because the fl ow of O 2 depends on some variables that were assumed constant and or were not taken into account. To overcome this error and to reach more acceptable regression values it is necessary to perform more simulations and combinations of variables.
From the error distribution and doing a hypothesis test on the errors it can be stated that with a 95% certainty the absolute error of the data is between a 1 and 1.5% of error comparing the data from the software and the plant values, this is comparable to other studies 9 where absolute errors less than 5% were achieved using Aspen Hysys ® software in natural gas plants.
As in other investigations 4 , it was possible to simulate the cryogenic distillation stage, fi nding in our simulation a linear regression adjustment above 0.85. We also simulated a compression system fi nding a regression adjustment of 0.99. In other investigations we found absolute errors of between 1.5% and 1.3% 8 Figure 6. Linear regression of washing stage

CONCLUSION
Using operational data, a simulator was made for the Oxinor plant with the Aspen Hysys ® software tool.
Validating the simulator with historical data of the plant shows that the errors between the process variables and the values delivered by it have values that fl uctuates between 1% and 1.5%.
The use of this software allowed consideration of the main stages of the process, which allowed evaluation and validation of parts each system, and the entire productive process, which consist of the following stages: -Compression; -Washed; -Second compression; -Cold box; -Compressor/turbine system; and -Water cooling plant. Finally, the simulated tool was validated with the data of the Oxinor I plant, which allows it to be used as a base tool to study any operational or equipment changes in this production process.
The simulated plant might also help with future works on the Air Liquide Chile S.A Oxinor I plant using the software Aspen Hysys® without wasting time and efforts on validating the software on the plant. cially to Mr. Andrés Roque for his contributions to the project research.   Table 3. LOX production design data