Jun-2016

How dynamic pressure can affect vacuum column readings

Computational fluid dynamics confirmed field measurements to explain a vacuum unit’s poor quality HVGO and was used to redesign the tower internals

Grant Niccun and Steve White
Process Consulting Services

Article Summary

A vacuum unit was producing HVGO with high end-point, high metals, and high Conradson carbon – all clear signs of feed entrainment. In an effort to diagnose the root cause, detailed pressure measurements were taken in the field. The pressure survey of the vacuum column flash zone and wash bed yielded some unexpected results. To explain the results, computational fluid dynamics (CFD) was used. Study of the CFD results clearly explained the reason for the unexpected measurements: dynamic pressure. In high velocity, low pressure systems, ignoring dynamic pressure can be dangerous. The fact that CFD confirmed the field measurements inspired confidence in using it to redesign the tower internals to prevent future HVGO quality issues.

A mystery in the field
In 2015, a large North American refiner was struggling to meet vacuum unit HVGO specifications after coming out of a turnaround. The HVGO product was often black and contained high levels of contaminants (Conradson carbon and metals) and high boiling point material. Figure 1 shows the TBP curve of the high end-point HVGO in this unit versus the HVGO TBP curve from the refinery’s other vacuum unit. The 9m diameter vacuum column was processing about 60000 b/d of reduced crude with flash zone conditions of 370°C and 23 mmHg. Poor HVGO quality is a clear indication that entrainment is passing through the wash section to the HVGO section.

The HVGO quality issues coincided with a scheduled pressure survey of the entire crude/vacuum unit. Could the vacuum column pressure measurements help precisely diagnose the root cause of the HVGO issues?

Here is a list of possible diagnoses and expected observations:
1. A blown-out wash bed would exhibit a differential pressure close to zero.
2. A coked wash bed would exhibit a differential pressure above the expected value. The higher the pressure drop, the more extensive the coking.
3. Poor wash oil and/or vapour distribution to the wash bed would not result in an abnormal differential pressure, unless the distribution was so flawed as to result in wash bed coking. The trough distributor design in this case was supported from the top of the packing, so any shift in the bed could easily disturb the distributor from an initially level position. Furthermore, the vapour inlet distributor was an unconventional design that did not intuitively seem like it would be effective.
4. Inadequate wash oil flow to the top of the wash bed would be obvious based on the flowmeter value, as the flowmeter was believed to be correct in this case.

In the field, we initially found accessible pressure taps above the HVGO draw tray (A) and near the bottom of the flash zone (B), as shown in Figure 2.

A precise DP measurement was performed by installing calibrated digital manometers at each pressure tap and taking several simultaneous readings. The pressure drop was calculated by subtraction of the two absolute pressure values. The total pressure drop across the wash collector, wash bed, and HVGO collector was thus found to be 4.0 mmHg. If, as a rule of thumb, the pressure drop across each collector tray was 0.5 mmHg, the wash bed DP would be 3.0 mmHg. As the unit had very recently started up, a wash bed pressure drop of 3.0 mmHg would indicate a high rate of fouling (coking). All-new packing would be needed, and the wash oil distributor and flow meter would need to be carefully inspected as one of these would be the root cause of the wash bed coking.

Based on the measurements described above, our initial diagnosis of wash bed coking seemed to be the obvious answer. As a double check, we scoured the tower and found another pressure tap (C) just above the slop wax collector tray (see Figure 3). By connecting the digital manometers to the flash zone (B) and this new point (C) between the slop wax collector and the wash bed, the DP across the slop wax collector was isolated. The pressure drop across the collector tray was found to be 3.1 mmHg – significantly higher than expected! Taking this measurement into account, the DP across the wash bed and HVGO collector tray would be reduced to 0.9 mmHg. Assume 0.5 mmHg of pressure drop for the HVGO collector tray, and the wash bed DP would be around 0.4 mmHg, indicating that it was relatively healthy.

Later, for reference, the predicted packed bed pressure drop was checked using Sulzer Sulcol. The predicted packed bed pressure drop with the appropriate vapour and liquid loads was 0.26 mmHg.

The extra set of measurements invalidated the initial diagnosis of rapid wash bed coking and raised new questions. Why was the measured collector tray DP so high? Was the tray plugged or damaged during turnaround? Was this tray producing significant flow maldistribution leading to the HVGO quality problem? Rather than a quick conclusion in the field, additional analysis in the office would be required to understand the situation.

Seeking understanding
Back in the office, the first task was to look closely at the design of the slop wax collector tray. The DP across a collector tray is a function of the open area and riser geometry, and can be roughly estimated using Equation 1:
            
ΔP = C * VR2 * ρV                       [1]
∆P = collector tray pressure drop
C = empirical coefficient
VR = riser velocity
ρV = vapour density

The riser open area was 25%, giving VR = 11.2 m/s. The calculated tray pressure drop was 0.25 mmHg – very far from the 3.1 mmHg differential that we measured in the field.

Equation 1 assumes an even velocity distribution below the collector tray. We hypothesised that the vacuum tower entrance geometry could create complex flow patterns and maldistribution below the collector tray, increasing the pressure drop. A good tool for such an evaluation is computational fluid dynamics (CFD), so a 3D model of the bottom section of the tower was constructed and subjected to CFD analysis.

CFD analysis
CFD analysis was performed in Ansys Fluent Version 16.1. The three-dimensional flow was modelled as incompressible with realisable k-ε for turbulence. A symmetry plane allowed for only half of the tower to be modelled, greatly reducing the computational expense. The packed bed was simulated as a porous zone to ensure that the flow through the top of the collector tray would behave realistically. Even with the advantage of symmetry, the final mesh contained approximately 4.7 million cells.