Chris Boljkovac
Distillation BW 1920x900

Product Lift in Vacuum Columns

|

Background

A common process configuration in Refinery crude distillation units is an atmospheric distillation unit (ADU) with a downstream vacuum distillation unit (VDU). In the VDU, vacuum distillates are recovered from the long residue (LR) for use as feed to a conversion unit, such as a hydrocracker (HCU). Due to the good separation in a VDU column, it’s not unusual that the vacuum gas oil (VGO) properties are similar to the atmospheric gas oil (AGO) produced in the ADU. The short residue (SR) stream from the bottom of the vacuum column can be routed to coking or visbreaking process units.

A key operating parameter for an ADU is the cutpoint between LR and the AGO. The front-end (i.e. the lightest) components of the LR stream have a significant lifting effect in the VDU flash zone and, therefore, on the total VGO recovery.

Figure 1: Simplified Refinery Crude Unit (ADU and VDU)

Lift Effect

With the shift of gas oil from the ADU to the VDU, the partial pressure of the heavy-boiling components in the LR stream is reduced due to the higher amount of gas oil and lighter components. This reduction in partial pressure facilitates the vaporization of the heavy components (i.e. “lifts” the heavy components), particularly the heavy vacuum distillate components. The same effect is achieved when stripping steam is injected at the bottom of an ADU or VDU column.

“The reduction in partial pressure of heavy-boiling components due to a higher quantity of lighter-boiling components.”

For a constant flash zone temperature and pressure, a 1.0% weight shift on crude (woc) from the ADU to the VDU will result in an approximate 0.2% woc higher vacuum distillate yield at the expense of SR. Therefore, a gas oil shift from the ADU to the VDU improves the vacuum distillate recovery.

It should be noted that shifting ADU gas oil has several drawbacks that must be considered when optimizing the ADU-VDU cutpoint. These include,

  1. The increase in VDU column load can be problematic, especially if any column sections are already operating near their loading factor limits. This can result in poor separation and increased entrainment of SR in the HVGO and wash oil (WO).
  2. The latent heat of the gas oil shifted from the ADU to the VDU is recovered as lower-level heat (60-100°C/140-212°F in the VDU-TCR instead of 200-290°C/392-554°F in the ADU-LCR). As a result, the ADU furnace inlet temperature will decrease. This can be problematic if the furnace is limited, although the resulting lower furnace outlet temperature will compensate.

Recommendation

As a general guideline, a VGO yield of 4-8% weight on LR feed typically gives a good balance between the lifting effect and the drawbacks listed above. This VGO yield normally corresponds to an ADU furnace outlet temperature (FOT) of approximately 360°C/680°F.

To continuously optimize the cutpoint between LR and AGO (in addition to other crude unit operating parameters) given changing economics, equipment constraints, and product quality requirements, the benefits of a real-time optimization (RTO) system should be evaluated.

PipeStraight BW 1920x900

Skin Temperature Measurement Model

|

Background

Sometimes, it’s necessary to measure a process temperature at a location without a thermowell. One option is to install a skin thermocouple mounted under the insulation on the outer surface of the pipe or equipment. As expected, this will lead to a less-accurate measurement. To improve the accuracy, a measurement model can be developed that relates the fluid temperature, skin temperature, and heat transfer in the system.

A common approach to estimate the fluid temperature from the skin temperature is,

T_f=T_s+C_t(T_s-T_a)     (1)

where,

T_f is the fluid temperature estimate
T_s is the measured skin temperature
T_a is the ambient temperature
C_t is the temperature correction factor

A correction factor of C_t=0.06 is often used but it would be expected to vary with the fluid properties and flowrate, pipe and insulation thickness and materials, ambient environment, etc.

Heat Transfer Model

A simple heat transfer model can be developed to determine the factors that affect the temperature correction factor C_t. See Figure 1 for an example of fluid flowing through an insulated pipe.

Figure 1: Pipe Skin Temperature Measurement

Heat transfer from the fluid through the pipe wall can be expressed as,

\tilde{Q}_{s}=U_{o,s}(T_f-T_s)     (2)

where,

\tilde{Q}_{s} is the heat flux through the pipe wall
U_{o,s} is the overall fluid-pipe heat transfer coefficient

Heat transfer from the outside pipe wall through the insulation can be expressed as,

\tilde{Q}_{a}=U_{o,a}(T_s-T_a)     (3)

where,

\tilde{Q}_{a} is the heat flux through the pipe wall to the ambient atmosphere
U_{o,a} is the overall pipe-ambient heat transfer coefficient

Note: U_{o,s} and U_{o,a} are based on outside pipe area

At steady-state, these heat fluxes are equal, and therefore,

U_{o,s}(T_f-T_s)=U_{o,a}(T_s-T_a)     (4)

and, re-arranging for T_f,

T_f=\dfrac{U_{o,a}}{U_{o,s}}(T_s-T_a)     (5)

Therefore, by comparing equations (1) and (5), the temperature correction factor C_t is,

C_t=\dfrac{U_{o,a}}{U_{o,s}}      (6)

the ratio of the overall heat transfer coefficients.

The overall heat transfer coefficients for flat surfaces (and, large diameter pipes),

U_{o,a}= \dfrac{x_a}{k_a}+\dfrac{1}{h_a}     (7)

U_{o,s}=\dfrac{1}{h_i}+\dfrac{x_w}{k_w}     (8)

where,

x_a is the thickness of the insulation
x_w is the thickness of the pipe wall
k_a is the thermal conductivity of the insulation
k_i is the thermal conductivity of the pipe wall
h_i is the fluid-pipe convective heat transfer coefficient
h_a is the insulation-ambient convective heat transfer coefficient

and, for circular cross sections (as in our pipe example),

U_{o,a}= \dfrac{D_o\ln(D_o/D_a)}{2k_a}+\dfrac{(D_o/D_a)}{h_a}     (9)

U_{o,s}=\dfrac{(D_o/D_i)}{h_i}+\dfrac{D_o\ln(D_o/D_i)}{2k_w}     (10)

where,

D_i is the inside pipe diameter
D_o is the outside pipe diameter
D_a is the outside insulation diameter

Therefore, from equations (9) and (10), the temperature correction factor C_t can be expected to vary with,

  1. pipe wall and insulation thickness,
  2. pipe wall and insulation thermal conductivity, and,
  3. fluid-pipe and insulation-ambient convective heat transfer coefficients.

For item (3) above, the convective heat transfer coefficients are strongly affected by fluid velocity but are also functions of fluid properties (e.g. density, heat capacity, viscosity, and thermal conductivity).

Recommendation

As shown above, the temperature correction factor C_t is affected by many system-specific factors. Using the nominal value C_t=0.06 in equation (1) is not expected to result in accurate fluid temperature estimates and is not recommended. When using skin temperatures, the heat transfer models should first be considered.

HeatExchanger BW 1920x900

Thermosiphon Reboiler Modeling

|

Background

Distillation column reboilers are typically one of the following types,

  1. kettle,
  2. fired,
  3. forced-circulation, or,
  4. thermosiphon.

In the operation of thermosiphon reboilers, flow through the reboiler is achieved naturally via the density difference between,

  • the higher-density single-phase (liquid) feed from the column bottoms, and,
  • the lower-density two-phase (vapor-liquid) reboiler outlet.

In the case of a horizontal thermosiphon, the liquid from the column passes in crossflow over the tube bundle with the heating fluid passing through the tubes. The resulting vapor-liquid mixture then flows back to the column. We’ll consider modeling a horizontal thermosiphon reboiler but the general principles are the same for vertical reboilers. A flow diagram with a typical set of temperature, pressure, and flow measurements for a thermosiphon reboiler is shown in Figure 1.

Figure 1: Horizontal Thermosiphon Reboiler

Process Modeling

The circulation rate is rarely measured for thermosiphon reboilers but specifying a constant flowrate (e.g. the design circulation rate) isn’t appropriate because the circulation rate varies with reboiler heat duty in the operation of the column. Therefore, in most cases, it’s preferable to specify a constant liquid fraction (\Phi_L) for the reboiler outlet stream.

When the return temperature T_{Co} is measured it may be possible to calculate the vapor fraction \Phi_L but the success of this approach depends on the sensitivity of the return temperature to the liquid fraction. Figure 2 illustrates the sensitivity of equilibrium temperature to liquid fraction for narrow- and wide-boiling hydrocarbon mixtures at 103 and 115 KPa-a pressure. As can be seen, the wide-boiling mixture has a well-defined liquid fraction at 110 C that’s much less sensitive to pressure changes compared to the narrow-boiling mixture.

Figure 2: Wide- and Narrow-Boiling Mixtures

Recommendation

When the bottoms stream is narrow-boiling (i.e. relatively pure) the outlet liquid fraction should be fixed at a nominal value. When the bottoms stream is wide-boiling (i.e. relatively impure) calculation of the outlet liquid fraction should be investigated.

ProcessEquipment BW 1920x900

Operational Decision Support with Online Models

|

Abstract

A review of the wide range of AVEVA ROMeo® projects at Shell’s Scotford site that moved offline production and safety systems online. Projects include,

  • fired heater outlet vaporization,
  • column relief load,
  • column loading,
  • preheat hydraulics,
  • heater and exchanger cleaning, and,
  • advanced process control (APC) coordination.

The motivation, implementation details, benefits achieved, and future work is discussed.

Presentation, AVEVA World Conference – North America, November 12, 2019

Flowmeter BW 1920x900

Flow Compensation of an Orifice Meter

|

Background

The calculations that underlie flow measurements typically assume that the fluid being measured is at the same conditions as the fluid used in the meter’s design. Flow meter compensation is the process of adjusting a measured flow for changes in flowing conditions (e.g. fluid temperature, pressure, and composition) to improve accuracy.

Flow meter compensation factors vary with the meter service, flow units, and meter type. As an example, we’ll consider the following case,

  • orifice meter (see Figure 1),
  • vapor service, and,
  • configured to indicate mass flow.

Figure 1: Orifice Meter

Flow Calculation

The mass flowrate of a vapor through an orifice meter can be described by the following,

{\dot{m}}={N}{C_d}{E_v}{Y}{d^2}\sqrt{\rho}\sqrt{\Delta{P}}     (1)

where,

\dot{m} is the vapor mass flowrate
N is a units conversion factor
C_d is the discharge coefficient
E_v is the velocity approach factor
Y is the vapor expansion factor
d is the orifice bore size
\rho is the vapor density
\Delta{P} is the pressure drop across the orifice, {P_1}-{P_2}

Orifice meters are designed by specifying typical operating conditions. If we consider that the terms in Equation (1) change little over the range of operating conditions for the meter they can be considered constant and we can calculate a maximum mass flowrate as,

{\dot{m}_{max}={N}{C_d}{E_v}{Y}{d^2}\sqrt{\rho_d}\sqrt{\Delta{P_{max}}}}     (2)

and, Equation (1) becomes simply,

\dot{m}_{d} = \dot{m}_{max}\sqrt{\dfrac{\Delta{P}}{\Delta{P}_{max}}}     (3)

where,

\dot{m}_{max} is the maximum mass flowrate (i.e. flowrate at full-scale pressure drop)
\rho_d is the vapor density, design conditions
\dot{m}_d is the mass flowrate, design conditions
\Delta{P} is the pressure drop
\Delta{P_{max}} is the full-scale pressure drop

Equation (3) is then implemented online but assumes that the vapor density differs little from the design value. Compensation for deviations in fluid properties from design to flowing conditions is described below.

Flow Compensation

Flow meters don’t operate at the conditions used for meter design. If suitable measurements are available, a compensation factor can be applied to compensate for changes in flowing density (i.e. fluids not at density \rho_d in Equation (2)) and improve meter accuracy.

The density of a vapor can be expressed as,

\rho=M\dfrac{ZRT}{P}     (4)

where,

\rho is vapor density
M is molecular weight
Z is ideal gas compressibility factor
R is the ideal gas constant
T is temperature (absolute)
P is pressure (absolute)

The density at flowing conditions can then be calculated as a deviation from design conditions as,

\rho_f=\rho_d\dfrac{M_f}{M_d}\dfrac{P_f}{P_d}\dfrac{Z_d}{Z_f}\dfrac{T_d}{T_f}     (5)

where,

\rho_f is the vapor density, flowing conditions
\rho_d is the vapor density, design conditions
M_f is molecular weight, flowing conditions
M_d is molecular weight, design conditions
P_f is operating pressure (absolute), flowing conditions
P_d is operating pressure (absolute), design conditions
T_f is operating temperature (absolute), flowing conditions
T_d is operating temperature (absolute), design conditions
Z_f is ideal gas compressibility factor, flowing conditions
Z_d is ideal gas compressibility factor, design conditions

Therefore, for orifice meters in vapor service configured to indicate mass flow, the compensated vapor mass flowrate is,

\dot{m}_f={K_f}{\dot{m}_d}     (6)

K_f=\sqrt{\dfrac{M_f}{M_d}\dfrac{P_f}{P_d}\dfrac{Z_d}{Z_f}\dfrac{T_d}{T_f}}     (7)

where,

\dot{m}_f is the mass flowrate, flowing conditions
\dot{m}_d is the mass flowrate, design conditions
K_f is the flow meter compensation factor

Recommendation

Flow meter compensation should be implemented to improve meter accuracy. This is especially true for meters that do not always operate at or near their design conditions. Measurements include fluid temperature, pressure, and composition but intermittent lab analysis and online estimators can also be used.

An extensive list of flow meter compensation factors (K_f) for various flow units and meter types in both liquid and vapor service is given in Table 1.

Table 1: Flow Meter Compensation Factors, K_f

where,

\bar\rho_d is the standard liquid density, design conditions
\bar\rho_f is the standard liquid density, flowing conditions

Note that there are eight (8) situations when flow compensation is never required (K_f=1.0).