As part of our refinery expansion, we must confirm the capacity of a schedule-80 crude-oil-desalter wastewater line. It's a 2,500-ft pipeline consisting of two heat exchangers and a mix of 3-in. and 4-in. pipe emptying into a slop-oil decanter operating at 5 psig. We're hoping that capacity can be increased by 25%. We collected some data using existing pressure gauges. An initial hydraulic study with software using equations from Crane Technical Paper No. 410 indicated there is an "extra" 65 psi when 125 gal/min of brine flows at 160 psig. However, another study done using different Crane-based software gave about-50-psig extra pressure. How could two programs produce such different results?
"Why is there a difference between two programs that both use the Darcy method?" This should be expanded to a more general question: "Why isn't my system working the way I think it should?" It doesn't matter whether you are comparing the results of two different programs, calculated results against system instrumentation, or how the system works now versus two years ago. The troubleshooting method is the same.
The first assumption I will make is that difference in pressure is due to a problem in building the model used by each program rather than the accuracy of the programs, especially since both use the Darcy method for calculating the pipeline head loss.
So let's start by seeing what system items could result in a large difference in head loss or pressure difference between what you expect and what you see.
The system is described as pipeline and heat exchangers, so let's concentrate on those two elements. Items that can affect the model are:
1. differences in pipe diameter — since the system includes both 3-in. and 4-in. pipe, the diameters used in the models may not match, which can significantly impact head loss;
2. schedule — while the system uses schedule-80 pipe, some programs default to schedule 40, again resulting in a large difference in head loss;
3. elevations — if those of system boundaries and junctions between pipelines aren't the same in the models, calculated pressure will vary;
4. head loss characteristics in the heat exchangers — if not modeled comparably, pressure values will differ; and
5. pipe roughness — while the value used normally doesn't have a major impact, for a 2,500-ft pipeline it could account for the difference.
The best way to troubleshoot a model is to start at one end, and compare the pressures and flow rates in each section of the system. For example, let's assume the inlet pressures in both models are 160 psig, and the flow rate through the system is 125 gal/min. Based on those input values each program should calculate the same head loss for the first section of pipeline. If the head losses differ, check the diameters, schedule and roughness until the results match. Next, if the outlet pressures differ between the two models, check the elevations. Repeat these two steps for each pipe segment in the system. With the heat exchangers, the inlet pressures should match, as should the calculated head loss and differential pressures. If not, check the head loss characteristics used.
Engineering programs provide a rapid way of performing tedious calculations but it is important that the person using the results has the ability to troubleshoot the system and model and to determine if the difference is caused by the model or the actual system.
Ray Hardee, chief engineer
Engineered Software, Inc., Lacey, Wash.
GET SOME DATA
I would start by collecting data. To understand the pipe system, two curves are necessary — the friction-flow curve and the pump head-flow curve. Other information will also be needed, including: the elevation profile; a detailed description of all pipe and fittings as an isometric or table; the pressure drops for the heat exchangers; the pipe schedule; and the condition of the pipe inside surface (it could be rusted or scaled). This information will be required to create a pressure drop versus flow rate curve.
If it is an existing pipeline, it probably employs a control valve. In the days of low energy costs, some companies specified the valve dP = 50% of the line friction loss for "better control." This could be one cause of the unexpectedly high pressure drop.
Art Krugler, president
Krugler Engineering Group, Whittier, Calif.
REDO THE MEASUREMENTS
"Some field data was collected?" Anytime you use field data, including those from wired instruments, the results will be conflicting and open to interpretation. The refinery may not even be in steady state; inflow of fresh water may not match outflows. A brine line like this tends to be the waste dump for the entire refinery — so establishing inflows into this network may be a challenge, especially when many streams run in flow-meter optional mode. No wonder the models are misleading. The best solution is to redo the measurements after the refinery has been running in steady state for a few days. Isolate as many stray flows as practical.
First, if you have the time, consider the anomaly of extra pressure. The explanation could come from either the calculation or the measurement.
The calculation in the models could differ for several reasons: 1) fittings are often unique to each model — Crane, Miller and others have developed friction coefficients (Ks) for fittings; 2) assuming the lengths are the same in each model, the roughness factor (ε) could differ — use commercial pipe with an ε = 0.0018 in., not hydraulically smooth with ε ≈ 0.0001 in. or corroded (iron) pipe with an ε = 0.02 in; 3) temperatures drop after an exchanger increases friction factor, viscosity, density and pressure drop — without taking this into account the pressure drop could be lower than actual; 4) some models can adjust for the effect of flow rate on pressure, leading to an obvious and embarrassing error if programmed badly; 5) look for common errors like using the wrong r/d ratio for elbows — 1.5 is correct for welded pipe; 6) not including Ks, or the right ones, for entrances and exits; and 7) Crane fitting Ks tend to underestimate pressure drop (there was an attempt to improve this approach in the 1980s but Crane is still used widely).
Now, let's evaluate sources of experimental errors: 1) relying on old field gauges could result in higher errors than the percent of full-scale allowed by ANSI; 2) during unsteady state, level control valves could be closing, yielding erroneous rates if used to establish flow by Q = Cv (ΔP/SG)½; 3) usually it's wise to use a single instrument for measuring pressures and a single instrument for measuring flows; 4) there are probably additional flow streams you can't locate in the pipe rack (most plant isometrics are a mess) — these streams could come from a known source or from a leak; and 5) an unexpected pressure drop such as a fouled heat exchanger, a spectacle blind or even misaligned gaskets.
If you don't have the option of a second test then, rather than chasing down streams, adjust the models to estimate the maximum flow rate through the pipeline. If you can't specify the source of the fouling, which is the most probable cause of the pressure drop, consider reducing the cross-sectional area and attributing the drop to scale. This approach meets the requirement of establishing the capacity of the pipeline and leaving it to the refinery to sort out its errant flows or fouling.
Dirk Willard, senior process engineer
Middough Consultants, Holland, Ohio
We neutralize a waste product with calcium carbonate prior to processing for disposal. The neutralization set-up consists of a weigh-feeder, twin-impeller agitator in a pressure vessel, recirculation pump, heat exchanger (cooler) and centrifugal exhaust fan. The weigh-feeder adds the carbonate to the baffled agitated tank. The reaction is complete when the spike in solution temperature dissipates and the volumes of hydrogen gas decrease. The process poses several problems: the slurry butterfly feed valve normally operates at 30% open but was designed to operate at 50%, leading to severe erosion; the pump often fails because of either gas blockage or suction fouling; the reaction frequently is incomplete, causing safety problems from the erupting hydrogen (the agitator is designed for about 0.5 hp/1,000 gallons with a impeller-diameter/tank (D/T) ratio of only about 0.4); the heat exchanger isn't performing well, although it checked out during commissioning. What can we do to make the process run smoothly? How might we limp by until the next available outage? Can you explain possible causes of the equipment failures?
Send us your comments, suggestions or solutions for this question by September 12, 2011. We'll include as many of them as possible in the October 2011 issue and all on www.chemicalprocessing.com. Send visuals — a sketch is fine. E-mail us at ProcessPuzzler@putman.net or mail to Process Puzzler, Chemical Processing, 555 W. Pierce Road, Suite 301, Itasca, IL 60143. Fax: (630) 467-1120. Please include your name, title, location and company affiliation in the response.
And, of course, if you have a process problem you'd like to pose to our readers, send it along and we'll be pleased to consider it for publication.