Transient cooling simulation of atmospheric residue during pipeline shutdowns

Applied Thermal Engineering 106 (2016) 22–32

Contents lists available at ScienceDirect

Applied Thermal Engineering journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Transient cooling simulation of atmospheric residue during pipeline shutdowns Tommy Firmansyah ⇑, Mohammad Abdur Rakib, Abraham George, Mohamed Al Musharfy, Mabruk Issa Suleiman Takreer Research Centre (TRC), Abu Dhabi Oil Refining Company (Takreer), PO Box 3593, Abu Dhabi, United Arab Emirates

h i g h l i g h t s
CFD model solves transient heat transfer of residue during pipeline shutdowns.
Cooling curves are predicted to estimate time allowance before full solidification.
Pour point is used as a criterion of residue solidification.

a r t i c l e

i n f o

Article history: Received 17 December 2015 Revised 28 April 2016 Accepted 29 May 2016 Available online 30 May 2016 Keywords: Computational Fluid Dynamics (CFD) Transient heat transfer Transient cooling Cooling curve Straight Run Residue (SRR) Solidification time

a b s t r a c t In this paper, a Computational Fluid Dynamics (CFD) model was developed for solving transient conjugate heat transfer of pre-heated atmospheric residue from a crude distillation unit in buried pipelines following pipeline shutdowns. The emphasis of the simulations was to predict cooling curves of the residue in a pipeline during shutdowns to help predict time allowance for operations and maintenance activities to troubleshoot the problem and resume operations before full solidification. Winter conditions were chosen to estimate duration for initiation of solidification of the residue using a criterion of pour point derived from laboratory analysis. Two different modeling approaches have been presented; axisymmetric and full pipe cross-section and both have given comparable results in estimating the residue solidification times. A scenario of energy savings by reducing initial temperature of the pre-heated residue and its impacts to the solidification time has also been presented. The results from transient CFD simulations showed cooling temperature fronts across all thermal layers including estimated time taken for the residue to reach a safe temperature margin above pour point. The model can be used as a valuable flow assurance tool to avoid the risk of residue solidification in pipelines and thus negative impact on economic resources. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Wax deposition and solidification remains a major challenge to oil and petroleum industries, ranging from damage to oil reservoir formations to blockage of pipelines and process equipment [1]. There are several driving forces that induce wax deposition in pipelines, such as the difference between the bulk oil temperature and the temperature of pipe wall or the outside temperature [2] and the hydrocarbon stream properties. When wax precipitates within pipelines at and below Wax Appearance Temperature Abbreviations: CFD, Computational Fluid Dynamics; SRR, Straight Run Residue; ADR, Abu Dhabi Refinery; RR, Ruwais Refinery. ⇑ Corresponding author. E-mail address: [email protected] (T. Firmansyah). http://dx.doi.org/10.1016/j.applthermaleng.2016.05.179 1359-4311/Ó 2016 Elsevier Ltd. All rights reserved.

(WAT), wax gelation starts to form and inhibit flow by causing non-Newtonian behavior and increasing viscosities as the temperature of a waxy crude oil approaches its pour point [3]. The formation of wax gelation or solid wax column during a pipeline shutdown can completely block the pipeline and could lead to major pipeline restarting problems, if insufficient pressure is available at the pipeline inlet to break the gel and allow the waxy oil to flow [4]. Problems caused by solidification and deposition of waxes during production and transportation of crude oils cost billions of dollars yearly to petroleum industry due to increasing cost of chemical wax inhibitors, production loss, well shut-in, less utilization of capacity, flow lines choking, equipment failure, extra horsepower requirement and increased manpower attention [5].

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

23

Nomenclature cp h k r R T To t

q

specific heat (J/kg K) heat transfer coefficient (W/m2 K) thermal conductivity (W/m K) radial coordinate (m) pipeline radius (m) temperature (K) reference temperature (K) time (s) density (kg/m3)

Preventing the wax deposition to happen in the first place is considered the most cost effective way [6]. Some prevention strategies such as use of chemical inhibitors [7,8] or combined chemical inhibitor with pigging [8] are often implemented. However, when wax deposition cannot be prevented and wax gelation starts to form and inhibits flow then wax removal strategies have to be enforced to avoid risk of full pipeline blockage. Most common practice is pigging to remove wax build-up in pipelines, sometimes in combination with chemical treatment [9]. Some heat treatment to remove wax in pipelines such as inductive heating through an external coil [10] has also been used and a preheating treatment such as injecting fresh warm oil at the pipeline entry has also been practiced to resume the flow of a compressible gel-like material [11]. Many researchers have investigated the complex process involved in wax deposition and solidification using experimental [12,13] as well as numerical approaches [5,14–17]. These studies [5,12–17] developed empirical based correlations and mathematical/numerical models describing the formation of wax deposits in a pipeline carrying waxy oil, and subjected to temperature drops due to low external temperature conditions surrounding the pipeline. Most models of wax deposition are based on molecular diffusion driven mechanism [6], although shear dispersion may play a role in wax deposit removal, which would affect the rate at which wax accumulates [18]. While the wax precipitation is mainly a function of thermodynamic variables such as composition, pressure and temperature, wax deposition is also dependent on the flow hydrodynamics, heat and mass transfer, and solid-solid and surface-solid interactions [19]. The biggest problems in waxy crude oils transportation are complete blockage of the pipelines and restarting of gelled flowlines [20]. As the rheology of waxy oils is highly temperature dependent, the heat transfer system must be considered under normal uninterrupted/steady flow operations as well as during restarting flow conditions after downtime. In fact, drastic heat transfer during shutdown makes static cooling much more problematic than dynamic cooling during uninterrupted flow [20]. Pipeline shutdowns may occur for maintenance, operational or emergency reasons [11]. Under non-flowing conditions when the pipeline is surrounded by low external temperature, the temperature in the pipeline starts to drop. This temperature decline causes the crystallization of paraffinic components eventually leading to wax gelation build-up as the temperature continues to drop below pour point. When the shutdown period extends too long the wax gel formation or solid column can completely block the pipe, and hinder restarting of the pipeline. Therefore, it is clear that the temperature is the key parameter of the whole shutdown and restart processes [17]. An early study of transient cooling in an insulated oil pipeline during shutdowns was described by Szilas [21] using an analytical model based on linear-heat-source correlations. The model

l lo

viscosity (kg/m s) reference viscosity (kg/m s)

Subscripts m, n two adjacent conducting layers (between two adjacent sides of pipe wall, insulator, HDPE outer jacket, soil medium)

assumed homogeneous soil properties, constant temperature of oil all over a certain cross-section of the pipeline including the pipeline wall cross-section and also constant heat flux around the pipeline wall. Although several simplifications of model and approximation of variables that determined the cooling rates were done, this analytical method nevertheless gave general cooling trends of waxy oil in pipeline during shutdowns. Many numerical studies [22–24] have been conducted recently to describe the transient cooling behavior of waxy oil pipelines in non-flow condition. These studies allow prediction of temperature drop of the waxy oil during shutdowns and generation of temperature profiles along both radial and axial directions. Evaluation of the temperature drop during such events makes prediction of wax solidification time also possible [24]. Cheng et al. [22] employed a two-dimensional Finite Volume Method (FVM) to discretize the governing equations with an unstructured grid. Soil was represented as a finite thermal influence region and assumed to be homogeneous and isotropic. Their mathematical model, however, required an artificial heat coefficient to represent natural convection of the crude oil in the same way as considered in conduction. Furthermore, their numerical model used a stagnation point concept by dividing the pipeline into liquid and solid regions and the layer of wax deposit on the pipeline wall was assumed uniformly distributed along the entire pipeline. Lu and Wang [24] developed a two-dimensional FVM based heat transfer model covering phase changes both in watersaturated soil around the pipeline and in crude oil inside the pipeline during pipeline hindrances in winter. An axially symmetrical boundary condition was used by assuming negligible heat transfer in the axial direction of the pipeline. Fixed physical boundaries/ interfaces had to be implemented explicitly to represent different regions in the soil based on the phase state of water, i.e. frozen soil, freezing soil, and water saturated soil and also three regions representing the crude oil in pipeline, i.e. solidified oil, solidifying oil and liquid oil. Han et al. [23] adopted a different approach with a Proper Orthogonal Decomposition (POD) Galerkin reduced-order model (ROM) for solving unsteady heat conduction problems of oil pipeline. Using body-fitted coordinate and a similar heat influence region approach of Cheng et al. [22] the model resulted in more efficient and faster computational time. Assumptions such as negligible heat transfer in the axial direction of pipeline and use of equivalent heat coefficient to represent natural convection of the crude oil still needed to be made. However, just like its predecessor, the POD reduced-order model also used pre-defined layer of wax, i.e. fixed segregated liquid and solid regions. The physical phenomena surrounding the solidification of waxy oil in pipeline during shutdown are highly complex involving wax crystal formation, wax deposition near the pipeline wall, wax gelation with yield-stress plastic rheology, and phase changing with

24

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

moving boundary whilst the oil temperature dropped below its pour point or Wax Appearance Temperature (WAT) [22]. For this reason, past numerical studies [22–24] had to make several assumptions and simplifications to reduce modeling complexities such as: 1. The surrounding environment, e.g. soil was assumed to be homogeneous and isotropic. 2. Simplifying a three-dimensional unsteady heat transfer problem into a two-dimensional model by neglecting the axial temperature drop, i.e. zero heat flux in the axial direction. 3. Use of predefined segregated liquid and solid regions so as to avoid complexities when applying specific physical properties, e.g. density and viscosity to each region. The viscosity of crude oil is particularly a very important physical property, which needs to be modeled accurately. It is generally understood that waxy crude oils have a very complex rheological behavior [25]. Above the WAT they behave as simple incompressible Newtonian viscous fluids. As the temperature drops below the WAT, the viscosity starts to increase sharply and then becomes stress dependent, in relation with the presence of paraffin crystals and of the related gel-like structure of the material. So for a realistic physical description of waxy oil solidification process a numerical model should include an appropriate nonNewtonian fluid property when the oil temperature drops below the WAT/pour point. Alternatively, an empirical correlation describing a temperature-dependent viscosity model could be implemented as an approximation to the increasing viscosity trend when the oil temperature drops below pour point. Numerical studies using Computational Fluid Dynamics (CFD) for steady-state heat transfer of moving crude oils in pipelines [15,16] have adopted this approach. This paper extends the application of this technique to transient heat transfer simulations of high pourpoint atmospheric residue in a buried pipeline during pipeline shutdowns. The atmospheric crude column Straight Run Residue (SRR) from Abu Dhabi Refinery (ADR) is pumped for further processing to Ruwais Refinery (RR) via a 1200 diameter and 235 km long pipeline. Fig. 1 illustrates a schematic figure of the SRR pipeline including seven valve stations, SV-1 to SV-7 depicting the corresponding distances between the stations. SRR is normally available at 60°–80 °C in the storage tanks at the ADR station. In order to avoid SRR solidification in the pipeline, SRR is heated to a temperature of 85°–90 °C in the SRR electric heaters at ADR station to achieve a minimum arrival temperature at the RR station of above 45 °C. In any emergency situation involving long shutdown periods in the peak winter conditions the SRR in the pipeline would have to be replaced by injecting hot gas oil to flush out the SRR remaining in the pipeline and thus avoid pipeline blockage problem. The objective of the current study is to utilize transient CFD simulations to predict the SRR solidification time in the event of long shutdowns at peak winter conditions, which could lead to disastrous impacts on the economic resources. The transient CFD

simulations were performed by solving energy equation, i.e. conduction and free/natural convection in non-flow condition. The soil medium is assumed to be isotropic and homogenous and acts as a heat sink with a predefined temperature from regional meteorological data. By acting as a cooling source the soil was positioned in the outermost layer of the CFD model and thus the effect of ground atmospheric temperature was not taken into account. The model also assumed negligible heat flux in the axial direction. The whole set-up of CFD model was two dimensional, which helped reduce computational time significantly especially for the computationally intensive and time-consuming transient simulations. The emphasis of CFD simulations was on predicting cooling curves of SRR in non-flow/static conditions during pipeline shutdowns. This was accomplished by performing unsteady conjugate heat transfer calculations encompassing all conducting layers of pipe wall, insulator, HDPE jacket and soil medium as well as the convective SRR fluid. The following metrics have been used to characterize the simulation results. Threshold temperature is defined as a safe margin higher than pour point. The safety margin has been adopted to be 5 °C, as described in Eq. (1).

Threshold Temperature ¼ pour point þ 5  C

ð1Þ

SRR solidification time is the time elapsed before the SRR temperature in the pipeline at any certain location drops down to the threshold temperature. Thus, SRR solidification time can be interpreted from the cooling plot as the time where the threshold temperature crosses with the SRR cooling curve. SRR cooling curve is a curve derived from transient SRR temperatures following a pipeline shutdown. Corresponding to each station, a CFD simulation case was run independently using a specified initial local SRR temperature as recorded at each station, as shown in Fig. 2 in a typical day in winter. 2. Model governing equations The governing heat transfer equation for the SRR inside the pipeline, pipeline wall, insulator, HDPE outer jacket and the soil environment in non-flow condition can be described by the following equation:

qcp

  @T @ @T ¼0  k @t @r @r

ð2Þ

where q is density (kg/m3), cp specific heat (J/kg K), T temperature (K), t time (s), r radial coordinate (m) and k thermal conductivity (W/m K). All solid conducting layers except the soil medium were represented with the actual corresponding thicknesses and sufficient mesh resolution across each layer. Solving the energy equation directly in all the meshed solid conducting layers will ensure good modeling accuracy and avoid distortion of crude oil transport profile at the wall boundary condition [16]. The soil medium at the outermost layer of the model was treated as a cooling source, i.e.

SRR PIPELINE SV-7

Abu Dhabi

SV-6

SV-5

SV-3

SV-4

SV-2

SV-1

Ruwais 2.31

51.98

25.733

27.6

31.22

Fig. 1. SRR pipeline and valve stations.

31.50

31.39

31.87

Distance in Km

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

25

was performed using the corresponding actual recorded SRR temperature shown in Fig. 2 as its initial condition, i.e. at time = 0 before the pipeline shutdowns. This avoids having to model the entire 235 km pipeline explicitly, which would require massive computational resources. All transient CFD simulations were based on nonflow condition. 3.1. Axisymmetric model 3.1.1. Geometry Fig. 3 shows the geometry of the axisymmetric model of the pipeline including all representative thermal conducting layers of the pipe wall, the insulator, the HDPE jacket and the soil. Thickness of each layer is schematically shown in Fig. 3a. Total length of the model domain is 20 m, as depicted in Fig. 3b. As the heat flux was assumed negligible in the axial direction this pipe length was considered sufficient to represent a section of the pipeline. Total radius is 0.5 m, which is the total sum of all layers’ thicknesses.

Fig. 2. Recorded SRR temperatures at different locations.

heat sink with a specified finite thickness whilst the heat leaks out from the SRR through the thermally conducting materials, i.e. the pipe wall, the insulator, and the HDPE jacket. A ‘‘coupled” wall heat transfer boundary condition was employed at the interfaces between the inner wall of the SRR pipe and the SRR fluid and between the two adjacent sides of the solid conducting layers to allow exchange of heat transfer rates, as described by the following equation:

k

@T ¼ hm ðT n  TÞ @r

ð3Þ

where h is the heat transfer coefficient (W/m2 K) and subscripts m and n belong to two adjacent conducting layers. The natural convection of SRR in static condition occurred through gravity-induced pressure gradients, i.e. buoyancy-driven flows. Two different modeling approaches were used in solving the natural convection. The first assumed constant density through Boussinesq approximation and this was carried out using an axisymmetric CFD model. The alternative approach employed temperature-dependent density at a full pipe cross-section model. Temperature-dependent viscosity was implemented through a power-law viscosity law relationship as:

l ¼ lo

 n T To

ð4Þ

where l denotes viscosity (kg/m s), T temperature (K), To reference temperature (K), and lo reference viscosity (kg/m s). The soil medium was modeled as a solid concentric block and assumed to be isotropic and homogeneous. The CFD model also assumed negligible heat flux in the axial direction. The CFD simulations in this study were performed using a Finite Volume Method (FVM) based CFD software, ANSYS Fluent 15.0. A pressure-based solver with ‘‘Coupled” algorithm and ‘‘Body Force Weighted” scheme was used to solve the governing equation. The ‘‘Body Force Weighted” scheme works well for body forces type of applications such as natural convection or buoyancydriven flows whilst the ‘‘Coupled” algorithm provides robust and stable calculations for transient flows with large time steps. Structured hexahedral computational mesh was used in all CFD simulations to achieve good accuracy and quicker solution times. 3. CFD model Both the axisymmetric and the full pipe cross-section models described below solved the transient heat transfer in the radial direction at different stations along the pipeline, as depicted in Fig. 1. For each station/location, a separate transient CFD simulation

3.1.2. Computational mesh A well-resolved mesh is required especially close to the pipe wall to capture the heat flow and thermal gradients accurately. For this purpose, three different mesh densities, as listed in Table 1, were generated and tested to check mesh sensitivity and dependency to solution accuracy. Fig. 4 shows a typical computational mesh, which has predominantly hexahedral cells for better resolution and hence good accuracy. As depicted in Fig. 4, finer and higher resolution cells are located near to the pipe wall to capture the correct thermal gradients. The cell thickness adjacent to the pipe wall is about 0.5 mm with the medium size mesh and the mesh representing SRR fluid gradually increases in size away from the wall with a maximum cell size of approximately 3 mm near the axis. 3.2. Full pipe cross-section model 3.2.1. Geometry Fig. 5 shows a sketch of the physical domain and overall geometry of the full pipe cross-section model covering the whole crosssections of pipe, insulator, HDPE jacket and soil with the same thickness in each layer as in Fig. 3a. 3.2.2. Computational mesh Fig. 6 shows the computational mesh for the full pipe crosssection model with similar number of cells across each conducting layer as in the medium mesh of the axisymmetric model. Similar cell sizes as of the medium mesh in the axisymmetric model were also used near to the pipe wall. 3.3. CFD boundary conditions and material properties All geometric layers and their corresponding material properties are listed in Table 2. Soil was represented as a cooling source (i.e. heat sink) with a specified temperature of 13 °C based on regional meteorological data during peak winter conditions. Several SRR samples were taken from the ADR station and analyzed in laboratory to achieve representative results for property estimation. For the study described in this paper only the highest pour point SRR sample was used as input for the CFD simulations in terms of SRR density, viscosity and pour point. The change of viscosity as temperature drops was taken into account in the CFD model by parameterizing the viscosity-temperature relationship based on the laboratory results. Fig. 7a depicts this relationship using the power-law viscosity law, as described in Eq. (4).

26

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

Fig. 3a. CFD axisymmetric geometry.

Fig. 3b. Full length and radius of CFD axisymmetric model.

Table 1 Computational mesh cases. No.

Mesh case

Total no. of cells

No. of cells in fluid domain

1 2 3

Coarse Medium Fine

12,004 17,458 22,575

6000 12,000 18,000

Starting SRR temperature in each transient simulation was specified according to the recorded SRR temperature at a particular station, as shown in Fig. 2. Soil as a heat sink started and remained at constant temperature of 13 °C.

4. Simulation results Temperature-dependent density was characterized by piecewise linear relationship based on laboratory analysis as shown in Fig. 7b. Table 3 summarizes the SRR properties used in the CFD model. 3.4. CFD initial conditions All transient CFD simulations were initialized with zero SRR fluid velocity representing the condition of pipeline shutdowns.

4.1. Mesh sensitivity analysis results Using initial SRR temperature of 89 °C at ADR station, results of radial temperature profile after 24 h of pipeline shutdown from the three simulation cases with different mesh densities are shown in Fig. 8. This clearly demonstrates that increasing the mesh resolution after 17,458 computational cells, i.e. the medium mesh, had no longer change nor affect the results and hence can be considered to reach a mesh independent solution. Therefore, the medium

Fig. 4. CFD computational mesh in axisymmetric model.

27

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

Fig. 5. Sketch of physical domain and overall geometry of the full pipe cross-section model.

Fig. 6. Computational mesh in full pipe cross-section model.

mesh size was used for all transient CFD simulations in both axisymmetric and full pipe cross-section models. 4.2. Transient radial temperature profiles across pipe cross-section during shutdown Radial temperature profiles at RR station after 24, 48, 72, 96 and 120 h of pipeline shutdowns are shown in Fig. 9a from CFD results of both axisymmetric and full pipe cross-section models. While the profiles from both numerical approaches match in the conducting layers of insulator, HDPE outer jacket and soil medium, it looks different in the convection layer of the SRR fluid. The full pipe

cross-section model produces almost flat profiles, i.e. low temperature gradients in the convection SRR layer whereas the axisymmetric model shows more gradual decrease in temperature near the pipe axis and steeper temperature drops towards the pipe wall. As a result, the axisymmetric model generates lower temperatures near the wall than that of the full pipe cross-section model. However, as the time progressed the resulting temperatures near the pipe wall from both modeling approaches came close to agreement. Ultimately all radial temperature profiles from both models flatten out when the steady-state temperature, i.e. the sub-soil temperature is reached. This trend of radial temperature profiles remains the same at the other stations along the pipeline. As clearly shown in Fig. 9a, the lowest SRR temperatures are always near to the pipe wall independent of the modeling approaches. In fact, the resulting radial temperature profile from the full pipe cross-section model also showed noticeable decrease near the pipe wall. This confirms the expectation that the rate of SRR cooling near the wall is faster than those near the pipe axis. This is clearly due to varying local temperature gradients in the fluid zone. Based on this phenomenon, the SRR temperature near the wall was taken in CFD as a criterion to estimate the elapsed time before the threshold temperature above the pour point is reached, i.e. when SRR starts to solidify. During the course of transient CFD simulations, the calculated SRR temperatures near the wall were fully recorded in time to produce the SRR cooling curves. Looking more closely at the radial temperature profiles along the x and y axis of the full pipe cross-section model in Fig. 9b, it is obvious that the resulting radial temperature profiles are almost symmetrical about the pipe axis. Therefore, the axisymmetric assumption in the axisymmetric model is a good approximation to achieve modeling efficiency. The resulting temperature distributions across all geometric layers at RR station before and after 24, 48, and 72 h of pipeline shutdown are shown in Figs. 10 and 11, from the axisymmetric and full pipe cross-section models respectively. The red color

Table 2 Material and thermal properties for CFD model. Geometry layer

Material

Thickness (mm)

Density (kg/m3)

Thermal conductivity (W/m K)

Specific heat (J/kg K)

Pipe wall Insulator HDPE Jacket Soil

Steel Polyurethane foam (PUF) High density polyethylene (HDPE) Soil medium

6.4 100 7.8 233

8030 40 970 2660

50 0.025 0.45 2.2

490 1500 1600 800

28

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

Fig. 7a. Temperature-dependent viscosity plots. Fig. 9a. Comparison of radial temperature profiles at RR station at different time intervals after shutdown between axisymmetric and full pipe cross-section models.

shows the starting SRR temperature and the blue one is the subsoil temperature. As the cooling started to take effect the colors across the SRR, the pipe wall, the insulator, and the HDPE jacket layers also changed gradually until eventually the steady state temperature reached (all blues). 4.3. Buoyancy due to natural convection

Fig. 7b. Temperature-dependent density plots.

Table 3 SRR fluid and thermal properties. Viscosity (Pa s)

Pour point (°C)

Thermal conductivity (W/m K)

Specific heat (J/kg K)

0.4146(T/To)14.49

+27

0.1573

2200

Since the pipeline is oriented horizontally underground, a radial temperature distribution due to radial heat conduction also imposes a vertical temperature gradient as well. This further creates vertical density gradients for any cross-section in the pipeline thereby inducing local convection currents. However, the SRR movement due to natural convection is not significant, as shown in Fig. 12 depicting the SRR velocity magnitude contours after 24 and 48 h of the pipeline shutdowns. The maximum velocity generated by this natural convection was merely around 5 mm/s. In reality, thus, this resulted in almost flat profiles of radial temperature distribution and quite symmetrical about the pipe axis. 4.4. Transient axial temperature profiles during shutdown Variations of resulting SRR axial temperature profiles along the pipeline after 24, 48, 72, 96, 120 and 144 h of pipeline shutdowns

Fig. 8. Mesh dependency study on radial velocity profile after 24 h of pipeline shutdown.

Fig. 9b. Comparison of radial temperature profiles at RR station along X and Y axis of full pipe cross-section model.

29

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

Time = 0 hrs

Time = 48 hrs

Time = 24 hrs

Time = 72 hrs

Fig. 10. Contours of temperature distributions at RR station before and after 24, 48 and 72 h of shutdown (contours span from 13 °C (blue) to 59 °C (red)) with axisymmetric model. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Time = 0 hrs

Time = 48 hrs

Time = 24 hrs

Time = 72 hrs

Fig. 11. Contours of temperature distributions at RR station before and after 24, 48 and 72 h of shutdown (contours span from 13 °C (blue) to 59 °C (red)) with full pipe crosssection model. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

30

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

Time = 24 hrs

Time = 48 hrs

Fig. 12. Contours of SRR velocity magnitude across the pipe cross-section after 24 h and 48 h of shutdown (contours span from 0 (blue) to 0.005 m/s (red)). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

are shown in Fig. 13 from both axisymmetric and full pipe crosssection modeling approaches. The axisymmetric model produced slightly lower temperatures along the pipeline, i.e. faster cooling rates than that of the full pipe cross-section model from the start until 96 h of the shutdowns. The deviation between the two profiles became smaller as the time progressed and good agreement between the results of both models was more obvious after 96 h of shutdowns. This trend finding is similar to the resulting radial temperature profiles and is related to the impact of natural convection in the full pipe cross-section model, which tends to slow down the cooling rates, as discussed in the above section. It is also to note that the rates of temperature drops were faster at the start and middle part of the pipeline than those near the end of the pipeline. This is more visible after 24 h of shutdown as temperature gaps between the ADR and RR stations shortened and the gradients of the axial temperature profiles became smaller. 4.5. Cooling curves and SRR solidification times The resulting SRR cooling curves from the CFD transient simulations are shown in Fig. 14 for both axisymmetric and full pipe cross-section models at the beginning, the middle and the end of the pipeline, i.e. ADR, SV-04 and RR stations. The full pipe crosssection model clearly produced slower cooling curves than that of the axisymmetric model from the start until about 100 h of the pipeline shutdown. This comes as a direct consequence of the low temperature gradients in the convection layer of the full pipe cross-sectional model, as previously mentioned in the discussion of

Fig. 13. Variation of axial temperature profiles along pipeline during shutdown.

the radial and axial temperature profiles. As shown in Fig. 9a, the near-wall SRR temperatures from both models were getting closed after about 100 h of the pipeline shutdown and consequently the respective cooling curves were also closely matching. In general, the results from both models showed that the cooling rates were steep at the beginning of the shutdown and then gradually slowed down until the steady state temperature reached. These predicted cooling curves are then used to estimate duration for initiation of SRR solidification using a criterion of threshold temperature. The threshold temperature is defined as a safe margin higher than pour point, which is the temperature at which SRR fluid becomes semi solid and loses its flow characteristics. This pour point was determined from laboratory analysis and the safety margin for the threshold temperature has been adopted to be 5 °C above pour point. SRR solidification time is thus defined as the time elapsed before the SRR temperature in the pipeline drops down to the threshold temperature. Therefore, SRR solidification time can be interpreted from the cooling plot as the time where the threshold temperature crosses with the SRR cooling curve. Table 4 lists the estimated SRR solidification times at ADR, SV-04 and RR stations from both axisymmetric and full pipe cross-section models. Both models showed good agreement in the predicted SRR solidification times as their respective cooling curves nearly matched after 100 h of the pipeline shutdown. The shortest SRR solidification time occurred at the RR station since its initial SRR temperature was the lowest along the pipeline before the shutdown.

Fig. 14. Predicted cooling curves from axisymmetric and full pipe cross-section models at ADR, SV-04 and RR stations.

31

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32 Table 4 Summary of estimated SRR solidification times. Station

Distance from ADR (km)

SRR solidification time (h) Axisymmetry

Full pipe

ADR SV-4 RR

0 94 203

204 168 127

200 167 128

4.6. Effects of decreasing initial axial SRR temperatures To explore opportunities of energy savings in the pre-heating SRR treatment, three CFD simulations were conducted for a scenario of reducing initial SRR temperatures at the ADR station, which will result in lowering SRR temperatures at the RR station. Three initial SRR temperatures (in case of pipeline shutdown) of 55 °C, 50 °C and 45 °C were considered at the RR station. For illustrative purpose, only the axisymmetric model has been presented and the results from these simulations with reduced starting SRR temperatures at the RR station are shown in Fig. 15 and Table 5. As expected, decreasing the initial inlet axial SRR temperatures led to shorter SRR solidification time. Since the rates of cooling are very high in the initial stage of the pipeline shutdown, i.e. steep decrease of SRR temperatures immediately after the shutdown, decreasing the initial axial SRR temperatures too quickly will significantly shorten the SRR solidification time. For example, decreasing temperature by 23% from 59 °C to 45 °C shortens the SRR solidification time by 43%. So, careful benefit and risk analyses should first be done before reducing the initial axial SRR temperatures. Results from the numerical study in this paper can be utilized to prepare and anticipate with a predicted response time required to avoid SRR solidification and thus pipeline blockage. 5. Conclusions In this paper, a CFD model was utilized to calculate unsteady heat transfer from a high pour point atmospheric residue in a buried pipeline during shutdowns in peak winter conditions. The model was developed with a Finite Volume Method (FVM) in ANSYS Fluent software and structured hexahedral grid for better accuracy and faster solution convergence. Temperature drops in the atmospheric residue during pipeline shutdowns in winter conditions were investigated and estimations of solidification times of the residue were obtained through pour point criterion derived from laboratory analysis. The study employed two different modeling approaches, the axisymmetric and the full pipe crosssection, and both generate comparable results in estimating the residue solidification times. Results have also shown that crosssectional convection currents do not significantly disturb the axisymmetricity of the temperature distribution. Hence, a computationally less demanding axisymmetric modeling approach could be reliably applied for predictive studies. The paper also discussed the impacts of reducing initial axial SRR temperatures to the solidification time of SRR during the pipeline shutdowns. The model can be used as a flow assurance tool to avoid the risk of high pour point residue solidification in pipelines and thus negative impact on economic resources during planned shutdowns. Acknowledgements The authors gratefully acknowledge the support of Takreer Research Centre (TRC) and Abu Dhabi Oil Refining Company (Takreer) management. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Fig. 15. Predicted SRR cooling curves at RR station with reduced initial SRR temperatures.

Table 5 Predicted SRR solidification times with reduced initial SRR temperatures at RR station. Initial SRR temperature (°C)

SRR solidification time (h)

59 55 50 45

127 114 94 72

References [1] H.O. Bidmus, A.K. Mehrotra, Heat transfer analogy for wax deposition from paraffinic mixtures, Ind. Eng. Chem. Res. 43 (2004) 791–803. [2] C.H. Wu, K.S. Wang, J. Shuler, Y. Tang, Measurement of wax deposition in paraffin solutions, AIChE J. 48 (2002) 2107–2110. [3] K.S. Pedersen, H.P. Ronningsen, Influence of wax inhibitors in wax appearance temperature, pour point, and viscosity of waxy crude oils, Energy Fuels 17 (2003) 321–328. [4] T.S. Golczynski, E.C. Kempton, Understanding wax problems leads to deepwater flow assurance solutions, World Oil (2006) D7–D10. [5] A. Seyfaee, M. Lashkarbolooki, F. Esmaeilzadeh, D. Mowla, Investigation of the effect of insulation on wax deposition in an iranian crude oil pipeline with OLGA simulator, J. Dispersion Sci. Technol. 33 (2012) 1218–1224. [6] A. Aiyejina, D.P. Chakrabarti, A. Pilgrim, M.K.S. Sastry, Wax formation in oil pipelines: a critical review, Int. J. Multiph. Flow 37 (2011) 671–694. [7] O.O. Bello, S.O. Fasesan, C. Teodoriu, K.M. Reinicke, An evaluation of the performance of selected wax inhibitors on paraffin deposition of nigerian crude oils, Pet. Sci. Technol. 24 (2006) 195–206. [8] D.W. Jennings, J. Breitigam, Paraffin inhibitor formulations for different application environments, Energy Fuels 24 (2009) 2337–2349. [9] H.P. Ronningsen, Production of waxy oils on the Norwegian continental shelf: experiences, challenges, and practices, Energy Fuels 26 (2012) 4124–4136. [10] R.C. Sarmento, G.A.S. Ribbe, L.F.A. Azevado, Wax blockage removal by inductive heating of subsea pipelines, Heat Transfer Eng. 25 (2004) 2–12. [11] C. Chang, Q.D. Nguyen, H.P. Ronningsen, Isothermal start-up of pipeline transporting waxy crude oil, J. Non-Newtonian Fluid Mech. 87 (1999) 127– 154. [12] Z. Guozhong, L. Gang, Study on the wax deposition of waxy crude in pipelines and its application, J. Petrol. Sci. Eng. 70 (2010) 1–9. [13] W. Wang, Q. Huang, Prediction for wax deposition in oil pipelines validated by field pigging, J. Energy Inst. 87 (2014) 196–207. [14] S. Haj-Shafiei, D. Serafini, A.K. Mehrotra, A steady-state heat-transfer model for solids deposition from waxy mixtures in a pipeline, Fuel 137 (2014) 346–359. [15] H. Zhu, X. Yang, J. Li, N. Li, Simulation analysis of thermal influential factors on crude oil temperature when double pipelines are laid in one ditch, Adv. Eng. Softw. 65 (2013) 23–31. [16] W. Rukthong, W. Weearapakkaroon, U. Wongsiriwan, P. Piumsomboon, B. Chalermsinsuwan, Integration of computational fluid dynamics simulation and statistical factorial experimental design of thick-wall crude oil pipeline with heat loss, Adv. Eng. Softw. 86 (2015) 49–54. [17] G. Vinay, A. Wachs, J.-F. Agassant, Numerical simulation of non-isothermal viscoplastic waxy crude oil flows, J. Non-Newtonian Fluid Mech. 128 (2005) 144–162. [18] L.F.A. Azevedo, A.M. Teixeira, A critical review of the modeling of wax deposition mechanisms, Pet. Sci. Technol. 21 (2003) 393–408.

32

T. Firmansyah et al. / Applied Thermal Engineering 106 (2016) 22–32

[19] A. Hammami, J. Ratulowski, J.A.P. Coutinho, Cloud points: can we measure or model them?, Pet Sci. Technol. 21 (3–4) (2003) 345–358. [20] J.A. Ajienka, C.U. Ikoku, Criteria for the design of waxy crude oil pipelines: maximum pump (horsepower) pressure requirement, J. Petrol. Sci. Eng. 13 (1995) 87–94. [21] A.P. Szilas, Production and Transport of Oil and Gas, second ed., 1986, pp. 240– 245. [22] X. Cheng, Y. Bo, Z. Zhengwei, Z. Jinjun, W. Jinjia, S. Shuyu, Numerical simulation of a buried hot crude oil pipeline during shutdown, Petrol. Sci. 7 (2010) 73–82.

[23] D. Han, B. Yu, Y. Wang, Y. Zhao, G. Yu, Fast thermal simulation of a heated crude oil pipeline with a BFC-Based POD reduced-order model, Appl. Therm. Eng. 88 (2015) 217–229. [24] T. Lu, K.S. Wang, Numerical analysis of the heat transfer associated with freezing/solidifying phase changes for a pipeline filled with crude oil in soil saturated with water during pipeline shutdown in winter, J. Petrol. Sci. Eng. 62 (2008) 52–58. [25] R. Glowinski, J. Xu, Special volume: numerical methods for non-Newtonian fluids, in: Handbook of Numerical Analysis XVI, 2011, p. 571.