Sie sind auf Seite 1von 10

AIChE Paper Number 291195 Simulating Field Performance Data from an Ethylene Producer for a Cold Box

Douglas Decker Staff Engineer Chart Energy and Chemicals

Vishwas V Wadekar Technology Director, HTFS Research Aspen Technology

Prepared for Presentation at the 2013 Spring National Meeting San Antonio, Texas, April, 30, 2013

AIChE and EPC shall not be responsible for statements or opinions contained in papers or printed in its publications

Simulating Field Performance Data from an Ethylene Producer for a Cold Box
Douglas Decker Staff Engineer Chart Energy and Chemicals

Vishwas V Wadekar Technology Director, HTFS Research Aspen Technology

Abstract: This paper shows how the field performance data for a plate-fin heat exchanger used in hydrogen enrichment is successfully simulated. Such simulation requires that detailed geometry of the exchanger is taken into account, and rigorous simulation of the exchanger is carried out directly in the process simulation environment. In order to model the exchanger performance, a detailed incremental analysis is essential because the heat exchange processes invariably involve multicomponent boiling and condensing streams, with their consequent nonlinear temperature and heat transfer coefficient profiles. It is demonstrated that, as always, engineering judgment and process knowledge are vital to assess the relative confidence levels of different measured plant variables, so that more meaningful reconciliation of the measured data and performance predictions could be carried out. In the step by step approach each measured plant variable is assessed and then used in the simulation as either an input item or as a target predicted item. The final outcome of this step by step approach is a rigorous simulation model of the hydrogen enrichment process, which is validated with the measured field data. This type of validated model can form the basis for investigating the effect of off-design conditions on the plant operation, establishing the optimum operating conditions or for dynamic simulation.

Introduction Ethylene plants have a cold train separation unit to separate light components like hydrogen and methane from the process gas in order to extract the ethylene. Plate-fin heat exchangers and knock-out vessels are utilized in this cryogenic separation process. Figure 1 illustrates a typical ethylene process diagram for the cold section

Figure 1 Ethylene Plant Schematic for the Cold Section

Plate-fin heat exchangers used in this and other cryogenic processes are the key equipments in the cold boxes. Some of the unique features of plate-fin heat exchangers, which make them suitable for these applications, include Ability to work effectively at very low temperatures because of their aluminum based construction Multistream ability which means that the concept of process integration can be used to the fullest extent to save energy, capital and operating costs. High thermal effectiveness arising from the high heat transfer coefficients and ability to incorporate large heat transfer area, both primary as well as secondary. High thermal effectiveness means that the exchangers can be operated for duties involving very close temperature approaches.

This paper shows how the field performance data for a plate-fin heat exchanger used in hydrogen enrichment could be successfully simulated using a rigorous

multistream exchanger model, performing detailed incremental analysis in the process simulator environment. Hydrogen Enrichment Process and Field Data Figure 2 illustrates a process simulator model of the hydrogen product heat exchanger used in the hydrogen enrichment process in an ethylene plant.

Figure 2 Hydrogen Product Heat Exchanger Model Figure 3 provides an enlarged view of the process schematic shown in Figure 2. The process gas enters the heat exchanger as material stream PG-IN and is cooled to the state point PG-OUT. The cooled fluid enters vessel V-100 where the liquid and gas are separated. The gas that exits the vessel V-100 at state point 40 is higher in concentration of hydrogen than the process gas and returns to the heat exchanger as material stream H2-IN where it is heated in the heat exchanger to material state point H2-OUT. The liquid from V-100 is flashed across a valve and returns to the heat exchanger as tail gas at material state point TG-IN where the heat exchanger heats the fluid to material state point TGOUT. A nonperm material stream 53 is available if more refrigeration is needed to control the process temperatures. Similarly there is also a valve (VLV-101)

that can return part of the hydrogen stream through the tail gas stream to control the process temperatures if required.

Figure 3 Hydrogen Product Heat Exchanger Schematic In this model there are only two material streams that are independent variables: the process gas stream at state point PG-IN and the nonperm fluid at state point 53. The other independent controls are the valves VLV-100 and VLV201. Otherwise the heat exchanger geometry, vessel geometry and piping are fixed. The field data included temperature measurements at state points PG-IN, PG-OUT, H2-IN, H2-OUT, TG-IN and TG-OUT. There were thermocouple calibration records for these measurements. Therefore, there is a high degree of confidence in the temperature measurements, with known accuracy. The location of the thermocouples with respect to the heat exchanger was also known. Pressure measurements are at state points PG-IN, PG-OUT and TG-IN. The accuracy and location of the pressure measurements was not as well known but they were perceived to have significantly less impact on the analysis. Mass

flow data was provided at state points PG-IN, H2-IN, TG-IN and nonperm fluid. The exact method of flow measurement was not documented; some of the flow rates were measured and others were calculated. Stream compositions were not measured during data collection. In summary the field data is perceived to have the following order of decreasing level of confidence: temperatures, pressures, flow rates and finally stream compositions. Simulation of Plant Data Modeling field data is difficult at best and sometimes it appears near impossible. Therefore a stepwise approach is essential to make the problem tractable. The following stepwise approach, consisting of three steps was adopted in the present case. 1. Simulate the design case. 2. Simulate the field data without any adjustment to variables 3. Try to reduce difference between calculated and measured variables It should be mentioned that a commercially available process simulator with a plate fin unit operation that has the ability to rigorously model heat transfer and the process, allows this paper to concentrate on a simple method to analyze field data. Without an integrated tool many more steps and iterations would be required, along with significant amount of effort. Assuming the use of an integrated tool a detailed description of these steps is given below. Step 1 Simulate the design case. This step is probably the most important but often the most ignored one. There is often a tendency to jump into the next step of simulation of plant data without properly calibrating the model. It is always a good investment of time to know the model replicates the design condition, especially when one is puzzling over the differences between the model and the field data, in the later stages of analysis. In this particular case it was reassuring to be able to replicate the original design conditions. Step 2 Simulate the field data without any adjustment to variables. The manner in which the data for temperature, pressure, and flow rate were selected is described here with help of Table 1 and Table 2. Please note that both Table 1 and 2 contain two parts, one showing the measured data and other showing simulation results. Table 1 lists the field temperature data and Table 2 pressure and flow rate data. Note that except for the nonperm fluid temperature, all other stream temperature data were provided. The first 6 rows of data were from steady state operation, running from 0 to 100 minutes. The 60 minute time was arbitrarily selected as the steady state condition to model and hence data in this row are highlighted with a background color, in both tables.

Referring to the second part of the Tables, showing simulation results, the values that are bold (and blue) are input values; normal black items are calculated values. The differences between the calculated model values and the field data are listed in the rows labeled Diff in Table 1. Please note that such differences shown in Table 2 on the mass flow rate is because of the lower confidence level associated with flow rate measurements. Case I is an attempt to model the 60 minute case. The model is constrained to temperature PG-IN and the nonperm temperature. The nonperm temperature was not provided but was initially assumed to equal the design nonperm temperature. Another model constraint is pressure at PG-IN and TG-IN. The flow rates are constrained at PGIN, H2-IN and the nonperm flow rate. The customer provided an initial estimate of the composition at state point PG-IN and this was used in the analysis. The design composition of the nonperm stream was assumed for Case I. Table 1 Field Data Temperatures Case Time Temp PG-IN
0

min

Temp PGOUT 0 F

Temp H2-IN
0

Temp H2OUT 0 F

Temp TG-IN
0

Temp TGOUT 0 F

T Nonperm
0

Measured data 0 20 40 60 80 100 Case I Diff Case II Diff Case III Diff -139.7 -140.4 -140.8 -141.4 -138.3 -135.0 -200.5 -202.8 -152.5 -200.5 -202.9 -153.2 -200.8 -203.1 -153.5 -200.8 -203.0 -154.4 -200.1 -202.7 -151.6 -199.8 -201.9 -148.2 Results of simulation -141.4 -236.8 -236.8 -159.7 -36.0 -33.8 -5.3 -214.1 -214.2 -214.0 -213.7 -213.4 -214.0 -155.0 -155.6 -156.3 -157.6 -154.3 -150.4
-191

-244.4 -151.9 -30.7 5.7

-141.4 -200.8 -200.6 -150.3 -211.6 -147.5

-120

0.0

2.4

4.1

2.1

10.1
-120

-141.4 -201.3 -201.2 -159.7 -205.9 -154.7

-0.5

1.8

-5.3

7.8

2.9

It is interesting to note that the measured temperature at H2-IN is lower than the temperature at PG-OUT. The temperature at H2-IN should equal PG-

OUT unless energy is added to or removed from the stream. It was believed there was no physical basis for removing energy when the stream flowed from H2-IN to PG-OUT. While temperatures were perceived to have the highest level of confidence, no modeling technique would estimate this 2.2 0F temperature difference between the measured data for PG-OUT and H2-IN. Table 2 Field Data Pressures and Flow Rates Case Time Pressure Pressure Pressure PG-IN PG-OUT TG-IN psig psig psig Flow PG-IN k lb/hr Flow H2-IN k lb/hr Flow TG-IN k lb/hr Flow NONPERM k lb/hr

min

Measured data 0 20 40 60 80 100 Case I Diff Case II Diff Case III Diff 411.6 412.0 410.6 411.9 411.1 412.4 411.9 411.4 411.6 410.4 411.5 410.3 411.8 82.8 19.1 82.8 19.4 82.8 17.5 83.1 24.2 83.7 19.5 84.4 20.2 Results of simulation 411 83.1 24.2 -0.5 83.1 24.2 11.2 11.5 11.0 11.2 11.1 10.8 11.2 9.6 9.6 8.3 14.8 10.2 10.9 14.8 0 20.15 5.35 1.8 1.8 1.8 1.8 1.8 1.5 1.8

411.9

411.4 -0.1

11.2

7.15

411.9

411.4 -0.1

83.1

24.2

11.2

20.15 5.35

7.15

Step 3 Compare simulation results with field data, adjust variables with low measurement confidence and run simulation again until there is reasonable agreement with the field data. For Case I the field data suggests the temperature at PG-OUT is -200.8 0F, where as the model actually calculates a temperature of -236.8 0F, i.e. a temperature difference of 36.0 0F. Other temperature differences are also listed in the Table 1. Clearly it is desirable to determine if the model can be adjusted

to reduce the differences between the predicted model temperatures and the field data temperatures. The temperature of the nonperm gas was unknown. Case I was defaulted to the design condition of -191 0F. For Case II this nonperm fluid temperature was changed to -120 0F. Note the model now predicts the stream PG-OUT measured temperature. The other calculated temperatures also agree better with the measured field data except for the temperature at TG-OUT. The nonperm fluid composition was not measured. In Case III the nonperm composition was altered. The maximum difference in temperature between the calculated temperatures and measured temperatures is now reduced from 10.1 to 7.8 0F. Engineering judgment is required to determine when the results are satisfactory. It was noted earlier that there is an unexplained 2.2 0F temperature difference between the measurements for PG-OUT and H2-IN. If one were to continue to pursue reducing the temperature differences further, the process gas composition would be another variable that could be adjusted.

Further Remarks An interesting aspect in this modeling is one must understand the impact on constraining the model. The hydrogen product heat exchanger is at the coldest end of the cold train separation unit, so it is fairly simple to isolate the variables that influence its performance. Note that this example is really limited to two streams influencing the modeling results. If one were to model the next warmest heat exchanger in the cold train separation unit, the performance of the hydrogen product heat exchanger would also influence the warmer heat exchanger performance. The number of variables and the difficultly in resolving the differences between measured and calculated temperatures would also increase correspondingly. Another sidebar in this study was it challenged the traditional approach of assigning design margin to each piece of equipment in isolation. Traditionally the inlet temperatures are held constant and the excess heat load or area of each piece of equipment is used to assess design margin. What is apparent in this example is the inlet state point of the hydrogen and tail gas stream is dependent on the process gas outlet state point. For example, not only is the hydrogen inlet temperature affected by the process gas outlet temperature but the composition of hydrogen is also affected by the process gas outlet temperature. The performance of each piece of equipment impacts the performance of the other pieces of equipment in the process. Therefore,

assessing equipment performance in isolation could result in some unexpected process performance issues. The final outcome of this step by step approach is rigorous simulation model of the hydrogen enrichment process, which is validated with measured field data. This type of validated model can form the basis for investigating the effect of dynamic events and off-design conditions on the plant operation, as well as establishing the optimum operating conditions. Conclusions The need to use detailed incremental analysis for simulating complex multistream heat exchangers is well known. It is essential that such detailed heat exchanger simulation is carried out within the simulation of the whole process, so that the interactions with other unit operations including other heat exchangers, and the effect of process feedback loops are taken into account. Because of the availability of integrated commercial tools for detailed simulation of plate-fin heat exchanger within a process simulator, it was possible to use the simple method outlined in the paper to simulate the field performance data. Another important point that needs emphasizing is that, as always, engineering judgment and process knowledge are vital to assess the relative confidence levels of different measured plant variables, so that more meaningful reconciliation of the measured data and performance predictions could be carried out.

Das könnte Ihnen auch gefallen